Recent advances in parameter-efficient transfer learning have demonstrated the utility of composing LoRA adapters from libraries of pretrained modules. However, most existing approaches rely on simple retrieval heuristics or uniform averaging, which overlook the latent structure of task relationships in representation space. We propose a new framework for adapter reuse that moves beyond retrieval, formulating adapter composition as a geometry-aware sparse reconstruction problem. Specifically, we represent each task by a latent prototype vector derived from the base model’s encoder and aim to approximate the target task prototype as a sparse linear combination of retrieved reference prototypes, under an L1-regularized optimization objective. The resulting combination weights are then used to blend the corresponding LoRA adapters, yielding a composite adapter tailored to the target task. This formulation not only preserves the local geometric structure of the task representation manifold, but also promotes interpretability and efficient reuse by selecting a minimal set of relevant adapters. We demonstrate the effectiveness of our approach across multiple domains—including medical image segmentation, medical report generation and image synthesis. Our results highlight the benefit of coupling retrieval with latent geometry-aware optimization for improved zero-shot generalization.

Addressing missing modalities is a critical challenge in multimodal brain tumor segmentation. Most existing approaches merely handle modality-incomplete inputs during inference, assuming a full set of modalities for all training samples. However, this unrealistic assumption limits the usage of abundant modality-incomplete data commonly observed in clinical practice. In this paper, we explore a more practical task of tackling missing modalities during both training and inference. We propose a universal model featuring robust modality reconstruction and prompt-guided modality adaptation. Our mask-reconstruction pre-training enables robust modality-invariant representation learning, during which we design a novel distribution approximation method that supervises the reconstruction of absent modalities without requiring full-modal training data. Afterwards, when adapting our model to the segmentation task, we introduce the complete-then-distill (CTD) paradigm, which first estimates missing modalities in training samples from the available ones, and then distills the knowledge from the reconstructed full-modal representations to enhance learning from modality-incomplete data. Moreover, we propose prompt-guided modality adaptation to personalize a subset of model parameters during CTD, enabling the model to adapt to each distinct modality input scenario by using prompts with rich visual-textual information. Extensive experiments on two brain tumor segmentation benchmarks show our method consistently surpasses previous state-of-the-art approaches under dual-stage missing modality settings across various missing ratios.

Semi-supervised learning (SSL) has achieved notable progress in medical image segmentation. To achieve effective SSL, a model needs to be able to efficiently learn from limited labeled data and effectively exploit knowledge from abundant unlabeled data. Recent developments in visual foundation models, such as the Segment Anything Model (SAM), have demonstrated remarkable adaptability with improved sample efficiency. To seamlessly harness foundation models in SSL, we propose a SAM-driven cross prompting framework with adaptive sampling and prompt consistency for semi-supervised medical image segmentation, named CPAC-SAM. Our method employs SAM's unique prompt design and innovates a cross prompting strategy within a dual-branch framework to automatically generate prompts and supervision across two decoder branches, enabling effective learning from both scarce labeled and valuable unlabeled data. To ensure the quality of prompts for unlabeled data and provide meaningful supervision in the cross prompting scheme, we propose an innovative prototype-guided grid sampling strategy with adaptive intervals to simultaneously improve the reliability of the prompt selection area and ensure both adequate prompt density and complete target coverage. We further design a novel prompt consistency regularization to reduce SAM's prompt sensitivity and to enhance the output invariance under different prompts. We validate our method on five medical image segmentation tasks, encompassing both 2D and 3D scenarios. The extensive experiments with different labeled-data ratios and modalities demonstrate the superiority of our proposed method over the state-of-the-art SSL methods, with more than 4.1% and 3.8% Dice improvement on the breast cancer segmentation task and left atrium segmentation task, respectively. Our code is available at: https://github.com/JuzhengMiao/CPAC-SAM.

Deep learning-based positron emission tomography (PET) image denoising offers the potential to reduce radiation exposure and scanning time by transforming low-count images into high-count equivalents. However, existing methods typically blur crucial details, leading to inaccurate lesion quantification. This paper proposes a lesion-perceived and quantification-consistent modulation (LeqMod) strategy for enhanced PET image denoising, via employing downstream lesion quantification analysis as auxiliary tools. The LeqMod is a plug-and-play design adaptable to a wide range of model architectures, modulating the sampling and optimization procedures of model training without adding any computational burden to the inference phase. Specifically, the LeqMod consists of two components, the lesion-perceived modulation (LeMod) and the multiscale quantification-consistent modulation (QuMod). The LeMod enhances lesion contrast and visibility by allocating higher sampling weights and stricter loss criteria to lesion-present samples determined by an auxiliary segmentation network than lesion-absent ones. The QuMod further emphasizes quantification accuracy for both the mean and maximum standardized uptake value ( ${\mathrm {SUV}}_{{\textit {mean}}}$ and ${\mathrm {SUV}}_{{\textit {max}}}$ ) across multiscale sub-regions throughout the entire image, thereby reducing biases of denoised results relative to high-count references. Experiments conducted on large PET datasets from multiple centers and vendors, and varying noise levels demonstrated the LeqMod efficacy across various denoising frameworks. Compared to frameworks without LeqMod, the integration of LeqMod reduces the lesion ${\mathrm {SUV}}_{{\textit {max}}}$ bias by 5.92% on average and increases the peak signal-to-noise ratio (PSNR) by 0.36 on average, when denoising images across participating sites. (Code is available at https://github.com/mhxiaaa/LeqMod_PET_denoising).

Semi-supervised learning (SSL) has greatly advanced 3D medical image segmentation by alleviating the need for intensive labeling by radiologists. While previous efforts focused on model-centric advancements, the emergence of foundational generalist models like the Segment Anything Model (SAM) is expected to reshape the SSL landscape. Although these generalists usually show performance gaps relative to previous specialists in medical imaging, they possess impressive zero-shot segmentation abilities with manual prompts. Thus, this capability could serve as "free lunch" for training specialists, offering future SSL a promising data-centric perspective, especially revolutionizing both pseudo and expert labeling strategies to enhance the data pool. In this regard, we propose the Generalist Model-driven Active Barely Supervised (GM-ABS) learning paradigm, for developing specialized 3D segmentation models under extremely limited (barely) annotation budgets, e.g., merely cross-labeling three slices per selected scan. In specific, building upon a basic mean-teacher SSL framework, GM-ABS modernizes the SSL paradigm with two key data-centric designs: (i) Specialist-generalist collaboration, where the in-training specialist leverages class-specific positional prompts derived from class prototypes to interact with the frozen class-agnostic generalist across multiple views to achieve noisy-yet-effective label augmentation. Then, the specialist robustly assimilates the augmented knowledge via noise-tolerant collaborative learning. (ii) Expert-model collaboration that promotes active cross-labeling with notably low labeling efforts. This design progressively furnishes the specialist with informative and efficient supervision via a human-in-the-loop manner, which in turn benefits the quality of class-specific prompts. Extensive experiments on three benchmark datasets highlight the promising performance of GM-ABS over recent SSL approaches under extremely constrained labeling resources.

Large Language Models (LLMs) have shown remarkable capabilities in general domains but often struggle with tasks requiring specialized knowledge. Conventional Retrieval-Augmented Generation (RAG) techniques typically retrieve external information from static knowledge bases, which can be outdated or incomplete, missing fine-grained clinical details essential for accurate medical question answering. In this work, we propose SearchRAG, a novel framework that overcomes these limitations by leveraging real-time search engines. Our method employs synthetic query generation to convert complex medical questions into search-engine-friendly queries and utilizes uncertainty-based knowledge selection to filter and incorporate the most relevant and informative medical knowledge into the LLM's input. Experimental results demonstrate that our method significantly improves response accuracy in medical question answering tasks, particularly for complex questions requiring detailed and up-to-date knowledge. We provide our code here11https://github.com/sycny/SearchRAG.

Echocardiography (ECHO) is essential for cardiac assessments, but its video quality and interpretation heavily relies on manual expertise, leading to inconsistent results from clinical and portable devices. ECHO video generation offers a solution by improving automated monitoring through synthetic data and generating high-quality videos from routine health data. However, existing models often face high computational costs, slow inference, and rely on complex conditional prompts that require experts' annotations. To address these challenges, we propose ECHOPulse, an ECG-conditioned ECHO video generation model. ECHOPulse introduces two key advancements: (1) it accelerates ECHO video generation by leveraging VQ-VAE tokenization and masked visual token modeling for fast decoding, and (2) it conditions on readily accessible ECG signals, which are highly coherent with ECHO videos, bypassing complex conditional prompts. To the best of our knowledge, this is the first work to use time-series prompts like ECG signals for ECHO video generation. ECHOPulse not only enables controllable synthetic ECHO data generation but also provides updated cardiac function information for disease monitoring and prediction beyond ECG alone. Evaluations on three public and private datasets demonstrate state-of-the-art performance in ECHO video generation across both qualitative and quantitative measures. Additionally, ECHOPulse can be easily generalized to other modality generation tasks, such as cardiac MRI, fMRI, and 3D CT generation. We will make the synthetic ECHO dataset, along with the code and model, publicly available upon acceptance.

Large Multimodal Models (LMMs), or Vision-Language Models (VLMs), have shown impressive capabilities in a wide range of visual tasks. However, they often struggle with fine-grained visual reasoning, failing to identify domain-specific objectives and provide justifiable explanations for their predictions. To address the above challenge, we propose a novel visual rejection sampling framework to improve the cognition and explainability of LMMs using self-synthesized data. Specifically, visual fine-tuning requires images, queries, and target answers. Our approach begins by synthesizing interpretable answers that include human-verifiable visual features. These features are based on expert-defined concepts, and carefully selected based on their alignment with the image content. After each round of fine-tuning, we apply a reward model-free filtering mechanism to select the highest-quality interpretable answers for the next round of tuning. This iterative process of synthetic data generation and fine-tuning progressively improves the model's ability to generate accurate and reasonable explanations. Experimental results demonstrate the effectiveness of our method in improving both the accuracy and explainability of specialized visual classification tasks.

Ensuring fairness in medical image segmentation is critical due to biases in imbalanced clinical data acquisition caused by demographic attributes (e.g., age, sex, race) and clinical factors (e.g., disease severity). To address these challenges, we introduce Distribution-aware Mixture of Experts (dMoE), inspired by optimal control theory. We provide a comprehensive analysis of its underlying mechanisms and clarify dMoE's role in adapting to heterogeneous distributions in medical image segmentation. Furthermore, we integrate dMoE into multiple network architectures, demonstrating its broad applicability across diverse medical image analysis tasks. By incorporating demographic and clinical factors, dMoE achieves state-of-the-art performance on two 2D benchmark datasets and a 3D in-house dataset. Our results highlight the effectiveness of dMoE in mitigating biases from imbalanced distributions, offering a promising approach to bridging control theory and medical image segmentation within fairness learning paradigms. The source code is available at https://github.com/tvseg/dMoE.

Solving inverse problems—recovering signals from incomplete or noisy measurements—is fundamental in science and engineering. Score-based generative models (SGMs) have recently emerged as a powerful framework for this task. Two main paradigms have formed: unsupervised approaches that adapt pretrained generative models to inverse problems, and supervised bridge methods that train stochastic processes conditioned on paired clean and corrupted data. While the former typically assume knowledge of the measurement model, the latter have largely overlooked this structural information. We introduce System-embedded Diffusion Bridge Models (SDBs), a new class of supervised bridge methods that explicitly embed the known linear measurement system into the coefficients of a matrix-valued SDE. This principled integration yields consistent improvements across diverse linear inverse problems and demonstrates robust generalization under system misspecification between training and deployment, offering a promising solution to real-world applications.

Diffusion models have achieved state-of-the-art performance in generative modeling, yet their sampling procedures remain vulnerable to hallucinations—often stemming from inaccuracies in score approximation. In this work, we reinterpret diffusion sampling through the lens of optimization and introduce RODS (Robust Optimization–inspired Diffusion Sampler), a novel method that detects and corrects high-risk sampling steps using geometric cues from the loss landscape. RODS enforces smoother sampling trajectories and \textit{adaptively} adjusts perturbations, reducing hallucinations without retraining and at minimal additional inference cost. Experiments on AFHQv2, FFHQ, and 11k-hands demonstrate that RODS maintains comparable image quality and preserves generation diversity. More importantly, it improves both sampling fidelity and robustness, detecting over 70\% of hallucinated samples and correcting more than 25\%, all while avoiding the introduction of new artifacts. We release our code at https://github.com/Yiqi-Verna-Tian/RODS.

Text data augmentation is an effective strategy for overcoming the challenge of limited sample sizes in many natural language processing (NLP) tasks. This challenge is especially prominent in the few-shot learning (FSL) scenario, where the data in the target domain is generally much scarcer and of lowered quality. A natural and widely used strategy to mitigate such challenges is to perform data augmentation to better capture data invariance and increase the sample size. However, current text data augmentation methods either can’t ensure the correct labeling of the generated data (lacking faithfulness), or can’t ensure sufficient diversity in the generated data (lacking compactness), or both. Inspired by the recent success of large language models (LLM), especially the development of ChatGPT, we propose a text data augmentation approach based on ChatGPT (named ”AugGPT”). AugGPT rephrases each sentence in the training samples into multiple conceptually similar but semantically different samples. The augmented samples can then be used in downstream model training. Experiment results on multiple few-shot learning text classification tasks show the superior performance of the proposed AugGPT approach over state-of-the-art text data augmentation methods in terms of testing accuracy and distribution of the augmented samples.

Recently, ChatGPT and GPT-4 have emerged and gained immense global attention due to their unparalleled performance in language processing. Despite demonstrating impressive capability in various open-domain tasks, their adequacy in highly specific fields like radiology remains untested. Radiology presents unique linguistic phenomena distinct from open-domain data due to its specificity and complexity. Assessing the performance of large language models (LLMs) in such specific domains is crucial not only for a thorough evaluation of their overall performance but also for providing valuable insights into future model design directions: whether model design should be generic or domain-specific. To this end, in this study, we evaluate the performance of ChatGPT/GPT-4 on a radiology natural language inference (NLI) task and compare it to other models fine-tuned specifically on task-related data samples. We also conduct a comprehensive investigation on ChatGPT/GPT-4’s reasoning ability by introducing varying levels of inference difficulty. Our results show that 1) ChatGPT and GPT-4 outperform other LLMs in the radiology NLI task and 2) other specifically fine-tuned Bert-based models require significant amounts of data samples to achieve comparable performance to ChatGPT/GPT-4. These findings not only demonstrate the feasibility and promise of constructing a generic model capable of addressing various tasks across different domains, but also highlight several key factors crucial for developing a unified model, particularly in a medical context, paving the way for future artificial general intelligence (AGI) systems. We release our code and data to the research community.

Despite achieving impressive results in general-purpose semantic segmentation with strong generalization on natural images, the Segment Anything Model (SAM) has shown less precision and stability in medical image segmentation. In particular, the original SAM architecture is designed for 2D natural images and is therefore not support to handle three-dimensional information, which is particularly important for medical imaging modalities that are often volumetric or video data. In this paper, we introduce MediViSTA, a parameter-efficient fine-tuning method designed to adapt the vision foundation model for medical video, with a specific focus on echocardiography segmentation. To achieve spatial adaptation, we propose a frequency feature fusion technique that injects spatial frequency information from a CNN branch. For temporal adaptation, we integrate temporal adapters within the transformer blocks of the image encoder. Using a fine-tuning strategy, only a small subset of pre-trained parameters is updated, allowing efficient adaptation to echocardiography data. The effectiveness of our method has been comprehensively evaluated on three datasets, comprising two public datasets and one multi-center in-house dataset. Our method consistently outperforms various state-of-the-art approaches without using any prompts. Furthermore, our model exhibits strong generalization capabilities on unseen datasets, surpassing the second-best approach by 2.15% in Dice and 0.09 in temporal consistency. The results demonstrate the potential of MediViSTA to significantly advance echocardiography video segmentation, offering improved accuracy and robustness in cardiac assessment applications.

Echocardiography is the first-line non-invasive cardiac imaging modality, providing rich spatio-temporal information on cardiac anatomy and physiology. Recently, foundation model trained on extensive and diverse datasets has shown strong performance in various downstream tasks. However, translating foundation models into the medical imaging domain remains challenging due to domain differences between medical and natural images, the lack of diverse patient and disease datasets. In this paper, we introduce EchoFM, a general-purpose vision foundation model for echocardiography trained on a large-scale dataset of over 20 million echocardiographic images from 6,500 patients. To enable effective learning of rich spatio-temporal representations from periodic videos, we propose a novel self-supervised learning framework based on a masked autoencoder with a spatio-temporal consistent masking strategy and periodic-driven contrastive learning. The learned cardiac representations can be readily adapted and fine-tuned for a wide range of downstream tasks, serving as a strong and flexible backbone model. We validate EchoFM through experiments across key downstream tasks in the clinical echocardiography workflow, leveraging public and multi-center internal datasets. EchoFM consistently outperforms SOTA methods, demonstrating superior generalization capabilities and flexibility. The code and checkpoints are available at: https://github.com/SekeunKim/EchoFM.git.

Logs are ubiquitous digital footprints, playing an indispensable role in system diagnostics, security analysis, and performance optimization. The extraction of actionable insights from logs is critically dependent on the log parsing process, which converts raw logs into structured formats for downstream analysis. Yet, the complexities of contemporary systems and the dynamic nature of logs pose significant challenges to existing automatic parsing techniques. The emergence of Large Language Models (LLM) offers new horizons. With their expansive knowledge and contextual prowess, LLMs have been transformative across diverse applications. Building on this, we introduce LogParser-LLM, a novel log parser integrated with LLM capabilities. This union seamlessly blends semantic insights with statistical nuances, obviating the need for hyper-parameter tuning and labeled training data, while ensuring rapid adaptability through online parsing. Further deepening our exploration, we address the intricate challenge of parsing granularity, proposing a new metric and integrating human interactions to allow users to calibrate granularity to their specific needs. Our method's efficacy is empirically demonstrated through evaluations on the Loghub-2k and the large-scale LogPub benchmark. In evaluations on the LogPub benchmark, involving an average of 3.6 million logs per dataset across 14 datasets, our LogParser-LLM requires only 272.5 LLM invocations on average, achieving a 90.6% F1 score for grouping accuracy and an 81.1% for parsing accuracy. These results demonstrate the method's high efficiency and accuracy, outperforming current state-of-the-art log parsers, including pattern-based, neural network-based, and existing LLM-enhanced approaches.

Recent advancements in multimodal foundation models have showcased impressive capabilities in understanding and reasoning with visual and textual information. Adapting these foundation models trained for general usage to specialized domains like biomedicine requires large-scale domain-specific instruction datasets. While existing works have explored curating such datasets automatically, the resultant datasets are not explicitly aligned with domain expertise. In this work, we propose a data-centric framework, Biomedical Visual Instruction Tuning with Clinician Preference Alignment (BioMed-VITAL), that incorporates clinician preferences into both stages of generating and selecting instruction data for tuning biomedical multimodal foundation models. First, during the generation stage, we prompt the GPT-4V generator with a diverse set of clinician-selected demonstrations for preference-aligned data candidate generation. Then, during the selection phase, we train a separate selection model, which explicitly distills clinician and policy-guided model preferences into a rating function to select high-quality data for medical instruction tuning. Results show that the model tuned with the instruction-following data from our method demonstrates a significant improvement in open visual chat (18.5% relatively) and medical VQA (win rate up to 81.73%). Our instruction-following data and models are available at https://BioMed-VITAL.github.io.

This paper aims to develop a Newton-type method to solve a class of nonconvex composite programs. In particular, the nonsmooth part is possibly nonconvex. To tackle the nonconvexity, we develop a notion of strong prox-regularity which is related to the singleton property and Lipschitz continuity of the associated proximal operator, and we verify it in various classes of functions, including weakly convex functions, indicator functions of proximally smooth sets, and two specific sphere-related nonconvex nonsmooth functions. In this case, the problem class we are concerned with covers smooth optimization problems on manifold and certain composite optimization problems on manifold. For the latter, the proposed algorithm is the first second-order type method. Combining with the semismoothness of the proximal operator, we design a projected semismooth Newton method to find a root of the natural residual induced by the proximal gradient method. Due to the possible nonconvexity of the feasible domain, an extra projection is added to the usual semismooth Newton step and new criteria are proposed for the switching between the projected semismooth Newton step and the proximal step. The global convergence is then established under the strong prox-regularity. Based on the BD regularity condition, we establish local superlinear convergence. Numerical experiments demonstrate the effectiveness of our proposed method compared with state-of-the-art ones.

Position emission tomography (PET) is widely used in clinics and research due to its quantitative merits and high sensitivity, but suffers from low signal-to-noise ratio (SNR). Recently convolutional neural networks (CNNs) have been widely used to improve PET image quality. Though successful and efficient in local feature extraction, CNN cannot capture long-range dependencies well due to its limited receptive field. Global multi-head self-attention (MSA) is a popular approach to capture long-range information. However, the calculation of global MSA for 3D images has high computational costs. In this work, we proposed an efficient spatial and channel-wise encoder-decoder transformer, Spach Transformer, that can leverage spatial and channel information based on local and global MSAs. Experiments based on datasets of different PET tracers, i.e., 18F-FDG, 18F-ACBC, 18F-DCFPyL, and 68Ga-DOTATATE, were conducted to evaluate the proposed framework. Quantitative results show that the proposed Spach Transformer framework outperforms state-of-the-art deep learning architectures.

To address the lack of high-quality training labels in positron emission tomography (PET) imaging, weakly-supervised reconstruction methods that generate network-based mappings between prior images and noisy targets have been developed. However, the learned model has an intrinsic variance proportional to the average variance of the target image. To suppress noise and improve the accuracy and generalizability of the learned model, we propose a conditional weakly-supervised multi-task learning (MTL) strategy, in which an auxiliary task is introduced serving as an anatomical regularizer for the PET reconstruction main task. In the proposed MTL approach, we devise a novel multi-channel self-attention (MCSA) module that helps learn an optimal combination of shared and task-specific features by capturing both local and global channel-spatial dependencies. The proposed reconstruction method was evaluated on NEMA phantom PET datasets acquired at different positions in a PET/CT scanner and 26 clinical whole-body PET datasets. The phantom results demonstrate that our method outperforms state-of-the-art learning-free and weakly-supervised approaches obtaining the best noise/contrast tradeoff with a significant noise reduction of approximately 50.0% relative to the maximum likelihood (ML) reconstruction. The patient study results demonstrate that our method achieves the largest noise reductions of 67.3% and 35.5% in the liver and lung, respectively, as well as consistently small biases in 8 tumors with various volumes and intensities. In addition, network visualization reveals that adding the auxiliary task introduces more anatomical information into PET reconstruction than adding only the anatomical loss, and the developed MCSA can abstract features and retain PET image details.

The knowledge graph (KG) is a highly needed basis to support the high-fidelity and high-interpretability modeling of various tasks in healthcare artificial intelligence. In this work, we focus on constructing an oncology knowledge graph that will be used in downstream cancer research and solution development. Modern supervised learning for knowledge graph construction requires a large amount of manually labeled data, which makes the process time-consuming and labor-intensive. Although there exists multiple research on named entity recognition and relation extraction based on distantly supervised learning, constructing a domain-specific knowledge graph from large collections of textual data without manual annotations is still an urgent problem to be solved. In response, we propose an integrated framework for adapting and re-learning knowledge graphs from a general domain (biomedical in our case) to a fine-defined domain (oncology). In this framework, we apply distant-supervision on cross-domain knowledge graph adaptation. Consequently, no manual data annotation is required to train the model. We introduce a novel iterative training strategy to facilitate the discovery of domain-specific named entities and triplets. Experimental results indicate that the proposed framework can perform domain adaptation and construction of knowledge graphs efficiently.

Though achieving excellent performance in some cases, current unsupervised learning methods for single image denoising usually have constraints in applications. In this paper, we propose a new approach which is more general and applicable to complicated noise models. Utilizing the property of score function, the gradient of logarithmic probability, we define a solving system for denoising. Once the score function of noisy images has been estimated, the denoised result can be obtained through the solving system. Our approach can be applied to multiple noise models, such as the mixture of multiplicative and additive noise combined with structured correlation. Experimental results show that our method is comparable when the noise model is simple, and has good performance in complicated cases where other methods are not applicable or perform poorly.

Federated Learning (FL) could solve the challenges of training AI models on large datasets for medical imaging due to data privacy and ownership concerns by allowing collaborative training without the need for sharing raw data. This Special Issue on Federated Learning for Medical Imaging features papers covering FL-related topics and discussing their implications for healthcare and medical imaging. The included articles focus on a broad range of federated scenarios and applications, such as semi-supervised and self-supervised learning, histopathology, image reconstruction, graph neural networks, privacy preservation, active learning, data auditing, multi-task learning, personalization, and swarm learning. The importance of training unbiased, privacy-preserving, and generalizable AI models that have the potential to be translated into clinical practice increases the need for collaborative training techniques such as FL. The articles included in this Special Issue have moved the needle markedly forward in this regard.

Direct reconstruction methods have been developed to estimate parametric images directly from the measured PET sinograms by combining the PET imaging model and tracer kinetics in an integrated framework. Due to limited counts received, signal-to-noise-ratio (SNR) and resolution of parametric images produced by direct reconstruction frameworks are still limited. Recently supervised deep learning methods have been successfully applied to medical imaging denoising/reconstruction when large number of high-quality training labels are available. For static PET imaging, high-quality training labels can be acquired by extending the scanning time. However, this is not feasible for dynamic PET imaging, where the scanning time is already long enough. In this work, we proposed an unsupervised deep learning framework for direct parametric reconstruction from dynamic PET, which was tested on the Patlak model and the relative equilibrium Logan model. The training objective function was based on the PET statistical model. The patient’s anatomical prior image, which is readily available from PET/CT or PET/MR scans, was supplied as the network input to provide a manifold constraint, and also utilized to construct a kernel layer to perform non-local feature denoising. The linear kinetic model was embedded in the network structure as a${1} \times {1} \times {1}$convolution layer. Evaluations based on dynamic datasets of18F-FDG and11C-PiB tracers show that the proposed framework can outperform the traditional and the kernel method-based direct reconstruction methods.

Automatic quantification of the left ventricle (LV) from cardiac magnetic resonance (CMR) images plays an important role in making the diagnosis procedure efficient, reliable, and alleviating the laborious reading work for physicians. Considerable efforts have been devoted to LV quantification using different strategies that include segmentation-based (SG) methods and the recent direct regression (DR) methods. Although both SG and DR methods have obtained great success for the task, a systematic platform to benchmark them remains absent because of differences in label information during model learning. In this paper, we conducted an unbiased evaluation and comparison of cardiac LV quantification methods that were submitted to the Left Ventricle Quantification (LVQuan) challenge, which was held in conjunction with the Statistical Atlases and Computational Modeling of the Heart (STACOM) workshop at the MICCAI 2018. The challenge was targeted at the quantification of 1) areas of LV cavity and myocardium, 2) dimensions of the LV cavity, 3) regional wall thicknesses (RWT), and 4) the cardiac phase, from mid-ventricle short-axis CMR images. First, we constructed a public quantification dataset Cardiac-DIG with ground truth labels for both the myocardium mask and these quantification targets across the entire cardiac cycle. Then, the key techniques employed by each submission were described. Next, quantitative validation of these submissions were conducted with the constructed dataset. The evaluation results revealed that both SG and DR methods can offer good LV quantification performance, even though DR methods do not require densely labeled masks for supervision. Among the 12 submissions, the DR method LDAMT offered the best performance, with a mean estimation error of 301 mm2for the two areas, 2.15 mm for the cavity dimensions, 2.03 mm for RWTs, and a 9.5% error rate for the cardiac phase classification. Three of the SG methods also delivered comparable performances. Finally, we discussed the advantages and disadvantages of SG and DR methods, as well as the unsolved problems in automatic cardiac quantification for clinical practice applications.

Cell detection and cell type classification from biomedical images play an important role for high-throughput imaging and various clinical application. While classification of single cell sample can be performed with standard computer vision and machine learning methods, analysis of multi-label samples (region containing congregating cells) is more challenging, as separation of individual cells can be difficult (e.g. touching cells) or even impossible (e.g. overlapping cells). As multi-instance images are common in analyzing Red Blood Cell (RBC) for Sickle Cell Disease (SCD) diagnosis, we develop and implement a multi-instance cell detection and classification framework to address this challenge. The framework firstly trains a region proposal model based on Region-based Convolutional Network (RCNN) to obtain bounding-boxes of regions potentially containing single or multiple cells from input microscopic images, which are extracted as image patches. High-level image features are then calculated from image patches through a pre-trained Convolutional Neural Network (CNN) with ResNet-50 structure. Using these image features inputs, six networks are then trained to make multi-label prediction of whether a given patch contains cells belonging to a specific cell type. As the six networks are trained with image patches consisting of both individual cells and touching/overlapping cells, they can effectively recognize cell types that are presented in multi-instance image samples. Finally, for the purpose of SCD testing, we train another machine learning classifier to predict whether the given image patch contains abnormal cell type based on outputs from the six networks. Testing result of the proposed framework shows that it can achieve good performance in automatic cell detection and classification.

Missing value imputation is a challenging and well- researched topic in data mining. In this paper, we propose IFGAN, a missing value imputation algorithm based on Feature- specific Generative Adversarial Networks (GAN). Our idea is intuitive yet effective: a feature-specific generator is trained to impute missing values, while a discriminator is expected to distinguish the imputed values from observed ones. The proposed architecture is capable of handling different data types, data distributions, missing mechanisms, and missing rates. It also improves post-imputation analysis by preserving inter-feature correlations. We empirically show on several real-life datasets that IFGAN outperforms current state-of-the-art algorithm under various missing conditions.

Identifying heterogeneous cognitive impairment markers at an early stage is vital for Alzheimer's disease diagnosis. However, due to complex and uncertain brain connectivity features in the cognitive domains, it remains challenging to quantify functional brain connectomic changes during non-pharmacological interventions for amnestic mild cognitive impairment (aMCI) patients. We present a quantitative method for functional brain network analysis of fMRI data based on the multi-graph unsupervised Gaussian embedding method (MG2G). This neural network-based model can effectively learn low-dimensional Gaussian distributions from the original high-dimensional sparse functional brain networks, quantify uncertainties in link prediction, and discover the intrinsic dimensionality of brain networks. Using the Wasserstein distance to measure probabilistic changes, we discovered that brain regions in the default mode network and somatosensory/somatomotor hand, fronto-parietal task control, memory retrieval, and visual and dorsal attention systems had relatively large variations during non-pharmacological training, which might provide distinct biomarkers for fine-grained monitoring of aMCI cognitive alteration. An important finding of our study is the ability of the new method to capture subtle changes for individual patients before and after short-term intervention. More broadly, the MG2G method can be used in studying multiple brain disorders and injuries, e.g., in Parkinson's disease or traumatic brain injury (TBI), and hence it will be useful to the wider neuroscience community.

Early and accurate diagnosis of Coronavirus disease (COVID-19) is essential for patient isolation and contact tracing so that the spread of infection can be limited. Computed tomography (CT) can provide important information in COVID-19, especially for patients with moderate to severe disease as well as those with worsening cardiopulmonary status. As an automatic tool, deep learning methods can be utilized to perform semantic segmentation of affected lung regions, which is important to establish disease severity and prognosis prediction. Both the extent and type of pulmonary opacities help assess disease severity. However, manually pixel-level multi-class labelling is time-consuming, subjective, and non-quantitative. In this article, we proposed a hybrid weak label-based deep learning method that utilize both the manually annotated pulmonary opacities from COVID-19 pneumonia and the patient-level disease-type information available from the clinical report. A UNet was firstly trained with semantic labels to segment the total infected region. It was used to initialize another UNet, which was trained to segment the consolidations with patient-level information using the Expectation-Maximization (EM) algorithm. To demonstrate the performance of the proposed method, multi-institutional CT datasets from Iran, Italy, South Korea, and the United States were utilized. Results show that our proposed method can predict the infected regions as well as the consolidation regions with good correlation to human annotation.

This study aims to develop an automatic classifier based on deep learning for exacerbation frequency in patients with chronic obstructive pulmonary disease (COPD). A three-layer deep belief network (DBN) with two hidden layers and one visible layer was employed to develop classification models and the models' robustness to exacerbation was analyzed. Subjects from the COPDGene cohort were labeled with exacerbation frequency, defined as the number of exacerbation events per year. A total of 10 300 subjects with 361 features each were included in the analysis. After feature selection and parameter optimization, the proposed classification method achieved an accuracy of 91.99%, using a ten-fold cross validation experiment. The analysis of DBN weights showed that there was a good visual spatial relationship between the underlying critical features of different layers. Our findings show that the most sensitive features obtained from the DBN weights are consistent with the consensus showed by clinical rules and standards for COPD diagnostics. We, thus, demonstrate that DBN is a competitive tool for exacerbation risk assessment for patients suffering from COPD.

Red blood cell (RBC) segmentation and classification from microscopic images is a crucial step for the diagnosis of sickle cell disease (SCD). In this work, we adopt a deep learning based semantic segmentation framework to solve the RBC classification task. A major challenge for robust segmentation and classification is the large variations on the size, shape and viewpoint of the cells, combining with the low image quality caused by noise and artifacts. To address these challenges, we apply deformable convolution layers to the classic U-Net structure and implement the deformable U-Net (dU-Net). U-Net architecture has been shown to offer accurate localization for image semantic segmentation. Moreover, deformable convolution enables free-form deformation of the feature learning process, thus making the network more robust to various cell morphologies and image settings. dU-Net is tested on microscopic red blood cell images from patients with sickle cell disease. Results show that dU-Net can achieve highest accuracy for both binary segmentation and multi-class semantic segmentation tasks, comparing with both unsupervised and state-of-the-art deep learning based supervised segmentation methods. Through detailed investigation of the segmentation results, we further conclude that the performance improvement is mainly caused by the deformable convolution layer, which has better ability to separate the touching cells, discriminate the background noise and predict correct cell shapes without any shape priors.

Imaging-based early diagnosis of Alzheimer Disease (AD) has become an effective approach, especially by using nuclear medicine imaging techniques such as Positron Emission Topography (PET). In various literature it has been found that PET images can be better modeled as signals (e.g. uptake of florbetapir) defined on a network (non-Euclidean) structure which is governed by its underlying graph patterns of pathological progression and metabolic connectivity. In order to effectively apply deep learning framework for PET image analysis to overcome its limitation on Euclidean grid, we develop a solution for 3D PET image representation and analysis under a generalized, graph-based CNN architecture (PETNet), which analyzes PET signals defined on a group-wise inferred graph structure. Computations in PETNet are defined in non-Euclidean, graph (network) domain, as it performs feature extraction by convolution operations on spectral-filtered signals on the graph and pooling operations based on hierarchical graph clustering. Effectiveness of the PETNet is evaluated on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, which shows improved performance over both deep learning and other machine learning-based methods.

Computerized healthcare has undergone rapid development thanks to the advances in medical imaging and machine learning technologies. Especially, recent progress on deep learning opens a new era for multimedia based clinical decision support. In this paper, we use deep learning with brain network and clinical relevant text information to make early diagnosis of Alzheimer's Disease (AD). The clinical relevant text information includes age, gender, and ApoE gene of the subject. The brain network is constructed by computing the functional connectivity of brain regions using resting-state functional magnetic resonance imaging (R-fMRI) data. A targeted autoencoder network is built to distinguish normal aging from mild cognitive impairment, an early stage of AD. The proposed method reveals discriminative brain network features effectively and provides a reliable classifier for AD detection. Compared to traditional classifiers based on R-fMRI time series data, about 31.21 percent improvement of the prediction accuracy is achieved by the proposed deep learning method, and the standard deviation reduces by 51.23 percent in the best case that means our prediction model is more stable and reliable compared to the traditional methods. Our work excavates deep learning's advantages of classifying high-dimensional multimedia data in medical services, and could help predict and prevent AD at an early stage.

Automated detection of cancer metastases in lymph nodes has the potential to improve the assessment of prognosis for patients. To enable fair comparison between the algorithms for this purpose, we set up the CAMELYON17 challenge in conjunction with the IEEE International Symposium on Biomedical Imaging 2017 Conference in Melbourne. Over 300 participants registered on the challenge website, of which 23 teams submitted a total of 37 algorithms before the initial deadline. Participants were provided with 899 whole-slide images (WSIs) for developing their algorithms. The developed algorithms were evaluated based on the test set encompassing 100 patients and 500 WSIs. The evaluation metric used was a quadratic weighted Cohen's kappa. We discuss the algorithmic details of the 10 best pre-conference and two post-conference submissions. All these participants used convolutional neural networks in combination with pre- and postprocessing steps. Algorithms differed mostly in neural network architecture, training strategy, and pre- and postprocessing methodology. Overall, the kappa metric ranged from 0.89 to -0.13 across all submissions. The best results were obtained with pre-trained architectures such as ResNet. Confusion matrix analysis revealed that all participants struggled with reliably identifying isolated tumor cells, the smallest type of metastasis, with detection rates below 40%. Qualitative inspection of the results of the top participants showed categories of false positives, such as nerves or contamination, which could be targeted for further optimization. Last, we show that simple combinations of the top algorithms result in higher kappa metric values than any algorithm individually, with 0.93 for the best combination.

Recently, deep neural networks have been widely and successfully applied in computer vision tasks and have attracted growing interest in medical imaging. One barrier for the application of deep neural networks to medical imaging is the need for large amounts of prior training pairs, which is not always feasible in clinical practice. This is especially true for medical image reconstruction problems, where raw data are needed. Inspired by the deep image prior framework, in this paper, we proposed a personalized network training method where no prior training pairs are needed, but only the patient's own prior information. The network is updated during the iterative reconstruction process using the patient-specific prior information and measured data. We formulated the maximum-likelihood estimation as a constrained optimization problem and solved it using the alternating direction method of multipliers algorithm. Magnetic resonance imaging guided positron emission tomography reconstruction was employed as an example to demonstrate the effectiveness of the proposed framework. Quantification results based on simulation and real data show that the proposed reconstruction framework can outperform Gaussian post-smoothing and anatomically guided reconstructions using the kernel method or the neural-network penalty.

PET image reconstruction is challenging due to the ill-poseness of the inverse problem and limited number of detected photons. Recently, the deep neural networks have been widely and successfully used in computer vision tasks and attracted growing interests in medical imaging. In this paper, we trained a deep residual convolutional neural network to improve PET image quality by using the existing inter-patient information. An innovative feature of the proposed method is that we embed the neural network in the iterative reconstruction framework for image representation, rather than using it as a post-processing tool. We formulate the objective function as a constrained optimization problem and solve it using the alternating direction method of multipliers algorithm. Both simulation data and hybrid real data are used to evaluate the proposed method. Quantification results show that our proposed iterative neural network method can outperform the neural network denoising and conventional penalized maximum likelihood methods.

Deep neural networks have become the state-of-the-art models in numerous machine learning tasks. However, general guidance to network architecture design is still missing. In our work, we bridge deep neural network design with numerical differential equations. We show that many effective networks, such as ResNet, PolyNet, FractalNet and RevNet, can be interpreted as different numerical discretizations of differential equations. This finding brings us a brand new perspective on the design of effective deep architectures. We can take advantage of the rich knowledge in numerical analysis to guide us in designing new and potentially more effective deep networks. As an example, we propose a linear multi-step architecture (LM-architecture) which is inspired by the linear multi-step method solving ordinary differential equations. The LM-architecture is an effective structure that can be used on any ResNet-like networks. In particular, we demonstrate that LM-ResNet and LM-ResNeXt (i.e. the networks obtained by applying the LM-architecture on ResNet and ResNeXt respectively) can achieve noticeably higher accuracy than ResNet and ResNeXt on both CIFAR and ImageNet with comparable numbers of trainable parameters. In particular, on both CIFAR and ImageNet, LM-ResNet/LM-ResNeXt can significantly compress (>50%) the original networks while maintaining a similar performance. This can be explained mathematically using the concept of modified equation from numerical analysis. Last but not least, we also establish a connection between stochastic control and noise injection in the training process which helps to improve generalization of the networks. Furthermore, by relating stochastic training strategy with stochastic dynamic system, we can easily apply stochastic training to the networks with the LM-architecture. As an example, we introduced stochastic depth to LM-ResNet and achieve significant improvement over the original LM-ResNet on CIFAR10.

Motivated by the great potential of deep learning in medical imaging, we propose an iterative positron emission tomography reconstruction framework using a deep learning-based prior. We utilized the denoising convolutional neural network (DnCNN) method and trained the network using full-dose images as the ground truth and low dose images reconstructed from downsampled data by Poisson thinning as input. Since most published deep networks are trained at a predetermined noise level, the noise level disparity of training and testing data is a major problem for their applicability as a generalized prior. In particular, the noise level significantly changes in each iteration, which can potentially degrade the overall performance of iterative reconstruction. Due to insufficient existing studies, we conducted simulations and evaluated the degradation of performance at various noise conditions. Our findings indicated that DnCNN produces additional bias induced by the disparity of noise levels. To address this issue, we propose a local linear fitting function incorporated with the DnCNN prior to improve the image quality by preventing unwanted bias. We demonstrate that the resultant method is robust against noise level disparities despite the network being trained at a predetermined noise level. By means of bias and standard deviation studies via both simulations and clinical experiments, we show that the proposed method outperforms conventional methods based on total variation and non-local means penalties. We thereby confirm that the proposed method improves the reconstruction result both quantitatively and qualitatively.

Dose reduction in computed tomography (CT) is essential for decreasing radiation risk in clinical applications. Iterative reconstruction algorithms are one of the most promising way to compensate for the increased noise due to reduction of photon flux. Most iterative reconstruction algorithms incorporate manually designed prior functions of the reconstructed image to suppress noises while maintaining structures of the image. These priors basically rely on smoothness constraints and cannot exploit more complex features of the image. The recent development of artificial neural networks and machine learning enabled learning of more complex features of image, which has the potential to improve reconstruction quality. In this letter, K-sparse auto encoder was used for unsupervised feature learning. A manifold was learned from normal-dose images and the distance between the reconstructed image and the manifold was minimized along with data fidelity during reconstruction. Experiments on 2016 Low-dose CT Grand Challenge were used for the method verification, and results demonstrated the noise reduction and detail preservation abilities of the proposed method.

This study aims to employ deep learning for the development of an automatic classifier for the severity of chronic obstructive pulmonary disease (COPD) in patients. A three-layer deep belief network (DBN) with two hidden layers and one visible layer was employed to generate a model for classification, and the model's robustness against exacerbation was analyzed. Subjects from the COPDGene cohort were staged using the GOLD 2011 guidelines. 10,300 subjects with 361 features each were included in the analysis. After feature selection and parameter optimization, the proposed classification method achieved an accuracy of 97.2% by using a 10-fold cross validation experiment. The most sensitive features as revealed by the DBN weights were consistent with the clinical consensus as per previous studies and clinical diagnosis rules. In summary, we demonstrate that the DBN is a competitive tool for exacerbation risk assessment for patients suffering from, COPD.

Modern e-health systems have undergone rapid development thanks to the advances in communications, computing and machine learning technology. Especially, deep learning has great superiority in image analysis and disease prediction. In this paper, we use Alzheimer's Disease (AD) as an example to show advantages of deep learning in diagnosing brain diseases and providing clinical decision support. Firstly, we convert raw functional magnetic resonance imaging (fMRI) to a matrix to represent activity of 90 brain regions. Secondly, to represent the functional connectivity between different brain regions, a correlation matrix is obtained by calculating the correlation between each pair of brain regions. In the next, a targeted autoencoder network is built to classify the correlation matrix, which is sensitive to AD. Finally, the experiment results show that our proposed method for AD prediction achieves much better effects than traditional means. It finds the correlations between different brain regions efficiently, provides strong reference for AD prediction. Compared to Support Vector Machine (SVM), about 25% improvement is gained in prediction accuracy. The e-health field becomes more complete and effective owing to that. Our work helps predict AD at an early stage and take measures to slow down or even prevent the onset of it.

Spectral computed tomography (CT) is a promising technique with the potential for improving lesion detection, tissue characterization, and material decomposition. In this paper, we are interested in kVp switching-based spectral CT that alternates distinct kVp X-ray transmissions during gantry rotation. This system can acquire multiple X-ray energy transmissions without additional radiation dose. However, only sparse views are generated for each spectral measurement; and the spectra themselves are limited in number. To address these limitations, we propose a penalized maximum likelihood method using spectral patch-based low-rank penalty, which exploits the self-similarity of patches that are collected at the same position in spectral images. The main advantage is that the relatively small number of materials within each patch allows us to employ the low-rank penalty that is less sensitive to intensity changes while preserving edge directions. In our optimization formulation, the cost function consists of the Poisson log-likelihood for X-ray transmission and the nonconvex patch-based low-rank penalty. Since the original cost function is difficult to minimize directly, we propose an optimization method using separable quadratic surrogate and concave convex procedure algorithms for the log-likelihood and penalty terms, which results in an alternating minimization that provides a computational advantage because each subproblem can be solved independently. We performed computer simulations and a real experiment using a kVp switching-based spectral CT with sparse-view measurements, and compared the proposed method with conventional algorithms. We confirmed that the proposed method improves spectral images both qualitatively and quantitatively. Furthermore, our GPU implementation significantly reduces the computational cost.

The estimation and analysis of kinetic parameters in dynamic positron emission tomography (PET) is frequently confounded by tissue heterogeneity and partial volume effects. We propose a new constrained model of dynamic PET to address these limitations. The proposed formulation incorporates an explicit mixture model in which each image voxel is represented as a mixture of different pure tissue types with distinct temporal dynamics. We use Cramér-Rao lower bounds to demonstrate that the use of prior information is important to stabilize parameter estimation with this model. As a result, we propose a constrained formulation of the estimation problem that we solve using a two-stage algorithm. In the first stage, a sparse signal processing method is applied to estimate the rate parameters for the different tissue compartments from the noisy PET time series. In the second stage, tissue fractions and the linear parameters of different time activity curves are estimated using a combination of spatial-regularity and fractional mixture constraints. A block coordinate descent algorithm is combined with a manifold search to robustly estimate these parameters. The method is evaluated with both simulated and experimental dynamic PET data.

We investigate using dual time-point PET data to perform Patlak modeling. This approach can be used for whole body dynamic PET studies in which we compute voxel-wise estimates of Patlak parameters using two frames of data for each bed position. Our approach directly uses list-mode arrival times for each event to estimate the Patlak parametric image. We use a penalized likelihood method in which the penalty function uses spatially variant weighting to ensure a count independent local impulse response. We evaluate performance of the method in comparison to fractional changes in SUV values (%DSUV) between the two frames using Cramer Rao analysis and Monte Carlo simulation. Receiver operating characteristic (ROC) curves are used to compare performance in differentiating tumors relative to background based on the dynamic data sets. Using area under the ROC curve as a performance metric, we show superior performance of Patlak relative to %DSUV over a range of dynamic data sets and parameters. These results suggest that Patlak analysis may be appropriate for analysis of dual time-point whole body PET data and could lead to superior detection of tumors relative to %DSUV metrics.

Time-of-flight (TOF) positron emission tomography (PET) scanners offer the potential for significantly improved signal-to-noise ratio (SNR) and lesion detectability in clinical PET. However, fully 3D TOF PET image reconstruction is a challenging task due to the huge data size. One solution to this problem is to rebin TOF data into a lower dimensional format. We have recently developed Fourier rebinning methods for mapping TOF data into non-TOF formats that retain substantial SNR advantages relative to sinograms acquired without TOF information. However, mappings for rebinning into non-TOF formats are not unique and optimization of rebinning methods has not been widely investigated. In this paper we address the question of optimal rebinning in order to make full use of TOF information. We focus on FORET-3D, which approximately rebins 3D TOF data into 3D non-TOF sinogram formats without requiring a Fourier transform in the axial direction. We optimize the weighting for FORET-3D to minimize the variance, resulting in H(2)-weighted FORET-3D, which turns out to be the best linear unbiased estimator (BLUE) under reasonable approximations and furthermore the uniformly minimum variance unbiased (UMVU) estimator under Gaussian noise assumptions. This implies that any information loss due to optimal rebinning is as a result only of the approximations used in deriving the rebinning equation and developing the optimal weighting. We demonstrate using simulated and real phantom TOF data that the optimal rebinning method achieves variance reduction and contrast recovery improvement compared to nonoptimized rebinning weightings. In our preliminary study using a simplified simulation setup, the performance of the optimal rebinning method was comparable to that of fully 3D TOF MAP.

We describe a nonparametric framework for incorporating information from co-registered anatomical images into positron emission tomographic (PET) image reconstruction through priors based on information theoretic similarity measures. We compare and evaluate the use of mutual information (MI) and joint entropy (JE) between feature vectors extracted from the anatomical and PET images as priors in PET reconstruction. Scale-space theory provides a framework for the analysis of images at different levels of detail, and we use this approach to define feature vectors that emphasize prominent boundaries in the anatomical and functional images, and attach less importance to detail and noise that is less likely to be correlated in the two images. Through simulations that model the best case scenario of perfect agreement between the anatomical and functional images, and a more realistic situation with a real magnetic resonance image and a PET phantom that has partial volumes and a smooth variation of intensities, we evaluate the performance of MI and JE based priors in comparison to a Gaussian quadratic prior, which does not use any anatomical information. We also apply this method to clinical brain scan data using F(18) Fallypride, a tracer that binds to dopamine receptors and therefore localizes mainly in the striatum. We present an efficient method of computing these priors and their derivatives based on fast Fourier transforms that reduce the complexity of their convolution-like expressions. Our results indicate that while sensitive to initialization and choice of hyperparameters, information theoretic priors can reconstruct images with higher contrast and superior quantitation than quadratic priors.

Manually labeled landmark sets are often required as inputs for landmark-based image registration. Identifying an optimal subset of landmarks from a training dataset may be useful in reducing the labor intensive task of manual labeling. In this paper, we present a new problem and a method to solve it: given a set of N landmarks, find the k(< N) best landmarks such that aligning these k landmarks that produce the best overall alignment of all N landmarks. The resulting procedure allows us to select a reduced number of landmarks to be labeled as a part of the registration procedure. We apply this methodology to the problem of registering cerebral cortical surfaces extracted from MRI data. We use manually traced sulcal curves as landmarks in performing inter-subject registration of these surfaces. To minimize the error metric, we analyze the correlation structure of the sulcal errors in the landmark points by modeling them as a multivariate Gaussian process. Selection of the optimal subset of sulcal curves is performed by computing the error variance for the subset of unconstrained landmarks conditioned on the constrained set. We show that the registration error predicted by our method closely matches the actual registration error. The method determines optimal curve subsets of any given size with minimal registration error.

Detection of small lesions in fluorodeoxyglucose (FDG) positron emission tomography (PET) is limited by image resolution and low signal to noise ratio. We have previously described a matched subspace detection method that uses the time activity curve to distinguish tumors from background in dynamic FDG PET. Applying this algorithm on a voxel by voxel basis throughout the dynamic image produces a test statistic image or "map" which on thresholding indicates the potential locations of secondary or metastatic tumors. In this paper, we describe a thresholding method that controls familywise error rate (FWER) for the matched subspace detection statistical map. The method involves three steps. First, the PET image is segmented into several homogeneous regions. Then, the statistical map is normalized to a zero mean unit variance Gaussian random field. Finally, the images are thresholded at a fixed FWER by estimating their spatial smoothness and applying a random field theory maximum statistic approach. We evaluate this thresholding method using digital phantoms generated from clinical dynamic images. We also present an application of the proposed approach to clinical PET data from a breast cancer patient with metastatic disease.

We describe a matched subspace detection algorithm to assist in the detection of small tumors in dynamic positron emission tomography (PET) images. The algorithm is designed to differentiate tumors from background using the time activity curves (TACs) that characterize the uptake of PET tracers. TACs are modeled using linear subspaces with additive Gaussian noise. Using TACs from a primary tumor region of interest (ROI) and one or more background ROIs, each identified by a human observer, two linear subspaces are identified. Applying a matched subspace detector to these identified subspaces on a voxel-by-voxel basis throughout the dynamic image produces a test statistic at each voxel which on thresholding indicates potential locations of secondary or metastatic tumors. The detector is derived for three cases: using a single TAC with white noise of unknown variance, using a single TAC with known noise covariance, and detection using multiple TACs within a small ROI with known noise covariance. The noise covariance is estimated for the reconstructed image from the observed sinogram data. To evaluate the proposed method, a simulation-based receiver operating characteristic (ROC) study for dynamic PET tumor detection is designed. The detector uses a dynamic sequence of frame-by-frame 2-D reconstructions as input. We compare the performance of the subspace detectors with that of a Hotelling observer applied to a single frame image and of the Patlak method applied to the dynamic data. We also show examples of the application of each detection approach to clinical PET data from a breast cancer patient with metastatic disease.

We describe a fast forward and back projector pair based on inverse Fourier rebinning for use in iterative image reconstruction for fully 3-D positron emission tomography (PET). The projector pair is used as part of a factored system matrix that takes into account detector-pair response by using shift-variant sinogram blur kernels, thereby combining the computational advantages of Fourier rebinning with iterative reconstruction using accurate system models. The forward projector consists of a 2-D projector, which maps 3-D images into 2-D direct sinograms, followed by exact inverse rebinning which maps the 2-D into fully 3-D sinograms. The back projector is implemented as the transpose of the forward projector and differs from the true exact rebinning operator in the sense that it does not require reprojection to compute missing lines of response (LORs). We compensate for two types of inaccuracies that arise in a cylindrical PET scanner when using inverse Fourier rebinning: 1) nonuniform radial sampling and 2) nonconstant oblique angles in the radial direction in a single oblique sinogram. We examine the effects of these corrections on sinogram accuracy and reconstructed image quality. We evaluate performance of the new projector pair for maximum a posteriori (MAP) reconstruction of simulated and in vivo data. The new projector results in only a small loss in resolution towards the edge of the field-of-view when compared to the fully 3-D geometric projector and requires an order of magnitude less computation.

We describe a fast and globally convergent fully four-dimensional incremental gradient (4DIG) algorithm to estimate the continuous-time tracer density from list mode positron emission tomography (PET) data. Detection of 511-keV photon pairs produced by positron-electron annihilation is modeled as an inhomogeneous Poisson process whose rate function is parameterized using cubic B-splines. The rate functions are estimated by minimizing the cost function formed by the sum of the negative log-likelihood of arrival times, spatial and temporal roughness penalties, and a negativity penalty. We first derive a computable bound for the norm of the optimal temporal basis function coefficients. Based on this bound we then construct and prove convergence of an incremental gradient algorithm. Fully 4-D simulations demonstrate the substantially faster convergence behavior of the 4DIG algorithm relative to preconditioned conjugate gradient. Four-dimensional reconstructions of real data are also included to illustrate the performance of this method.

Randoms precorrected positron emission tomography (PET) data is formed as the difference of two Poisson random variables. Its exact probability mass function (PMF) is inconvenient for use in likelihood-based iterative image reconstruction as it contains an infinite summation. The shifted Poisson model is a tractable approximation to this PMF but requires that negative values are truncated, resulting in positively biased reconstructions in low count studies. Here we analyze the properties of the exact PMF and propose a simple but accurate approximation that allows negative valued data. We investigate the properties of this approximation and demonstrate its application to penalized maximum likelihood image reconstruction.

The Fisher information matrix (FIM) plays a key role in the analysis and applications of statistical image reconstruction methods based on Poisson data models. The elements of the FIM are a function of the reciprocal of the mean values of sinogram elements. Conventional plug-in FIM estimation methods do not work well at low counts, where the FIM estimate is highly sensitive to the reciprocal mean estimates at individual detector pairs. A generalized error look-up table (GELT) method is developed to estimate the reciprocal of the mean of the sinogram data. This approach is also extended to randoms precorrected data. Based on these techniques, an accurate FIM estimate is obtained for both Poisson and randoms precorrected data. As an application, the new GELT method is used to improve resolution uniformity and achieve near-uniform image resolution in low count situations.