Background Accurate segmentation of abdominopelvic visceral adipose tissue (visceral fat) is critical for assessing the health risks associated with central obesity using positron emission tomography and cone-beam computed tomography. However, low-dose cone-beam computed tomography images are difficult to analyze because of anatomical complexity and low contrast. We developed an Anatomical Structure-Guided Segmentation Network, a deep learning–based artificial intelligence framework that integrates anatomical priors through Spatially Adaptive Normalization within a residual encoder–decoder backbone to achieve anatomically consistent visceral fat segmentation. Methods Data from 150 individuals who underwent positron emission tomography and cone-beam computed tomography as part of health screening were retrospectively analyzed. Ground-truth segmentations were manually refined from TotalSegmentator outputs. Five-fold cross-validation was applied to ensure robustness and generalizability. Segmentation accuracy was evaluated using Dice Similarity Coefficient, Intersection over Union, and 95th-percentile Hausdorff Distance, and compared with representative convolutional and transformer-based architectures. Results Visceral fat volumes from the proposed model and reference volumes showed high agreement (concordance correlation coefficient = 0.999; 95% confidence interval: 0.998–0.999). The mean percentage difference was −1.0%, with 95% limits of agreement from −8.5% to +6.5%. The proposed framework achieved the highest overall segmentation accuracy (Dice = 0.965 ± 0.004; Intersection = 0.932 ± 0.007; Hausdorff = 1.632 ± 0.512) and maintained robustness across abdominal regions. Conclusion The Anatomical Structure-Guided Segmentation Network offers a robust, anatomically guided framework for accurate visceral fat segmentation, with the potential to stratify clinical risk in metabolic and oncologic conditions.

Aligning a pair of cross-modality images (e.g., MR-CT, CBCT-CT) is important, yet conventional approaches, including registration or image-to-image (I2I) translation methods often have limitations. To overcome these challenges, we introduce a "Register by Generation (RbG)" framework, a novel 2D deep learning approach designed to generate images that are structurally well-aligned with the fixed image while preserving the detailed intensity and contrast of the moving image, which we refer to as the reference image. Our approach operates in two sequential key stages: first, we employ a novel semi-global reference-augmented image synthesis network incorporating Patch Adaptive Instance Normalization (PAdaIN). This method leverages a down-sampled reference image to guide local adaptive synthesis, generating a more accurately aligned image with a reduced risk of hallucinations. In the second stage, we introduce a detailed refining reference-augmented network featuring a Deformation-Aware Cross-Attention (DACA) block, which aims to recover finer details and textures that may be missing from the initial stage. This unique component (DACA block) enables the transfer of corresponding relevant features from the reference image, effectively performing a "copy-and-paste" operation within the latent feature space. Additionally, we propose a novel combination of loss functions that enables self-supervised training on misaligned datasets, eliminating the need for pre-aligned data. We rigorously evaluate our method on multiple misaligned datasets using metrics focused on structural alignment and distributional consistency, demonstrating comprehensively superior performance. Furthermore, we test its robustness by simulating intentional misalignments in a well-aligned dataset. Additionally, experiments from a case study and downstream segmentation tasks highlight the broad applicability of our approach.

Curriculum learning is a training mechanism in reinforcement learning (RL) that facilitates the achievement of complex policies by progressively increasing the task difficulty during training. However, designing effective curricula for a specific task often requires extensive domain knowledge and human intervention, which limits its applicability across various domains. Our core idea is that large language models (LLMs), with their extensive training on diverse language data and ability to encapsulate world knowledge, present significant potential for efficiently breaking down tasks and decomposing skills across various robotics environments. Additionally, the demonstrated success of LLMs in translating natural language into executable code for RL agents strengthens their role in generating task curricula. In this work, we propose CurricuLLM, which leverages the high-level planning and programming capabilities of LLMs for curriculum design, thereby enhancing the efficient learning of complex target tasks. CurricuLLM consists of: (Step 1) Generating a sequence of subtasks that aid target task learning in natural language form, (Step 2) Translating natural language description of subtasks in executable task code, including the reward code and goal distribution code, and (Step 3) Evaluating trained policies based on trajectory rollout and subtask description. We evaluate Cur-ricuLLM in various robotics simulation environments, ranging from manipulation, navigation, and locomotion, to show that CurricuLLM can aid learning complex robot control tasks. In addition, we validate humanoid locomotion policy learned through CurricuLLM in the real-world. Project website is https://iconlab.negarmehr.com/CurricuLLM/

RegistFormer, our novel reference-augmented image synthesis framework, generates aligned pseudo-CT images (with respect to MR) from misaligned MR and CT pairs. RegistFormer addresses the limitations of intensity-based registration methods, which often fail due to dissimilar image features and complex deformation fields. Unlike conventional image-to-image (I2I) translation methods, our method uses a misaligned CT scan as an auxiliary input to guide the synthesis task through the Deformation-Aware Cross-Attention (DACA) mechanism. DACA integrates the deformation field from a registration method to aggregate spatially matched features from the misaligned CT into MR spatial coordinates. Additionally, we propose a novel combination of loss functions for training with datasets of misaligned MR-CT pairs in a self-supervised manner, eliminating the need for pre-aligned training data. Experiments were conducted with the synthRAD202311https://synthrad2023.grand-challenge.org/ MR-CT pelvis pair dataset. RegistFormer outperforms past state-of-the-art methods, including I2I, registration, and hybrid (registration + I2I), across metrics evaluating both structure alignment and distribution similarity. Moreover, RegistFormer demonstrates superior performance in zero-shot segmentation downstream tasks, highlighting its clinical value. Source code: https://github.com/danny4159/RegistFormer

Ensuring safe navigation in human-populated environments is crucial for autonomous mobile robots. Although recent advances in machine learning offer promising methods to predict human trajectories in crowded areas, it remains unclear how one can safely incorporate these learned models into a control loop due to the uncertain nature of human motion, which can make predictions of these models imprecise. In this work, we address this challenge and introduce a distributionally robust chance-constrained model predictive control (DRCC-MPC) which: (i) adopts a probability of collision as a pre-specified, interpretable risk metric, and (ii) offers robustness against discrepancies between actual human trajectories and their predictions. We consider the risk of collision in the form of a chance constraint, providing an interpretable measure of robot safety. To enable real-time evaluation of chance constraints, we consider conservative approximations of chance constraints in the form of distributionally robust Conditional Value at Risk constraints. The resulting formulation offers computational efficiency as well as robustness with respect to out-of-distribution human motion. With the parallelization of a sampling-based optimization technique, our method operates in real-time, demonstrating successful and safe navigation in a number of case studies with real-world pedestrian data.

The quality of medical images is critical for Computer-aided diagnosis (CAD) and Image-guided robotic interventions because accurate and high-quality images are required to perform each task. High resolution and Signal-to-Noise Ratio (SNR) images are required to analyze and navigate the robotic instruments to the accurate localization inside the body. However, medical images are often of substantially lower quality than clean photographic images due to various factors. In this study we focus on a post-processing based strategy for reducing the amount of noise in MRI images. We propose a method based on Denoising Diffusion Probablistic Models (DDPM), also known as diffusion models the reduce the amount of noise in the image. Specifically, a two-stage DDPM method is proposed – estimating the amount of noise and designating to the correct stage in the Marchov Chain in the reverse diffusion operation, and iteratively and gradually reducing noise by reversing the process. Our experiment was performed on an actually scanned images on a clinical MR scanner, with the reference image that were averaged to match the SNR. Our quantitative and qualitative comparison shows that our method outperforms previous methods including supervised training based on two different metrics (SSIM, PSNR). It demonstrates the effectiveness of the DDPM-based method in reducing noise in the image. Moreover, the resulting image quality achieved with the proposed approach shows that tissue sub-structures are clearer. The noise reduction performance of the proposed method for multiple adjacent slices and various contrasts was tested to show the modelś ability to reduce noise across a diverse set of imaging conditions, which is essential in real-world scenarios.