Variational autoencoders (VAE) encode data into lower-dimensional latent vectors before decoding those vectors back to data. Once trained, decoding a random latent vector from the prior usually does not produce meaningful data, at least when the latent space has more than a dozen dimensions. In this paper, we investigate this issue by drawing insight from high dimensional statistics: in these regimes, the latent vectors of a standard VAE are by construction distributed uniformly on a hypersphere. We propose to formulate the latent variables of a VAE using hyperspherical coordinates, which allows compressing the latent vectors towards an island on the hypersphere, thereby reducing the latent sparsity and we show that this improves the generation ability of the VAE. We propose a new parameterization of the latent space with limited computational overhead.

Managing chronic wounds is a global challenge that can be alleviated by the adoption of automatic systems for clinical wound assessment from consumer-grade videos. While 2D image analysis approaches are insufficient for handling the 3D features of wounds, existing approaches utilizing 3D reconstruction methods have not been thor-oughly evaluated. To address this gap, this paper presents a comprehensive study on 3D wound reconstruction from consumer-grade videos. Specifically, we introduce the SALVE dataset, comprising video recordings of realistic wound phantoms captured with different cameras. Using this dataset, we assess the accuracy and precision of state-of-the-art methods for 3D reconstruction, ranging from traditional photogrammetry pipelines to advanced neural rendering approaches. In our experiments, we observe that photogrammetry approaches do not provide smooth surfaces suitable for precise clinical measurements of wounds. Neural rendering approaches show promise in addressing this issue, advancing the use of this technology in wound care practices. We encourage the readers to visit the project page: https://rcmichierchia.github.io/SALVE/.

Neural Rendering representations have significantly contributed to the field of 3D computer vision. Given their potential, considerable efforts have been invested to improve their performance. Nonetheless, the essential question of selecting training views is yet to be thoroughly investigated. This key aspect plays a vital role in achieving high-quality results and aligns with the well-known tenet of deep learning: “garbage in, garbage out”. In this paper, we first illustrate the importance of view selection by demonstrating how a simple rotation of the test views within the most pervasive NeRF dataset can lead to consequential shifts in the performance rankings of state-of-the-art techniques. To address this challenge, we introduce a unified framework for view selection methods and devise a thorough benchmark to assess its impact. Significant improvements can be achieved without leveraging error or uncertainty estimation but focusing on uniform view coverage of the reconstructed object, resulting in a training-free approach. Using this technique, we show that high-quality renderings can be achieved faster by using fewer views. We conduct extensive experiments on both synthetic datasets and realistic data to demonstrate the effectiveness of our proposed method compared with random, conventional error-based, and uncertainty-guided view selection.

Saliency methods are critical tools that allow the estimation of the most important features of an input image that contribute to the network’s prediction. These tools are pivotal in high-stakes applications such as medical diagnosis or autonomous driving. Additionally, these tools can help identify models’ biasedness, such as a strong prior on object placement, easily distinguishable background features, or frequent object co-occurrence. We introduce RankPix, a novel saliency method for visual bias identification in image classification tasks. RankPix is a derivative-free approach that allows the identification of a minimum subset of pixels/features at a given network layer that changes the output of a classifier. Surprisingly, this approaches provides equivalent performance to gradient-based approaches on the standard pointing game benchmark. More interestingly, RankPix outperforms traditional approaches for systematic bias identification.

Convolutional Neural Networks (CNNs) are powerful in learning patterns of different vision tasks, but they are sensitive to label noise and may overfit to noisy labels during training. The early stopping strategy averts updating CNNs during the early training phase and is widely employed in the presence of noisy labels. Motivated by biological findings that the amplitude spectrum (AS) and phase spectrum (PS) in the frequency domain play different roles in the animal’s vision system, we observe that PS, which captures more semantic information, can increase the robustness of CNNs to label noise, more so than AS can. We thus propose early stops at different times for AS and PS by disentangling the features of some layer(s) into AS and PS using Discrete Fourier Transform (DFT) during training. Our proposed Phase-AmplituDe DisentangLed Early Stopping (PADDLES) method is shown to be effective on both synthetic and real-world label-noise datasets. PADDLES out-performs other early stopping methods and obtains state-of-the-art performance.

Deep learning models are powerful tools for addressing challenging medical imaging problems. However, for an ever-growing range of applications, interpreting a model’s prediction remains non-trivial. Understanding decisions made by black-box algorithms is critical, and assessing their fairness and susceptibility to bias is a key step towards healthcare deployment. In this paper, we propose DBCE (Deformation Based Counterfactual Explainability). We optimise a diffeomorphic transformation that deforms a given input image to change the prediction of the model. This provides anatomically meaningful saliency maps indicating tissue atrophy and expansion, which can be easily interpreted by clinicians. In our test case, DBCE replicates the transition of a patient from healthy control (HC) to Alzheimer’s disease (AD). We benchmark DBCE against three commonly used saliency methods. We show that it provides more meaningful saliency maps when applied to one subject and disease-consistent atrophy patterns when used over a larger cohort. In addition, our method fulfils a recent sanity check and is repeatable for different model initialisations in contrast to classical sensitivity-based methods.

Recent advances in geometric deep-learning introduce complex computational challenges for evaluating the distance between meshes. From a mesh model, point clouds are necessary along with a robust distance metric to assess surface quality or as part of the loss function for training models. Current methods often rely on a uniform random mesh discretization, which yields irregular sampling and noisy distance estimation. In this paper we introduce MongeNet, a fast and optimal transport based sampler that allows for an accurate discretization of a mesh with better approximation properties. We compare our method to the ubiquitous random uniform sampling and show that the approximation error is almost half with a very small computational overhead.

In this paper, we introduce CorticalFlow, a new geometric deep-learning model that, given a 3-dimensional image, learns to deform a reference template towards a targeted object. To conserve the template mesh’s topological properties, we train our model over a set of diffeomorphic transformations. This new implementation of a flow Ordinary Differential Equation (ODE) framework benefits from a small GPU memory footprint, allowing the generation of surfaces with several hundred thousand vertices. To reduce topological errors introduced by its discrete resolution, we derive numeric conditions which improve the manifoldness of the predicted triangle mesh. To exhibit the utility of CorticalFlow, we demonstrate its performance for the challenging task of brain cortical surface reconstruction. In contrast to the current state-of-the-art, CorticalFlow produces superior surfaces while reducing the computation time from nine and a half minutes to one second. More significantly, CorticalFlow enforces the generation of anatomically plausible surfaces; the absence of which has been a major impediment restricting the clinical relevance of such surface reconstruction methods.

The study of neurodegenerative diseases relies on the reconstruction and analysis of the brain cortex from magnetic resonance imaging (MRI). Traditional frameworks for this task like FreeSurfer demand lengthy runtimes, while its accelerated variant FastSurfer still relies on a voxel-wise segmentation which is limited by its resolution to capture narrow continuous objects as cortical surfaces. Having these limitations in mind, we propose DeepCSR, a 3D deep learning framework for cortical surface reconstruction from MRI. Towards this end, we train a neural network model with hypercolumn features to predict implicit surface representations for points in a brain template space. After training, the cortical surface at a desired level of detail is obtained by evaluating surface representations at specific coordinates, and subsequently applying a topology correction algorithm and an isosurface extraction method. Thanks to the continuous nature of this approach and the efficacy of its hypercolumn features scheme, DeepCSR efficiently reconstructs cortical surfaces at high resolution capturing fine details in the cortical folding. Moreover, DeepCSR is as accurate, more precise, and faster than the widely used FreeSurfer toolbox and its deep learning powered variant FastSurfer on reconstructing cortical surfaces from MRI which should facilitate large-scale medical studies and new healthcare applications.

Automated detection of small, low level shapes such as circular/spherical objects in images is a challenging computer vision problem. For many applications, especially microbleed detection in Alzheimer's disease, an automatic pre-screening scheme is required to identify potential seeds with high sensitivity and reasonable specificity. A new method is proposed to detect spherical objects in 3D medical images within the multi-scale Laplacian of Gaussian framework. The major contributions are(1)breaking down 3D sphere detection into 1D line profile detection along each coordinate dimension, (2) identifying center of structures bynormalizing the line response profile and (3) employing eigenvalues of the Hessian matrix at optimum scale for the center points to determine spherical objects. The method is validated both on simulated data and susceptibility weighted MRI images with ground truth provided by a medical expert. Validation results demonstrate that the current approach has higher performance in terms of sensitivity and specificity and is effective in detecting adjacent microbleeds, with invariance to intensity, orientation, translation and object scale.

OBJECTIVE: As large-scale medical imaging studies are becoming more common, there is an increasing reliance on automated software to extract quantitative information from these images. As the size of the cohorts keeps increasing with large studies, there is a also a need for tools that allow results from automated image processing and analysis to be presented in a way that enables fast and efficient quality checking, tagging and reporting on cases in which automatic processing failed or was problematic. MATERIALS AND METHODS: MilxXplore is an open source visualization platform, which provides an interface to navigate and explore imaging data in a web browser, giving the end user the opportunity to perform quality control and reporting in a user friendly, collaborative and efficient way. DISCUSSION: Compared to existing software solutions that often provide an overview of the results at the subject's level, MilxXplore pools the results of individual subjects and time points together, allowing easy and efficient navigation and browsing through the different acquisitions of a subject over time, and comparing the results against the rest of the population. CONCLUSIONS: MilxXplore is fast, flexible and allows remote quality checks of processed imaging data, facilitating data sharing and collaboration across multiple locations, and can be easily integrated into a cloud computing pipeline. With the growing trend of open data and open science, such a tool will become increasingly important to share and publish results of imaging analysis.

Accurate localization of the prostate and its surrounding tissue is essential in the treatment of prostate cancer. This paper presents a novel approach to fully automatically segment the prostate, including its seminal vesicles, within a few minutes of a magnetic resonance (MR) scan acquired without an endorectal coil. Such MR images are important in external beam radiation therapy, where using an endorectal coil is highly undesirable. The segmentation is obtained using a deformable model that is trained on-the-fly so that it is specific to the patient's scan. This case specific deformable model consists of a patient specific initialized triangulated surface and image feature model that are trained during its initialization. The image feature model is used to deform the initialized surface by template matching image features (via normalized cross-correlation) to the features of the scan. The resulting deformations are regularized over the surface via well established simple surface smoothing algorithms, which is then made anatomically valid via an optimized shape model. Mean and median Dice's similarity coefficients (DSCs) of 0.85 and 0.87 were achieved when segmenting 3T MR clinical scans of 50 patients. The median DSC result was equal to the inter-rater DSC and had a mean absolute surface error of 1.85 mm. The approach is showed to perform well near the apex and seminal vesicles of the prostate.

Denoising algorithms can alleviate the trade-off between noise-level and acquisition time that still exists for certain image types. Nonlocal means, a recently proposed technique, outperforms other methods in removing noise while retaining image structure, albeit at prohibitive computational cost. Modifications have been proposed to reduce the cost, but the method is still too slow for practical filtering of 3D images. This paper proposes a hashed approach to explicitly represent two summed frequency (hash) functions of local descriptors (patches), utilizing all available image data. Unlike other approaches, the hash spaces are discretized on a regular grid, so primarily linear operations are used. The large memory requirements are overcome by recursing the hash spaces. Additional speed gains are obtained by using a marginal linear interpolation method. Careful choice of the patch features results in high computational efficiency, at similar accuracies. The proposed approach can filter a 3D image in less than a minute versus 15 minutes to 3 hours for existing nonlocal means methods.

We report in this communication a new formulation for the cost function of the well-known fuzzy C-means classification technique whereby we introduce weights. We derive the equations of this new weighted fuzzy C-means algorithm (WFCM) in the presence of additive and multiplicative bias field. We show that the weights can be designed in the same manner as prior probabilities commonly used in maximum a posteriori classifier (MAP) to introduce prior knowledge (e.g. using atlas), and increase robustness to noise (e.g. using Markov random field). Using prior probabilities of three popular MAP algorithms, we compare the performances of our proposed WFCM scheme using the simulated MRI T1W BrainWeb datasets, as well as five T1W MR patient scans. Our results show that WFCM achieves superior performances for low SNR conditions, whereas a Gaussian mixture model is desirable for high noise levels. WFCM allows rigorous comparison of fuzzy and probabilistic classifiers, and offers a framework where improvements can be shared between those two types of classifier.

We are developing methods to characterize atherosclerotic disease in human carotid arteries using multiple MR images having different contrast mechanisms (T1W, T2W, PDW). To enable the use of voxel gray values for interpretation of disease, we created a new method, local entropy minimization with a bicubic spline model (LEMS), to correct the severe (approximately 80%) intensity inhomogeneity that arises from the surface coil array. This entropy-based method does not require classification and robustly addresses some problems that are more severe than those found in brain imaging, including noise, steep bias field, sensitivity of artery wall voxels to edge artifacts, and signal voids near the artery wall. Validation studies were performed on a synthetic digital phantom with realistic intensity inhomogeneity, a physical phantom roughly mimicking the neck, and patient carotid artery images. We compared LEMS to a modified fuzzy c-means segmentation based method (mAFCM), and a linear filtering method (LINF). Following LEMS correction, skeletal muscles in patient images were relatively isointense across the field of view. In the physical phantom, LEMS reduced the variation in the image to 1.9% and across the vessel wall region to 2.5%, a value which should be sufficient to distinguish plaque tissue types, based on literature measurements. In conclusion, we believe that the correction method shows promise for aiding human and computerized tissue classification from MR signal intensities.

This paper presents a use of geometric-based method integrated with classifier for lumen wall estimation using MR plaque volumes. The following are the new things the readers will observe when it comes to plaque imaging. (a) Application of three different sets of : classifiers (fuzzy, Markovian and graph-based) for lumen region classification in plaque MR volumes. These classifiers are used in multi-resolution framework. (b) Usage of rule-based region merging applied to the sub-classes of lumen region. (c) Rotational effect on region of interest in arterial bifurcation zones for accurate lumen region identification and boundary estimation. We have used our diagnostic system with three different classifying methods on actual patient data. We measure performance of the system by computing the mean distance error with respect to boundaries traced manually by human experts. Overall, the system consists of 22,500 boundary points. The in-plane pixel resolution is 0.25 millimeters. Using Markovian classifier method, the average error was 0.61 pixels; using fuzzy classifier method, the average error was 0.62 pixels; using graph-based classifier method, the average error was 0.74 pixels. All these methods lead to error less than 0.185 mm. We also validated our system by simulating the lumen images with additive Gaussian perturbations. This system works on a Linux platform and is written in C++.