Terrence Chen

Xuan Gong, Abhishek Sharma, Srikrishna Karanam et al.

Federated Learning (FL) is a machine learning paradigm where local nodes collaboratively train a central model while the training data remains decentralized. Existing FL methods typically share model parameters or employ co-distillation to address the issue of unbalanced data distribution. However, they suffer from communication bottlenecks. More importantly, they risk privacy leakage. In this work, we develop a privacy preserving and communication efficient method in a FL framework with one-shot offline knowledge distillation using unlabeled, cross-domain public data. We propose a quantized and noisy ensemble of local predictions from completely trained local models for stronger privacy guarantees without sacrificing accuracy. Based on extensive experiments on image classification and text classification tasks, we show that our privacy-preserving method outperforms baseline FL algorithms with superior performance in both accuracy and communication efficiency.

Xuan Gong, Liangchen Song, Rishi Vedula et al.

Federated Learning (FL) is a machine learning paradigm where many local nodes collaboratively train a central model while keeping the training data decentralized. This is particularly relevant for clinical applications since patient data are usually not allowed to be transferred out of medical facilities, leading to the need for FL. Existing FL methods typically share model parameters or employ co-distillation to address the issue of unbalanced data distribution. However, they also require numerous rounds of synchronized communication and, more importantly, suffer from a privacy leakage risk. We propose a privacy-preserving FL framework leveraging unlabeled public data for one-way offline knowledge distillation in this work. The central model is learned from local knowledge via ensemble attention distillation. Our technique uses decentralized and heterogeneous local data like existing FL approaches, but more importantly, it significantly reduces the risk of privacy leakage. We demonstrate that our method achieves very competitive performance with more robust privacy preservation based on extensive experiments on image classification, segmentation, and reconstruction tasks.

Qin Liu, Meng Zheng, Benjamin Planche et al.

The goal of click-based interactive image segmentation is to obtain precise object segmentation masks with limited user interaction, i.e., by a minimal number of user clicks. Existing methods require users to provide all the clicks: by first inspecting the segmentation mask and then providing points on mislabeled regions, iteratively. We ask the question: can our model directly predict where to click, so as to further reduce the user interaction cost? To this end, we propose {\PseudoClick}, a generic framework that enables existing segmentation networks to propose candidate next clicks. These automatically generated clicks, termed pseudo clicks in this work, serve as an imitation of human clicks to refine the segmentation mask.

Liangchen Song, Xuan Gong, Benjamin Planche et al.

Knowing the 3D motions in a dynamic scene is essential to many vision applications. Recent progress is mainly focused on estimating the activity of some specific elements like humans. In this paper, we leverage a neural motion field for estimating the motion of all points in a multiview setting. Modeling the motion from a dynamic scene with multiview data is challenging due to the ambiguities in points of similar color and points with time-varying color. We propose to regularize the estimated motion to be predictable. If the motion from previous frames is known, then the motion in the near future should be predictable. Therefore, we introduce a predictability regularization by first conditioning the estimated motion on latent embeddings, then by adopting a predictor network to enforce predictability on the embeddings. The proposed framework PREF (Predictability REgularized Fields) achieves on par or better results than state-of-the-art neural motion field-based dynamic scene representation methods, while requiring no prior knowledge of the scene.

Xuan Gong, Liangchen Song, Meng Zheng et al.

To date, little attention has been given to multi-view 3D human mesh estimation, despite real-life applicability (e.g., motion capture, sport analysis) and robustness to single-view ambiguities. Existing solutions typically suffer from poor generalization performance to new settings, largely due to the limited diversity of image-mesh pairs in multi-view training data. To address this shortcoming, people have explored the use of synthetic images. But besides the usual impact of visual gap between rendered and target data, synthetic-data-driven multi-view estimators also suffer from overfitting to the camera viewpoint distribution sampled during training which usually differs from real-world distributions. Tackling both challenges, we propose a novel simulation-based training pipeline for multi-view human mesh recovery, which (a) relies on intermediate 2D representations which are more robust to synthetic-to-real domain gap; (b) leverages learnable calibration and triangulation to adapt to more diversified camera setups; and (c) progressively aggregates multi-view information in a canonical 3D space to remove ambiguities in 2D representations. Through extensive benchmarking, we demonstrate the superiority of the proposed solution especially for unseen in-the-wild scenarios.

Xuan Gong, Meng Zheng, Benjamin Planche et al.

Fully supervised human mesh recovery methods are data-hungry and have poor generalizability due to the limited availability and diversity of 3D-annotated benchmark datasets. Recent progress in self-supervised human mesh recovery has been made using synthetic-data-driven training paradigms where the model is trained from synthetic paired 2D representation (e.g., 2D keypoints and segmentation masks) and 3D mesh. However, on synthetic dense correspondence maps (i.e., IUV) few have been explored since the domain gap between synthetic training data and real testing data is hard to address for 2D dense representation. To alleviate this domain gap on IUV, we propose cross-representation alignment utilizing the complementary information from the robust but sparse representation (2D keypoints). Specifically, the alignment errors between initial mesh estimation and both 2D representations are forwarded into regressor and dynamically corrected in the following mesh regression. This adaptive cross-representation alignment explicitly learns from the deviations and captures complementary information: robustness from sparse representation and richness from dense representation. We conduct extensive experiments on multiple standard benchmark datasets and demonstrate competitive results, helping take a step towards reducing the annotation effort needed to produce state-of-the-art models in human mesh estimation.

Mancheng Meng, Ziyan Wu, Terrence Chen et al.

Predicting the future trajectory of a person remains a challenging problem, due to randomness and subjectivity of human movement. However, the moving patterns of human in a constrained scenario typically conform to a limited number of regularities to a certain extent, because of the scenario restrictions and person-person or person-object interactivity. Thus, an individual person in this scenario should follow one of the regularities as well. In other words, a person's subsequent trajectory has likely been traveled by others. Based on this hypothesis, we propose to forecast a person's future trajectory by learning from the implicit scene regularities. We call the regularities, inherently derived from the past dynamics of the people and the environment in the scene, scene history. We categorize scene history information into two types: historical group trajectory and individual-surroundings interaction. To exploit these two types of information for trajectory prediction, we propose a novel framework Scene History Excavating Network (SHENet), where the scene history is leveraged in a simple yet effective approach. In particular, we design two components: the group trajectory bank module to extract representative group trajectories as the candidate for future path, and the cross-modal interaction module to model the interaction between individual past trajectory and its surroundings for trajectory refinement. In addition, to mitigate the uncertainty in ground-truth trajectory, caused by the aforementioned randomness and subjectivity of human movement, we propose to include smoothness into the training process and evaluation metrics. We conduct extensive evaluations to validate the efficacy of our proposed framework on ETH, UCY, as well as a new, challenging benchmark dataset PAV, demonstrating superior performance compared to state-of-the-art methods.

Zikui Cai, Zhongpai Gao, Benjamin Planche et al.

With the rise of cameras and smart sensors, humanity generates an exponential amount of data. This valuable information, including underrepresented cases like AI in medical settings, can fuel new deep-learning tools. However, data scientists must prioritize ensuring privacy for individuals in these untapped datasets, especially for images or videos with faces, which are prime targets for identification methods. Proposed solutions to de-identify such images often compromise non-identifying facial attributes relevant to downstream tasks. In this paper, we introduce Disguise, a novel algorithm that seamlessly de-identifies facial images while ensuring the usability of the modified data. Unlike previous approaches, our solution is firmly grounded in the domains of differential privacy and ensemble-learning research. Our method involves extracting and substituting depicted identities with synthetic ones, generated using variational mechanisms to maximize obfuscation and non-invertibility. Additionally, we leverage supervision from a mixture-of-experts to disentangle and preserve other utility attributes. We extensively evaluate our method using multiple datasets, demonstrating a higher de-identification rate and superior consistency compared to prior approaches in various downstream tasks.

Qin Liu, Meng Zheng, Benjamin Planche et al.

Automatic medical volume segmentation often lacks clinical accuracy, necessitating further refinement. In this work, we interactively approach medical volume segmentation via two decoupled modules: interaction-to-segmentation and segmentation propagation. Given a medical volume, a user first segments a slice (or several slices) via the interaction module and then propagates the segmentation(s) to the remaining slices. The user may repeat this process multiple times until a sufficiently high volume segmentation quality is achieved. However, due to the lack of human correction during propagation, segmentation errors are prone to accumulate in the intermediate slices and may lead to sub-optimal performance. To alleviate this issue, we propose a simple yet effective cycle consistency loss that regularizes an intermediate segmentation by referencing the accurate segmentation in the starting slice. To this end, we introduce a backward segmentation path that propagates the intermediate segmentation back to the starting slice using the same propagation network. With cycle consistency training, the propagation network is better regularized than in standard forward-only training approaches. Evaluation results on challenging AbdomenCT-1K and OAI-ZIB datasets demonstrate the effectiveness of our method.

Yuhao Su, Anwesa Choudhuri, Zhongpai Gao et al.

Large vision-language models struggle with medical video understanding, where spatial precision, temporal reasoning, and clinical semantics are critical. To address this, we first introduce \textbf{MedVidBench}, a large-scale benchmark of 531,850 video-instruction pairs across 8 medical sources spanning video, segment, and frame-level tasks, curated through a rigorous quality assurance pipeline with expert-guided prompting and dual-model validation. While supervised fine-tuning on MedVidBench yields noticeable gains, standard Reinforcement Learning (RL) fails due to imbalanced reward scales across datasets, which destabilizes optimization and leads to training collapse. To overcome this, we introduce \textbf{MedGRPO}, a novel RL framework for balanced multi-dataset training with two key innovations: (1) \emph{cross-dataset reward normalization} that maps each dataset's median performance to a common reward value, ensuring fair optimization regardless of difficulty, and (2) a \emph{medical LLM judge} that evaluates caption quality on five clinical dimensions through comparative similarity scoring. Supervised fine-tuning Qwen2.5-VL-7B on MedVidBench substantially outperforms GPT-4.1 and Gemini-2.5-Flash across all tasks, demonstrating MedVidBench's efficacy, while our MedGRPO framework further improves upon the SFT baseline across grounding and captioning tasks. Our work establishes a foundational benchmark and robust training methodology for advancing vision-language models in medical domains. Our project website is available at https://yuhaosu.github.io/MedGRPO/.

Lin Zhao, Xiao Chen, Eric Z. Chen et al.

Medical image segmentation is crucial for clinical decision-making, but the scarcity of annotated data presents significant challenges. Few-shot segmentation (FSS) methods show promise but often require training on the target domain and struggle to generalize across different modalities. Similarly, adapting foundation models like the Segment Anything Model (SAM) for medical imaging has limitations, including the need for finetuning and domain-specific adaptation. To address these issues, we propose a novel method that adapts DINOv2 and Segment Anything Model 2 (SAM 2) for retrieval-augmented few-shot medical image segmentation. Our approach uses DINOv2's feature as query to retrieve similar samples from limited annotated data, which are then encoded as memories and stored in memory bank. With the memory attention mechanism of SAM 2, the model leverages these memories as conditions to generate accurate segmentation of the target image. We evaluated our framework on three medical image segmentation tasks, demonstrating superior performance and generalizability across various modalities without the need for any retraining or finetuning. Overall, this method offers a practical and effective solution for few-shot medical image segmentation and holds significant potential as a valuable annotation tool in clinical applications.

Meng Zheng, Benjamin Planche, Zhongpai Gao et al.

Conventional 3D medical image segmentation methods typically require learning heavy 3D networks (e.g., 3D-UNet), as well as large amounts of in-domain data with accurate pixel/voxel-level labels to avoid overfitting. These solutions are thus extremely time- and labor-expensive, but also may easily fail to generalize to unseen objects during training. To alleviate this issue, we present MSFSeg, a novel few-shot 3D segmentation framework with a lightweight multi-surrogate fusion (MSF). MSFSeg is able to automatically segment unseen 3D objects/organs (during training) provided with one or a few annotated 2D slices or 3D sequence segments, via learning dense query-support organ/lesion anatomy correlations across patient populations. Our proposed MSF module mines comprehensive and diversified morphology correlations between unlabeled and the few labeled slices/sequences through multiple designated surrogates, making it able to generate accurate cross-domain 3D segmentation masks given annotated slices or sequences. We demonstrate the effectiveness of our proposed framework by showing superior performance on conventional few-shot segmentation benchmarks compared to prior art, and remarkable cross-domain cross-volume segmentation performance on proprietary 3D segmentation datasets for challenging entities, i.e., tubular structures, with only limited 2D or 3D labels.

Eric Z. Chen, Chi Zhang, Xiao Chen et al.

Compared with 2D MRI, 3D MRI provides superior volumetric spatial resolution and signal-to-noise ratio. However, it is more challenging to reconstruct 3D MRI images. Current methods are mainly based on convolutional neural networks (CNN) with small kernels, which are difficult to scale up to have sufficient fitting power for 3D MRI reconstruction due to the large image size and GPU memory constraint. Furthermore, MRI reconstruction is a deconvolution problem, which demands long-distance information that is difficult to capture by CNNs with small convolution kernels. The multi-layer perceptron (MLP) can model such long-distance information, but it requires a fixed input size. In this paper, we proposed Recon3DMLP, a hybrid of CNN modules with small kernels for low-frequency reconstruction and adaptive MLP (dMLP) modules with large kernels to boost the high-frequency reconstruction, for 3D MRI reconstruction. We further utilized the circular shift operation based on MRI physics such that dMLP accepts arbitrary image size and can extract global information from the entire FOV. We also propose a GPU memory efficient data fidelity module that can reduce $>$50$\%$ memory. We compared Recon3DMLP with other CNN-based models on a high-resolution (HR) 3D MRI dataset. Recon3DMLP improves HR 3D reconstruction and outperforms several existing CNN-based models under similar GPU memory consumption, which demonstrates that Recon3DMLP is a practical solution for HR 3D MRI reconstruction.

Lin Zhao, Xiao Chen, Eric Z. Chen et al.

Multi-contrast Magnetic Resonance Imaging (MRI) generates multiple medical images with rich and complementary information for routine clinical use; however, it suffers from a long acquisition time. Recent works for accelerating MRI, mainly designed for single contrast, may not be optimal for multi-contrast scenario since the inherent correlations among the multi-contrast images are not exploited. In addition, independent reconstruction of each contrast usually does not translate to optimal performance of downstream tasks. Motivated by these aspects, in this paper we design an end-to-end framework for accelerating multi-contrast MRI which simultaneously optimizes the entire MR imaging workflow including sampling, reconstruction and downstream tasks to achieve the best overall outcomes. The proposed framework consists of a sampling mask generator for each image contrast and a reconstructor exploiting the inter-contrast correlations with a recurrent structure which enables the information sharing in a holistic way. The sampling mask generator and the reconstructor are trained jointly across the multiple image contrasts. The acceleration ratio of each image contrast is also learnable and can be driven by a downstream task performance. We validate our approach on a multi-contrast brain dataset and a multi-contrast knee dataset. Experiments show that (1) our framework consistently outperforms the baselines designed for single contrast on both datasets; (2) our newly designed recurrent reconstruction network effectively improves the reconstruction quality for multi-contrast images; (3) the learnable acceleration ratio improves the downstream task performance significantly. Overall, this work has potentials to open up new avenues for optimizing the entire multi-contrast MR imaging workflow.

Siyuan Dong, Eric Z. Chen, Lin Zhao et al.

High-quality MRI reconstruction plays a critical role in clinical applications. Deep learning-based methods have achieved promising results on MRI reconstruction. However, most state-of-the-art methods were designed to optimize the evaluation metrics commonly used for natural images, such as PSNR and SSIM, whereas the visual quality is not primarily pursued. Compared to the fully-sampled images, the reconstructed images are often blurry, where high-frequency features might not be sharp enough for confident clinical diagnosis. To this end, we propose an invertible sharpening network (InvSharpNet) to improve the visual quality of MRI reconstructions. During training, unlike the traditional methods that learn to map the input data to the ground truth, InvSharpNet adapts a backward training strategy that learns a blurring transform from the ground truth (fully-sampled image) to the input data (blurry reconstruction). During inference, the learned blurring transform can be inverted to a sharpening transform leveraging the network's invertibility. The experiments on various MRI datasets demonstrate that InvSharpNet can improve reconstruction sharpness with few artifacts. The results were also evaluated by radiologists, indicating better visual quality and diagnostic confidence of our proposed method.

Xiaoling Hu, Xiao Chen, Yikang Liu et al.

Accurate segmentation and motion estimation of myocardium have always been important in clinic field, which essentially contribute to the downstream diagnosis. However, existing methods cannot always guarantee the shape integrity for myocardium segmentation. In addition, motion estimation requires point correspondence on the myocardium region across different frames. In this paper, we propose a novel end-to-end deep statistic shape model to focus on myocardium segmentation with both shape integrity and boundary correspondence preserving. Specifically, myocardium shapes are represented by a fixed number of points, whose variations are extracted by Principal Component Analysis (PCA). Deep neural network is used to predict the transformation parameters (both affine and deformation), which are then used to warp the mean point cloud to the image domain. Furthermore, a differentiable rendering layer is introduced to incorporate mask supervision into the framework to learn more accurate point clouds. In this way, the proposed method is able to consistently produce anatomically reasonable segmentation mask without post processing. Additionally, the predicted point cloud guarantees boundary correspondence for sequential images, which contributes to the downstream tasks, such as the motion estimation of myocardium. We conduct several experiments to demonstrate the effectiveness of the proposed method on several benchmark datasets.

Tianyu Luan, Zhongpai Gao, Luyuan Xie et al.

We introduce a novel bottom-up approach for human body mesh reconstruction, specifically designed to address the challenges posed by partial visibility and occlusion in input images. Traditional top-down methods, relying on whole-body parametric models like SMPL, falter when only a small part of the human is visible, as they require visibility of most of the human body for accurate mesh reconstruction. To overcome this limitation, our method employs a "Divide and Fuse (D&F)" strategy, reconstructing human body parts independently before fusing them, thereby ensuring robustness against occlusions. We design Human Part Parametric Models (HPPM) that independently reconstruct the mesh from a few shape and global-location parameters, without inter-part dependency. A specially designed fusion module then seamlessly integrates the reconstructed parts, even when only a few are visible. We harness a large volume of ground-truth SMPL data to train our parametric mesh models. To facilitate the training and evaluation of our method, we have established benchmark datasets featuring images of partially visible humans with HPPM annotations. Our experiments, conducted on these benchmark datasets, demonstrate the effectiveness of our D&F method, particularly in scenarios with substantial invisibility, where traditional approaches struggle to maintain reconstruction quality.

Yikang Liu, Eric Z. Chen, Xiao Chen et al.

The k-space data generated from magnetic resonance imaging (MRI) is only a finite sampling of underlying signals. Therefore, MRI images often suffer from low spatial resolution and Gibbs ringing artifacts. Previous studies tackled these two problems separately, where super resolution methods tend to enhance Gibbs artifacts, whereas Gibbs ringing removal methods tend to blur the images. It is also a challenge that high resolution ground truth is hard to obtain in clinical MRI. In this paper, we propose an unsupervised learning framework for both MRI super resolution and Gibbs artifacts removal without using high resolution ground truth. Furthermore, we propose regularization methods to improve the model's generalizability across out-of-distribution MRI images. We evaluated our proposed methods with other state-of-the-art methods on eight MRI datasets with various contrasts and anatomical structures. Our method not only achieves the best SR performance but also significantly reduces the Gibbs artifacts. Our method also demonstrates good generalizability across different datasets, which is beneficial to clinical applications where training data are usually scarce and biased.

Daniel H. Pak, Xiao Chen, Eric Z. Chen et al.

Dynamic Magnetic Resonance Imaging (dMRI) is widely used to assess various cardiac conditions such as cardiac motion and blood flow. To accelerate MR acquisition, techniques such as undersampling and Simultaneous Multi-Slice (SMS) are often used. Special reconstruction algorithms are needed to reconstruct multiple SMS image slices from the entangled information. Deep learning (DL)-based methods have shown promising results for single-slice MR reconstruction, but the addition of SMS acceleration raises unique challenges due to the composite k-space signals and the resulting images with strong inter-slice artifacts. Furthermore, many dMRI applications lack sufficient data for training reconstruction neural networks. In this study, we propose a novel DL-based framework for dynamic SMS reconstruction. Our main contributions are 1) a combination of data transformation steps and network design that effectively leverages the unique characteristics of undersampled dynamic SMS data, and 2) an MR physics-guided transfer learning strategy that addresses the data scarcity issue. Thorough comparisons with multiple baseline methods illustrate the strengths of our proposed methods.

Junyi Wu, Van Nguyen Nguyen, Benjamin Planche et al.

We introduce Consistent Instance Field, a continuous and probabilistic spatio-temporal representation for dynamic scene understanding. Unlike prior methods that rely on discrete tracking or view-dependent features, our approach disentangles visibility from persistent object identity by modeling each space-time point with an occupancy probability and a conditional instance distribution. To realize this, we introduce a novel instance-embedded representation based on deformable 3D Gaussians, which jointly encode radiance and semantic information and are learned directly from input RGB images and instance masks through differentiable rasterization. Furthermore, we introduce new mechanisms to calibrate per-Gaussian identities and resample Gaussians toward semantically active regions, ensuring consistent instance representations across space and time. Experiments on HyperNeRF and Neu3D datasets demonstrate that our method significantly outperforms state-of-the-art methods on novel-view panoptic segmentation and open-vocabulary 4D querying tasks.

Zhongpai Gao, Abhishek Sharma, Meng Zheng et al.

Positioning patients for scanning and interventional procedures is a critical task that requires high precision and accuracy. The conventional workflow involves manually adjusting the patient support to align the center of the target body part with the laser projector or other guiding devices. This process is not only time-consuming but also prone to inaccuracies. In this work, we propose an automated patient positioning system that utilizes a camera to detect specific hand gestures from technicians, allowing users to indicate the target patient region to the system and initiate automated positioning. Our approach relies on a novel multi-stage pipeline to recognize and interpret the technicians' gestures, translating them into precise motions of medical devices. We evaluate our proposed pipeline during actual MRI scanning procedures, using RGB-Depth cameras to capture the process. Results show that our system achieves accurate and precise patient positioning with minimal technician intervention. Furthermore, we validate our method on HaGRID, a large-scale hand gesture dataset, demonstrating its effectiveness in hand detection and gesture recognition.

Yikang Liu, Lin Zhao, Eric Z. Chen et al.

Dynamic coronary roadmapping is a technology that overlays the vessel maps (the "roadmap") extracted from an offline image sequence of X-ray angiography onto a live stream of X-ray fluoroscopy in real-time. It aims to offer navigational guidance for interventional surgeries without the need for repeated contrast agent injections, thereby reducing the risks associated with radiation exposure and kidney failure. The precision of the roadmaps is contingent upon the accurate alignment of angiographic and fluoroscopic images based on their cardiac phases, as well as precise catheter tip tracking. The former ensures the selection of a roadmap that closely matches the vessel shape in the current frame, while the latter uses catheter tips as reference points to adjust for translational motion between the roadmap and the present vessel tree. Training deep learning models for both tasks is challenging and underexplored. However, incorporating catheter features into the models could offer substantial benefits, given humans heavily rely on catheters to complete the tasks. To this end, we introduce a simple but effective method, auxiliary input in training (AIT), and demonstrate that it enhances model performance across both tasks, outperforming baseline methods in knowledge incorporation and transfer learning.

Xi Ouyang, Srikrishna Karanam, Ziyan Wu et al.

We consider the problem of abnormality localization for clinical applications. While deep learning has driven much recent progress in medical imaging, many clinical challenges are not fully addressed, limiting its broader usage. While recent methods report high diagnostic accuracies, physicians have concerns trusting these algorithm results for diagnostic decision-making purposes because of a general lack of algorithm decision reasoning and interpretability. One potential way to address this problem is to further train these models to localize abnormalities in addition to just classifying them. However, doing this accurately will require a large amount of disease localization annotations by clinical experts, a task that is prohibitively expensive to accomplish for most applications. In this work, we take a step towards addressing these issues by means of a new attention-driven weakly supervised algorithm comprising a hierarchical attention mining framework that unifies activation- and gradient-based visual attention in a holistic manner. Our key algorithmic innovations include the design of explicit ordinal attention constraints, enabling principled model training in a weakly-supervised fashion, while also facilitating the generation of visual-attention-driven model explanations by means of localization cues. On two large-scale chest X-ray datasets (NIH ChestX-ray14 and CheXpert), we demonstrate significant localization performance improvements over the current state of the art while also achieving competitive classification performance. Our code is available on https://github.com/oyxhust/HAM.

Meng Zheng, Samhita Marri, Anwesa Choudhuri et al.

Vision-language-action (VLA) models provide a promising paradigm for scalable robotic manipulation, yet their reliance on success-only behavioral cloning leaves them brittle; lacking corrective training signals, minor execution errors rapidly compound into unrecoverable, out-of-distribution failures. To address this limitation, we propose Adaptive Failure-Informed Learning (AFIL), an end-to-end framework that leverages failure trajectories as adaptive negative guidance for diffusion- and flow-based VLA policies. AFIL uses a pretrained VLA to generate failure rollouts online, avoiding the need for handcrafted failure-mode design or human-in-the-loop recovery. It then jointly trains Dual Action Generators (DAGs) for successful and failed behaviors while sharing a common vision-language backbone, enabling efficient failure-aware policy learning with limited parameter overhead. During sampling, the failure generator adaptively steers action generation away from failure-prone regions and toward more reliable success modes, with guidance strength determined by the per-diffusion-step distance between success and failure distributions. Experiments across in-domain and out-of-domain robotic manipulation tasks, covering both short- and long-horizon settings, show that AFIL consistently improves task success rates and robustness over existing VLA baselines, demonstrating its effectiveness, efficiency, and generality.

Ange Lou, Benjamin Planche, Zhongpai Gao et al.

Addressing the intricate challenge of modeling and re-rendering dynamic scenes, most recent approaches have sought to simplify these complexities using plane-based explicit representations, overcoming the slow training time issues associated with methods like Neural Radiance Fields (NeRF) and implicit representations. However, the straightforward decomposition of 4D dynamic scenes into multiple 2D plane-based representations proves insufficient for re-rendering high-fidelity scenes with complex motions. In response, we present a novel direction-aware representation (DaRe) approach that captures scene dynamics from six different directions. This learned representation undergoes an inverse dual-tree complex wavelet transformation (DTCWT) to recover plane-based information. DaReNeRF computes features for each space-time point by fusing vectors from these recovered planes. Combining DaReNeRF with a tiny MLP for color regression and leveraging volume rendering in training yield state-of-the-art performance in novel view synthesis for complex dynamic scenes. Notably, to address redundancy introduced by the six real and six imaginary direction-aware wavelet coefficients, we introduce a trainable masking approach, mitigating storage issues without significant performance decline. Moreover, DaReNeRF maintains a 2x reduction in training time compared to prior art while delivering superior performance.

Meng Zheng, Benjamin Planche, Xuan Gong et al.

3D patient body modeling is critical to the success of automated patient positioning for smart medical scanning and operating rooms. Existing CNN-based end-to-end patient modeling solutions typically require a) customized network designs demanding large amount of relevant training data, covering extensive realistic clinical scenarios (e.g., patient covered by sheets), which leads to suboptimal generalizability in practical deployment, b) expensive 3D human model annotations, i.e., requiring huge amount of manual effort, resulting in systems that scale poorly. To address these issues, we propose a generic modularized 3D patient modeling method consists of (a) a multi-modal keypoint detection module with attentive fusion for 2D patient joint localization, to learn complementary cross-modality patient body information, leading to improved keypoint localization robustness and generalizability in a wide variety of imaging (e.g., CT, MRI etc.) and clinical scenarios (e.g., heavy occlusions); and (b) a self-supervised 3D mesh regression module which does not require expensive 3D mesh parameter annotations to train, bringing immediate cost benefits for clinical deployment. We demonstrate the efficacy of the proposed method by extensive patient positioning experiments on both public and clinical data. Our evaluation results achieve superior patient positioning performance across various imaging modalities in real clinical scenarios.

Hao Ding, Zhongpai Gao, Benjamin Planche et al.

Surgical phase recognition (SPR) is crucial for applications in workflow optimization, performance evaluation, and real-time intervention guidance. However, current deep learning models often struggle with fragmented predictions, failing to capture the sequential nature of surgical workflows. We propose the Neural Finite-State Machine (NFSM), a novel approach that enforces temporal coherence by integrating classical state-transition priors with modern neural networks. NFSM leverages learnable global state embeddings as unique phase identifiers and dynamic transition tables to model phase-to-phase progressions. Additionally, a future phase forecasting mechanism employs repeated frame padding to anticipate upcoming transitions. Implemented as a plug-and-play module, NFSM can be integrated into existing SPR pipelines without changing their core architectures. We demonstrate state-of-the-art performance across multiple benchmarks, including a significant improvement on the BernBypass70 dataset - raising video-level accuracy by 0.9 points and phase-level precision, recall, F1-score, and mAP by 3.8, 3.1, 3.3, and 4.1, respectively. Ablation studies confirm each component's effectiveness and the module's adaptability to various architectures. By unifying finite-state principles with deep learning, NFSM offers a robust path toward consistent, long-term surgical video analysis.

Andong Deng, Zhongpai Gao, Anwesa Choudhuri et al.

Temporal awareness is essential for video large language models (LLMs) to understand and reason about events within long videos, enabling applications like dense video captioning and temporal video grounding in a unified system. However, the scarcity of long videos with detailed captions and precise temporal annotations limits their temporal awareness. In this paper, we propose Seq2Time, a data-oriented training paradigm that leverages sequences of images and short video clips to enhance temporal awareness in long videos. By converting sequence positions into temporal annotations, we transform large-scale image and clip captioning datasets into sequences that mimic the temporal structure of long videos, enabling self-supervised training with abundant time-sensitive data. To enable sequence-to-time knowledge transfer, we introduce a novel time representation that unifies positional information across image sequences, clip sequences, and long videos. Experiments demonstrate the effectiveness of our method, achieving a 27.6% improvement in F1 score and 44.8% in CIDEr on the YouCook2 benchmark and a 14.7% increase in recall on the Charades-STA benchmark compared to the baseline.

Arindam Dutta, Meng Zheng, Zhongpai Gao et al.

Reconstructing clothed humans from a single image is a fundamental task in computer vision with wide-ranging applications. Although existing monocular clothed human reconstruction solutions have shown promising results, they often rely on the assumption that the human subject is in an occlusion-free environment. Thus, when encountering in-the-wild occluded images, these algorithms produce multiview inconsistent and fragmented reconstructions. Additionally, most algorithms for monocular 3D human reconstruction leverage geometric priors such as SMPL annotations for training and inference, which are extremely challenging to acquire in real-world applications. To address these limitations, we propose CHROME: Clothed Human Reconstruction with Occlusion-Resilience and Multiview-ConsistEncy from a Single Image, a novel pipeline designed to reconstruct occlusion-resilient 3D humans with multiview consistency from a single occluded image, without requiring either ground-truth geometric prior annotations or 3D supervision. Specifically, CHROME leverages a multiview diffusion model to first synthesize occlusion-free human images from the occluded input, compatible with off-the-shelf pose control to explicitly enforce cross-view consistency during synthesis. A 3D reconstruction model is then trained to predict a set of 3D Gaussians conditioned on both the occluded input and synthesized views, aligning cross-view details to produce a cohesive and accurate 3D representation. CHROME achieves significant improvements in terms of both novel view synthesis (upto 3 db PSNR) and geometric reconstruction under challenging conditions.

Zhongpai Gao, Benjamin Planche, Meng Zheng et al.

Real-time rendering of dynamic scenes with view-dependent effects remains a fundamental challenge in computer graphics. While recent advances in Gaussian Splatting have shown promising results separately handling dynamic scenes (4DGS) and view-dependent effects (6DGS), no existing method unifies these capabilities while maintaining real-time performance. We present 7D Gaussian Splatting (7DGS), a unified framework representing scene elements as seven-dimensional Gaussians spanning position (3D), time (1D), and viewing direction (3D). Our key contribution is an efficient conditional slicing mechanism that transforms 7D Gaussians into view- and time-conditioned 3D Gaussians, maintaining compatibility with existing 3D Gaussian Splatting pipelines while enabling joint optimization. Experiments demonstrate that 7DGS outperforms prior methods by up to 7.36 dB in PSNR while achieving real-time rendering (401 FPS) on challenging dynamic scenes with complex view-dependent effects. The project page is: https://gaozhongpai.github.io/7dgs/.

Xiao Chen, Shunan Zhang, Eric Z. Chen et al.

In artificial intelligence (AI), especially deep learning, data diversity and volume play a pivotal role in model development. However, training a robust deep learning model often faces challenges due to data privacy, regulations, and the difficulty of sharing data between different locations, especially for medical applications. To address this, we developed a method called the Federated Data Model (FDM). This method uses diffusion models to learn the characteristics of data at one site and then creates synthetic data that can be used at another site without sharing the actual data. We tested this approach with a medical image segmentation task, focusing on cardiac magnetic resonance images from different hospitals. Our results show that models trained with this method perform well both on the data they were originally trained on and on data from other sites. This approach offers a promising way to train accurate and privacy-respecting AI models across different locations.

Shihan Qiu, Shaoyan Pan, Yikang Liu et al.

Current deep learning reconstruction for accelerated cardiac cine MRI suffers from spatial and temporal blurring. We aim to improve image sharpness and motion delineation for cine MRI under high undersampling rates. A spatiotemporal diffusion enhancement model conditional on an existing deep learning reconstruction along with a novel paired sampling strategy was developed. The diffusion model provided sharper tissue boundaries and clearer motion than the original reconstruction in experts evaluation on clinical data. The innovative paired sampling strategy substantially reduced artificial noises in the generative results.

Xiaoran Zhang, Eric Z. Chen, Lin Zhao et al.

We propose a novel approach that adapts hierarchical vision foundation models for real-time ultrasound image segmentation. Existing ultrasound segmentation methods often struggle with adaptability to new tasks, relying on costly manual annotations, while real-time approaches generally fail to match state-of-the-art performance. To overcome these limitations, we introduce an adaptive framework that leverages the vision foundation model Hiera to extract multi-scale features, interleaved with DINOv2 representations to enhance visual expressiveness. These enriched features are then decoded to produce precise and robust segmentation. We conduct extensive evaluations on six public datasets and one in-house dataset, covering both cardiac and thyroid ultrasound segmentation. Experiments show that our approach outperforms state-of-the-art methods across multiple datasets and excels with limited supervision, surpassing nnUNet by over 20\% on average in the 1\% and 10\% data settings. Our method achieves $\sim$77 FPS inference speed with TensorRT on a single GPU, enabling real-time clinical applications.

Ange Lou, Benjamin Planche, Zhongpai Gao et al.

Numerous recent approaches to modeling and re-rendering dynamic scenes leverage plane-based explicit representations, addressing slow training times associated with models like neural radiance fields (NeRF) and Gaussian splatting (GS). However, merely decomposing 4D dynamic scenes into multiple 2D plane-based representations is insufficient for high-fidelity re-rendering of scenes with complex motions. In response, we present DaRePlane, a novel direction-aware representation approach that captures scene dynamics from six different directions. This learned representation undergoes an inverse dual-tree complex wavelet transformation (DTCWT) to recover plane-based information. Within NeRF pipelines, DaRePlane computes features for each space-time point by fusing vectors from these recovered planes, then passed to a tiny MLP for color regression. When applied to Gaussian splatting, DaRePlane computes the features of Gaussian points, followed by a tiny multi-head MLP for spatial-time deformation prediction. Notably, to address redundancy introduced by the six real and six imaginary direction-aware wavelet coefficients, we introduce a trainable masking approach, mitigating storage issues without significant performance decline. To demonstrate the generality and efficiency of DaRePlane, we test it on both regular and surgical dynamic scenes, for both NeRF and GS systems. Extensive experiments show that DaRePlane yields state-of-the-art performance in novel view synthesis for various complex dynamic scenes.

Yuchun Liu, Benjamin Planche, Meng Zheng et al.

Deep implicit functions (DIFs) have emerged as a potent and articulate means of representing 3D shapes. However, methods modeling object categories or non-rigid entities have mainly focused on single-object scenarios. In this work, we propose MODIF, a multi-object deep implicit function that jointly learns the deformation fields and instance-specific latent codes for multiple objects at once. Our emphasis is on non-rigid, non-interpenetrating entities such as organs. To effectively capture the interrelation between these entities and ensure precise, collision-free representations, our approach facilitates signaling between category-specific fields to adequately rectify shapes. We also introduce novel inter-object supervision: an attraction-repulsion loss is formulated to refine contact regions between objects. Our approach is demonstrated on various medical benchmarks, involving modeling different groups of intricate anatomical entities. Experimental results illustrate that our model can proficiently learn the shape representation of each organ and their relations to others, to the point that shapes missing from unseen instances can be consistently recovered by our method. Finally, MODIF can also propagate semantic information throughout the population via accurate point correspondences

Basar Demir, Yikang Liu, Xiao Chen et al.

Many self-supervised denoising approaches have been proposed in recent years. However, these methods tend to overly smooth images, resulting in the loss of fine structures that are essential for medical applications. In this paper, we propose DiffDenoise, a powerful self-supervised denoising approach tailored for medical images, designed to preserve high-frequency details. Our approach comprises three stages. First, we train a diffusion model on noisy images, using the outputs of a pretrained Blind-Spot Network as conditioning inputs. Next, we introduce a novel stabilized reverse sampling technique, which generates clean images by averaging diffusion sampling outputs initialized with a pair of symmetric noises. Finally, we train a supervised denoising network using noisy images paired with the denoised outputs generated by the diffusion model. Our results demonstrate that DiffDenoise outperforms existing state-of-the-art methods in both synthetic and real-world medical image denoising tasks. We provide both a theoretical foundation and practical insights, demonstrating the method's effectiveness across various medical imaging modalities and anatomical structures.

Hieu Tran, Junda Wang, Yujan Ting et al.

Large language models (LLMs) have shown remarkable capabilities in various natural language processing tasks, yet they often struggle with maintaining factual accuracy, particularly in knowledge-intensive domains like healthcare. This study introduces LEAF: Learning and Evaluation Augmented by Fact-Checking, a novel approach designed to enhance the factual reliability of LLMs, with a focus on medical question answering (QA). LEAF utilizes a dual strategy to enhance the factual accuracy of responses from models such as Llama 3 70B Instruct and Llama 3 8B Instruct. The first strategy, Fact-Check-Then-RAG, improves Retrieval-Augmented Generation (RAG) by incorporating fact-checking results to guide the retrieval process without updating model parameters. The second strategy, Learning from Fact-Checks via Self-Training, involves supervised fine-tuning (SFT) on fact-checked responses or applying Simple Preference Optimization (SimPO) with fact-checking as a ranking mechanism, both updating LLM parameters from supervision. These findings suggest that integrating fact-checked responses whether through RAG enhancement or self-training enhances the reliability and factual correctness of LLM outputs, offering a promising solution for applications where information accuracy is crucial.

Bin Wang, Anwesa Choudhuri, Meng Zheng et al.

Interactive segmentation aims to accurately segment target objects with minimal user interactions. However, current methods often fail to accurately separate target objects from the background, due to a limited understanding of order, the relative depth between objects in a scene. To address this issue, we propose OIS: order-aware interactive segmentation, where we explicitly encode the relative depth between objects into order maps. We introduce a novel order-aware attention, where the order maps seamlessly guide the user interactions (in the form of clicks) to attend to the image features. We further present an object-aware attention module to incorporate a strong object-level understanding to better differentiate objects with similar order. Our approach allows both dense and sparse integration of user clicks, enhancing both accuracy and efficiency as compared to prior works. Experimental results demonstrate that OIS achieves state-of-the-art performance, improving mIoU after one click by 7.61 on the HQSeg44K dataset and 1.32 on the DAVIS dataset as compared to the previous state-of-the-art SegNext, while also doubling inference speed compared to current leading methods. The project page is https://ukaukaaaa.github.io/projects/OIS/index.html

Zhongpai Gao, Meng Zheng, Benjamin Planche et al.

Volumetric rendering of Computed Tomography (CT) scans is crucial for visualizing complex 3D anatomical structures in medical imaging. Current high-fidelity approaches, especially neural rendering techniques, require time-consuming per-scene optimization, limiting clinical applicability due to computational demands and poor generalizability. We propose Render-FM, a novel foundation model for direct, real-time volumetric rendering of CT scans. Render-FM employs an encoder-decoder architecture that directly regresses 6D Gaussian Splatting (6DGS) parameters from CT volumes, eliminating per-scan optimization through large-scale pre-training on diverse medical data. By integrating robust feature extraction with the expressive power of 6DGS, our approach efficiently generates high-quality, real-time interactive 3D visualizations across diverse clinical CT data. Experiments demonstrate that Render-FM achieves visual fidelity comparable or superior to specialized per-scan methods while drastically reducing preparation time from nearly an hour to seconds for a single inference step. This advancement enables seamless integration into real-time surgical planning and diagnostic workflows. The project page is: https://gaozhongpai.github.io/renderfm/.

Shaoyan Pan, Yikang Liu, Lin Zhao et al.

The accurate segmentation of guidewires in interventional cardiac fluoroscopy videos is crucial for computer-aided navigation tasks. Although deep learning methods have demonstrated high accuracy and robustness in wire segmentation, they require substantial annotated datasets for generalizability, underscoring the need for extensive labeled data to enhance model performance. To address this challenge, we propose the Segmentation-guided Frame-consistency Video Diffusion Model (SF-VD) to generate large collections of labeled fluoroscopy videos, augmenting the training data for wire segmentation networks. SF-VD leverages videos with limited annotations by independently modeling scene distribution and motion distribution. It first samples the scene distribution by generating 2D fluoroscopy images with wires positioned according to a specified input mask, and then samples the motion distribution by progressively generating subsequent frames, ensuring frame-to-frame coherence through a frame-consistency strategy. A segmentation-guided mechanism further refines the process by adjusting wire contrast, ensuring a diverse range of visibility in the synthesized image. Evaluation on a fluoroscopy dataset confirms the superior quality of the generated videos and shows significant improvements in guidewire segmentation.

Junda Wang, Yujan Ting, Eric Z. Chen et al.

Multimodal large language models (MLLMs) have made significant strides, yet they face challenges in the medical domain due to limited specialized knowledge. While recent medical MLLMs demonstrate strong performance in lab settings, they often struggle in real-world applications, highlighting a substantial gap between research and practice. In this paper, we seek to address this gap at various stages of the end-to-end learning pipeline, including data collection, model fine-tuning, and evaluation. At the data collection stage, we introduce SemiHVision, a dataset that combines human annotations with automated augmentation techniques to improve both medical knowledge representation and diagnostic reasoning. For model fine-tuning, we trained PMC-Cambrian-8B-AN over 2400 H100 GPU hours, resulting in performance that surpasses public medical models like HuatuoGPT-Vision-34B (79.0% vs. 66.7%) and private general models like Claude3-Opus (55.7%) on traditional benchmarks such as SLAKE and VQA-RAD. In the evaluation phase, we observed that traditional benchmarks cannot accurately reflect realistic clinical task capabilities. To overcome this limitation and provide more targeted guidance for model evaluation, we introduce the JAMA Clinical Challenge, a novel benchmark specifically designed to evaluate diagnostic reasoning. On this benchmark, PMC-Cambrian-AN achieves state-of-the-art performance with a GPT-4 score of 1.29, significantly outperforming HuatuoGPT-Vision-34B (1.13) and Claude3-Opus (1.17), demonstrating its superior diagnostic reasoning abilities.

Shihan Qiu, Shaoyan Pan, Yikang Liu et al.

The currently limited quality of accelerated cardiac cine reconstruction may potentially be improved by the emerging diffusion models, but the clinically unacceptable long processing time poses a challenge. We aim to develop a clinically feasible diffusion-model-based reconstruction pipeline to improve the image quality of cine MRI. A multi-in multi-out diffusion enhancement model together with fast inference strategies were developed to be used in conjunction with a reconstruction model. The diffusion reconstruction reduced spatial and temporal blurring in prospectively undersampled clinical data, as validated by experts inspection. The 1.5s per video processing time enabled the approach to be applied in clinical scenarios.

Zhongpai Gao, Huayi Zhou, Abhishek Sharma et al.

The detection of human parts (e.g., hands, face) and their correct association with individuals is an essential task, e.g., for ubiquitous human-machine interfaces and action recognition. Traditional methods often employ multi-stage processes, rely on cumbersome anchor-based systems, or do not scale well to larger part sets. This paper presents PBADet, a novel one-stage, anchor-free approach for part-body association detection. Building upon the anchor-free object representation across multi-scale feature maps, we introduce a singular part-to-body center offset that effectively encapsulates the relationship between parts and their parent bodies. Our design is inherently versatile and capable of managing multiple parts-to-body associations without compromising on detection accuracy or robustness. Comprehensive experiments on various datasets underscore the efficacy of our approach, which not only outperforms existing state-of-the-art techniques but also offers a more streamlined and efficient solution to the part-body association challenge.

Jiachen Tao, Benjamin Planche, Van Nguyen Nguyen et al.

Accurately modeling light transport is essential for realistic image synthesis. Photon mapping provides physically grounded estimates of complex global illumination effects such as caustics and specular-diffuse interactions, yet its per-view radiance estimation remains computationally inefficient when rendering multiple views of the same scene. The inefficiency arises from independent photon tracing and stochastic kernel estimation at each viewpoint, leading to inevitable redundant computation. To accelerate multi-view rendering, we reformulate photon mapping as a continuous and reusable radiance function. Specifically, we introduce the Gaussian Photon Field (GPF), a learnable representation that encodes photon distributions as anisotropic 3D Gaussian primitives parameterized by position, rotation, scale, and spectrum. GPF is initialized from physically traced photons in the first SPPM iteration and optimized using multi-view supervision of final radiance, distilling photon-based light transport into a continuous field. Once trained, the field enables differentiable radiance evaluation along camera rays without repeated photon tracing or iterative refinement. Extensive experiments on scenes with complex light transport, such as caustics and specular-diffuse interactions, demonstrate that GPF attains photon-level accuracy while reducing computation by orders of magnitude, unifying the physical rigor of photon-based rendering with the efficiency of neural scene representations.

Rong Liu, Zhongpai Gao, Benjamin Planche et al.

We introduce Universal Beta Splatting (UBS), a unified framework that generalizes 3D Gaussian Splatting to N-dimensional anisotropic Beta kernels for explicit radiance field rendering. Unlike fixed Gaussian primitives, Beta kernels enable controllable dependency modeling across spatial, angular, and temporal dimensions within a single representation. Our unified approach captures complex light transport effects, handles anisotropic view-dependent appearance, and models scene dynamics without requiring auxiliary networks or specific color encodings. UBS maintains backward compatibility by approximating to Gaussian Splatting as a special case, guaranteeing plug-in usability and lower performance bounds. The learned Beta parameters naturally decompose scene properties into interpretable without explicit supervision: spatial (surface vs. texture), angular (diffuse vs. specular), and temporal (static vs. dynamic). Our CUDA-accelerated implementation achieves real-time rendering while consistently outperforming existing methods across static, view-dependent, and dynamic benchmarks, establishing Beta kernels as a scalable universal primitive for radiance field rendering. Our project website is available at https://rongliu-leo.github.io/universal-beta-splatting/.

Yiwei Li, Yikang Liu, Jiaqi Guo et al.

Anatomical understanding through deep learning is critical for automatic report generation, intra-operative navigation, and organ localization in medical imaging; however, its progress is constrained by the scarcity of expert-labeled data. A promising remedy is to leverage an annotated reference image to guide the interpretation of an unlabeled target. Although recent vision-language models (VLMs) exhibit non-trivial visual reasoning, their reference-based understanding and fine-grained localization remain limited. We introduce RAU, a framework for reference-based anatomical understanding with VLMs. We first show that a VLM learns to identify anatomical regions through relative spatial reasoning between reference and target images, trained on a moderately sized dataset. We validate this capability through visual question answering (VQA) and bounding box prediction. Next, we demonstrate that the VLM-derived spatial cues can be seamlessly integrated with the fine-grained segmentation capability of SAM2, enabling localization and pixel-level segmentation of small anatomical regions, such as vessel segments. Across two in-distribution and two out-of-distribution datasets, RAU consistently outperforms a SAM2 fine-tuning baseline using the same memory setup, yielding more accurate segmentations and more reliable localization. More importantly, its strong generalization ability makes it scalable to out-of-distribution datasets, a property crucial for medical image applications. To the best of our knowledge, RAU is the first to explore the capability of VLMs for reference-based identification, localization, and segmentation of anatomical structures in medical images. Its promising performance highlights the potential of VLM-driven approaches for anatomical understanding in automated clinical workflows.

Lin Zhao, Xin Yu, Yikang Liu et al.

Accurate correspondence matching in coronary angiography images is crucial for reconstructing 3D coronary artery structures, which is essential for precise diagnosis and treatment planning of coronary artery disease (CAD). Traditional matching methods for natural images often fail to generalize to X-ray images due to inherent differences such as lack of texture, lower contrast, and overlapping structures, compounded by insufficient training data. To address these challenges, we propose a novel pipeline that generates realistic paired coronary angiography images using a diffusion model conditioned on 2D projections of 3D reconstructed meshes from Coronary Computed Tomography Angiography (CCTA), providing high-quality synthetic data for training. Additionally, we employ large-scale image foundation models to guide feature aggregation, enhancing correspondence matching accuracy by focusing on semantically relevant regions and keypoints. Our approach demonstrates superior matching performance on synthetic datasets and effectively generalizes to real-world datasets, offering a practical solution for this task. Furthermore, our work investigates the efficacy of different foundation models in correspondence matching, providing novel insights into leveraging advanced image foundation models for medical imaging applications.

Anwesa Choudhuri, Zhongpai Gao, Meng Zheng et al.

Early detection, accurate segmentation, classification and tracking of polyps during colonoscopy are critical for preventing colorectal cancer. Many existing deep-learning-based methods for analyzing colonoscopic videos either require task-specific fine-tuning, lack tracking capabilities, or rely on domain-specific pre-training. In this paper, we introduce PolypSegTrack, a novel foundation model that jointly addresses polyp detection, segmentation, classification and unsupervised tracking in colonoscopic videos. Our approach leverages a novel conditional mask loss, enabling flexible training across datasets with either pixel-level segmentation masks or bounding box annotations, allowing us to bypass task-specific fine-tuning. Our unsupervised tracking module reliably associates polyp instances across frames using object queries, without relying on any heuristics. We leverage a robust vision foundation model backbone that is pre-trained unsupervisedly on natural images, thereby removing the need for domain-specific pre-training. Extensive experiments on multiple polyp benchmarks demonstrate that our method significantly outperforms existing state-of-the-art approaches in detection, segmentation, classification, and tracking.

Meng Zheng, Jiajin Zhang, Benjamin Planche et al.

Image-Text Retrieval (ITR) finds broad applications in healthcare, aiding clinicians and radiologists by automatically retrieving relevant patient cases in the database given the query image and/or report, for more efficient clinical diagnosis and treatment, especially for rare diseases. However conventional ITR systems typically only rely on global image or text representations for measuring patient image/report similarities, which overlook local distinctiveness across patient cases. This often results in suboptimal retrieval performance. In this paper, we propose an Anatomical Location-Conditioned Image-Text Retrieval (ALC-ITR) framework, which, given a query image and the associated suspicious anatomical region(s), aims to retrieve similar patient cases exhibiting the same disease or symptoms in the same anatomical region. To perform location-conditioned multimodal retrieval, we learn a medical Relevance-Region-Aligned Vision Language (RRA-VL) model with semantic global-level and region-/word-level alignment to produce generalizable, well-aligned multi-modal representations. Additionally, we perform location-conditioned contrastive learning to further utilize cross-pair region-level contrastiveness for improved multi-modal retrieval. We show that our proposed RRA-VL achieves state-of-the-art localization performance in phase-grounding tasks, and satisfying multi-modal retrieval performance with or without location conditioning. Finally, we thoroughly investigate the generalizability and explainability of our proposed ALC-ITR system in providing explanations and preliminary diagnosis reports given retrieved patient cases (conditioned on anatomical regions), with proper off-the-shelf LLM prompts.

Zhongpai Gao, Benjamin Planche, Meng Zheng et al.

Digitally reconstructed radiographs (DRRs) are simulated 2D X-ray images generated from 3D CT volumes, widely used in preoperative settings but limited in intraoperative applications due to computational bottlenecks, especially for accurate but heavy physics-based Monte Carlo methods. While analytical DRR renderers offer greater efficiency, they overlook anisotropic X-ray image formation phenomena, such as Compton scattering. We present a novel approach that marries realistic physics-inspired X-ray simulation with efficient, differentiable DRR generation using 3D Gaussian splatting (3DGS). Our direction-disentangled 3DGS (DDGS) method separates the radiosity contribution into isotropic and direction-dependent components, approximating complex anisotropic interactions without intricate runtime simulations. Additionally, we adapt the 3DGS initialization to account for tomography data properties, enhancing accuracy and efficiency. Our method outperforms state-of-the-art techniques in image accuracy. Furthermore, our DDGS shows promise for intraoperative applications and inverse problems such as pose registration, delivering superior registration accuracy and runtime performance compared to analytical DRR methods.