T3: Test-Time Model Merging in VLMs for Zero-Shot Medical Imaging Analysis

In medical imaging, vision-language models face a critical duality: pretrained networks offer broad robustness but lack subtle, modality-specific characteristics, while fine-tuned expert models achieve high in-distribution accuracy yet falter under modality shift. Existing model-merging techniques, designed for natural-image benchmarks, are simple and efficient but fail to deliver consistent gains across diverse medical modalities; their static interpolation limits reliability in varied clinical tasks. To address this, we introduce Test-Time Task adaptive merging (T^3), a backpropagation-free framework that computes per-sample interpolation coefficients via the Jensen-Shannon divergence between the two models' output distributions. T^3 dynamically preserves local precision when models agree and defers to generalist robustness under drift. To overcome the inference costs of sample-wise merging, we further propose a batch-wise extension, T^3_B, that computes a merging coefficient across a batch of samples, dramatically reducing computational bottleneck. Recognizing the lack of a standardized medical-merging benchmark, we present a rigorous cross-evaluation protocol spanning in-domain, base-to-novel, and corruptions across four modalities. Empirically, T^3 sets new state-of-the-art in Top-1 accuracy and error reduction, outperforming strong baselines while maintaining efficiency, paving the way for adaptive MVLM deployment in clinical settings. Our code is available at https://github.com/Razaimam45/TCube.

BRIQA: Balanced Reweighting in Image Quality Assessment of Pediatric Brain MRI

Assessing the severity of artifacts in pediatric brain Magnetic Resonance Imaging (MRI) is critical for diagnostic accuracy, especially in low-field systems where the signal-to-noise ratio is reduced. Manual quality assessment is time-consuming and subjective, motivating the need for robust automated solutions. In this work, we propose BRIQA (Balanced Reweighting in Image Quality Assessment), which addresses class imbalance in artifact severity levels. BRIQA uses gradient-based loss reweighting to dynamically adjust per-class contributions and employs a rotating batching scheme to ensure consistent exposure to underrepresented classes. Through experiments, no single architecture performs best across all artifact types, emphasizing the importance of architectural diversity. The rotating batching configuration improves performance across metrics by promoting balanced learning when combined with cross-entropy loss. BRIQA improves average macro F1 score from 0.659 to 0.706, with notable gains in Noise (0.430), Zipper (0.098), Positioning (0.097), Contrast (0.217), Motion (0.022), and Banding (0.012) artifact severity classification. The code is available at https://github.com/BioMedIA-MBZUAI/BRIQA.

CardioBench: Do Echocardiography Foundation Models Generalize Beyond the Lab?

Foundation models (FMs) are reshaping medical imaging, yet their application in echocardiography remains limited. While several echocardiography-specific FMs have recently been introduced, no standardized benchmark exists to evaluate them. Echocardiography poses unique challenges, including noisy acquisitions, high frame redundancy, and limited public datasets. Most existing solutions evaluate on private data, restricting comparability. To address this, we introduce CardioBench, a comprehensive benchmark for echocardiography FMs. CardioBench unifies eight publicly available datasets into a standardized suite spanning four regression and five classification tasks, covering functional, structural, diagnostic, and view recognition endpoints. We evaluate several leading FM, including cardiac-specific, biomedical, and general-purpose encoders, under consistent zero-shot, probing, and alignment protocols. Our results highlight complementary strengths across model families: temporal modeling is critical for functional regression, retrieval provides robustness under distribution shift, and domain-specific text encoders capture physiologically meaningful axes. General-purpose encoders transfer strongly and often close the gap with probing, but struggle with fine-grained distinctions like view classification and subtle pathology recognition. By releasing preprocessing, splits, and public evaluation pipelines, CardioBench establishes a reproducible reference point and offers actionable insights to guide the design of future echocardiography foundation models.

Decoupling Clinical and Class-Agnostic Features for Reliable Few-Shot Adaptation under Shift

Medical vision-language models (VLMs) offer promise for clinical decision support, yet their reliability under distribution shifts remains a major concern for safe deployment. These models often learn task-agnostic correlations due to variability in imaging protocols and free-text reports, limiting their generalizability and increasing the risk of failure in real-world settings. We propose DRiFt, a structured feature decoupling framework that explicitly separates clinically relevant signals from task-agnostic noise using parameter-efficient tuning (LoRA) and learnable prompt tokens. To enhance cross-modal alignment and reduce uncertainty, we curate high-quality, clinically grounded image-text pairs by generating captions for a diverse medical dataset. Our approach improves in-distribution performance by +11.4% Top-1 accuracy and +3.3% Macro-F1 over prior prompt-based methods, while maintaining strong robustness across unseen datasets. Ablation studies reveal that disentangling task-relevant features and careful alignment significantly enhance model generalization and reduce unpredictable behavior under domain shift. These insights contribute toward building safer, more trustworthy VLMs for clinical use. The code is available at https://github.com/rumaima/DRiFt.

Advancing Fetal Ultrasound Image Quality Assessment in Low-Resource Settings

Accurate fetal biometric measurements, such as abdominal circumference, play a vital role in prenatal care. However, obtaining high-quality ultrasound images for these measurements heavily depends on the expertise of sonographers, posing a significant challenge in low-income countries due to the scarcity of trained personnel. To address this issue, we leverage FetalCLIP, a vision-language model pretrained on a curated dataset of over 210,000 fetal ultrasound image-caption pairs, to perform automated fetal ultrasound image quality assessment (IQA) on blind-sweep ultrasound data. We introduce FetalCLIP$_{CLS}$, an IQA model adapted from FetalCLIP using Low-Rank Adaptation (LoRA), and evaluate it on the ACOUSLIC-AI dataset against six CNN and Transformer baselines. FetalCLIP$_{CLS}$ achieves the highest F1 score of 0.757. Moreover, we show that an adapted segmentation model, when repurposed for classification, further improves performance, achieving an F1 score of 0.771. Our work demonstrates how parameter-efficient fine-tuning of fetal ultrasound foundation models can enable task-specific adaptations, advancing prenatal care in resource-limited settings. The experimental code is available at: https://github.com/donglihe-hub/FetalCLIP-IQA.

TransPrune: Token Transition Pruning for Efficient Large Vision-Language Model

Large Vision-Language Models (LVLMs) have advanced multimodal learning but face high computational costs due to the large number of visual tokens, motivating token pruning to improve inference efficiency. The key challenge lies in identifying which tokens are truly important. Most existing approaches rely on attention-based criteria to estimate token importance. However, they inherently suffer from certain limitations, such as positional bias. In this work, we explore a new perspective on token importance based on token transitions in LVLMs. We observe that the transition of token representations provides a meaningful signal of semantic information. Based on this insight, we propose TransPrune, a training-free and efficient token pruning method. Specifically, TransPrune progressively prunes tokens by assessing their importance through a combination of Token Transition Variation (TTV)-which measures changes in both the magnitude and direction of token representations-and Instruction-Guided Attention (IGA), which measures how strongly the instruction attends to image tokens via attention. Extensive experiments demonstrate that TransPrune achieves comparable multimodal performance to original LVLMs, such as LLaVA-v1.5 and LLaVA-Next, across eight benchmarks, while reducing inference TFLOPs by more than half. Moreover, TTV alone can serve as an effective criterion without relying on attention, achieving performance comparable to attention-based methods. The code will be made publicly available upon acceptance of the paper at https://github.com/liaolea/TransPrune.

Not Only Grey Matter: OmniBrain for Robust Multimodal Classification of Alzheimer's Disease

Alzheimer's disease affects over 55 million people worldwide and is projected to more than double by 2050, necessitating rapid, accurate, and scalable diagnostics. However, existing approaches are limited because they cannot achieve clinically acceptable accuracy, generalization across datasets, robustness to missing modalities, and explainability all at the same time. This inability to satisfy all these requirements simultaneously undermines their reliability in clinical settings. We propose OmniBrain, a multimodal framework that integrates brain MRI, radiomics, gene expression, and clinical data using a unified model with cross-attention and modality dropout. OmniBrain achieves $92.2 \pm 2.4\%$accuracy on the ANMerge dataset and generalizes to the MRI-only ADNI dataset with $70.4 \pm 2.7\%$ accuracy, outperforming unimodal and prior multimodal approaches. Explainability analyses highlight neuropathologically relevant brain regions and genes, enhancing clinical trust. OmniBrain offers a robust, interpretable, and practical solution for real-world Alzheimer's diagnosis.

crossMoDA Challenge: Evolution of Cross-Modality Domain Adaptation Techniques for Vestibular Schwannoma and Cochlea Segmentation from 2021 to 2023

The cross-Modality Domain Adaptation (crossMoDA) challenge series, initiated in 2021 in conjunction with the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), focuses on unsupervised cross-modality segmentation, learning from contrast-enhanced T1 (ceT1) and transferring to T2 MRI. The task is an extreme example of domain shift chosen to serve as a meaningful and illustrative benchmark. From a clinical application perspective, it aims to automate Vestibular Schwannoma (VS) and cochlea segmentation on T2 scans for more cost-effective VS management. Over time, the challenge objectives have evolved to enhance its clinical relevance. The challenge evolved from using single-institutional data and basic segmentation in 2021 to incorporating multi-institutional data and Koos grading in 2022, and by 2023, it included heterogeneous routine data and sub-segmentation of intra- and extra-meatal tumour components. In this work, we report the findings of the 2022 and 2023 editions and perform a retrospective analysis of the challenge progression over the years. The observations from the successive challenge contributions indicate that the number of outliers decreases with an expanding dataset. This is notable since the diversity of scanning protocols of the datasets concurrently increased. The winning approach of the 2023 edition reduced the number of outliers on the 2021 and 2022 testing data, demonstrating how increased data heterogeneity can enhance segmentation performance even on homogeneous data. However, the cochlea Dice score declined in 2023, likely due to the added complexity from tumour sub-annotations affecting overall segmentation performance. While progress is still needed for clinically acceptable VS segmentation, the plateauing performance suggests that a more challenging cross-modal task may better serve future benchmarking.

On the Robustness of Medical Vision-Language Models: Are they Truly Generalizable?

Medical Vision-Language Models (MVLMs) have achieved par excellence generalization in medical image analysis, yet their performance under noisy, corrupted conditions remains largely untested. Clinical imaging is inherently susceptible to acquisition artifacts and noise; however, existing evaluations predominantly assess generally clean datasets, overlooking robustness -- i.e., the model's ability to perform under real-world distortions. To address this gap, we first introduce MediMeta-C, a corruption benchmark that systematically applies several perturbations across multiple medical imaging datasets. Combined with MedMNIST-C, this establishes a comprehensive robustness evaluation framework for MVLMs. We further propose RobustMedCLIP, a visual encoder adaptation of a pretrained MVLM that incorporates few-shot tuning to enhance resilience against corruptions. Through extensive experiments, we benchmark 5 major MVLMs across 5 medical imaging modalities, revealing that existing models exhibit severe degradation under corruption and struggle with domain-modality tradeoffs. Our findings highlight the necessity of diverse training and robust adaptation strategies, demonstrating that efficient low-rank adaptation when paired with few-shot tuning, improves robustness while preserving generalization across modalities.

Kalman Filter Enhanced GRPO for Reinforcement Learning-Based Language Model Reasoning

Reward baseline is important for Reinforcement Learning (RL) algorithms to reduce variance in policy gradient estimates. Recently, for language modeling, Group Relative Policy Optimization (GRPO) is proposed to compute the advantage for each output by subtracting the mean reward, as the baseline, for all outputs in the group. However, it can lead to inaccurate advantage estimates in environments with highly noisy rewards, potentially introducing bias. In this work, we propose a model, called Kalman Filter Enhanced Group Relative Policy Optimization (KRPO), by using lightweight Kalman filtering to dynamically estimate the latent reward mean and variance. This filtering technique replaces the naive batch mean baseline, enabling more adaptive advantage normalization. Our method does not require additional learned parameters over GRPO. This approach offers a simple yet effective way to incorporate multiple outputs of GRPO into advantage estimation, improving policy optimization in settings where highly dynamic reward signals are difficult to model for language models. Through experiments and analyses, we show that using a more adaptive advantage estimation model, KRPO can improve the stability and performance of GRPO. The code is available at https://github.com/billhhh/KRPO_LLMs_RL