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

Efficient Parameter Adaptation for Multi-Modal Medical Image Segmentation and Prognosis

Cancer detection and prognosis relies heavily on medical imaging, particularly CT and PET scans. Deep Neural Networks (DNNs) have shown promise in tumor segmentation by fusing information from these modalities. However, a critical bottleneck exists: the dependency on CT-PET data concurrently for training and inference, posing a challenge due to the limited availability of PET scans. Hence, there is a clear need for a flexible and efficient framework that can be trained with the widely available CT scans and can be still adapted for PET scans when they become available. In this work, we propose a parameter-efficient multi-modal adaptation (PEMMA) framework for lightweight upgrading of a transformer-based segmentation model trained only on CT scans such that it can be efficiently adapted for use with PET scans when they become available. This framework is further extended to perform prognosis task maintaining the same efficient cross-modal fine-tuning approach. The proposed approach is tested with two well-known segementation backbones, namely UNETR and Swin UNETR. Our approach offers two main advantages. Firstly, we leverage the inherent modularity of the transformer architecture and perform low-rank adaptation (LoRA) as well as decomposed low-rank adaptation (DoRA) of the attention weights to achieve parameter-efficient adaptation. Secondly, by minimizing cross-modal entanglement, PEMMA allows updates using only one modality without causing catastrophic forgetting in the other. Our method achieves comparable performance to early fusion, but with only 8% of the trainable parameters, and demonstrates a significant +28% Dice score improvement on PET scans when trained with a single modality. Furthermore, in prognosis, our method improves the concordance index by +10% when adapting a CT-pretrained model to include PET scans, and by +23% when adapting for both PET and EHR data.

SALT: Singular Value Adaptation with Low-Rank Transformation

The complex nature of medical image segmentation calls for models that are specifically designed to capture detailed, domain-specific features. Large foundation models offer considerable flexibility, yet the cost of fine-tuning these models remains a significant barrier. Parameter-Efficient Fine-Tuning (PEFT) methods, such as Low-Rank Adaptation (LoRA), efficiently update model weights with low-rank matrices but may suffer from underfitting when the chosen rank is insufficient to capture domain-specific nuances. Conversely, full-rank Singular Value Decomposition (SVD) based methods provide comprehensive updates by modifying all singular values, yet they often lack flexibility and exhibit variable performance across datasets. We propose SALT (Singular Value Adaptation with Low-Rank Transformation), a method that selectively adapts the most influential singular values using trainable scale and shift parameters while complementing this with a low-rank update for the remaining subspace. This hybrid approach harnesses the advantages of both LoRA and SVD, enabling effective adaptation without relying on increasing model size or depth. Evaluated on 5 challenging medical datasets, ranging from as few as 20 samples to 1000, SALT outperforms state-of-the-art PEFT (LoRA and SVD) by 2% to 5% in Dice with only 3.9% trainable parameters, demonstrating robust adaptation even in low-resource settings. The code for SALT is available at: https://github.com/BioMedIA-MBZUAI/SALT

Segmentation-Guided CT Synthesis with Pixel-Wise Conformal Uncertainty Bounds

Accurate dose calculations in proton therapy rely on high-quality CT images. While planning CTs (pCTs) serve as a reference for dosimetric planning, Cone Beam CT (CBCT) is used throughout Adaptive Radiotherapy (ART) to generate sCTs for improved dose calculations. Despite its lower cost and reduced radiation exposure advantages, CBCT suffers from severe artefacts and poor image quality, making it unsuitable for precise dosimetry. Deep learning-based CBCT-to-CT translation has emerged as a promising approach. Still, existing methods often introduce anatomical inconsistencies and lack reliable uncertainty estimates, limiting their clinical adoption. To bridge this gap, we propose STF-RUE, a novel framework integrating two key components. First, STF, a segmentation-guided CBCT-to-CT translation method that enhances anatomical consistency by leveraging segmentation priors extracted from pCTs. Second, RUE, a conformal prediction method that augments predicted CTs with pixel-wise conformal prediction intervals, providing clinicians with robust reliability indicator. Comprehensive experiments using UNet++ and Fast-DDPM on two benchmark datasets demonstrate that STF-RUE significantly improves translation accuracy, as measured by a novel soft-tissue-focused metric designed for precise dose computation. Additionally, STF-RUE provides better-calibrated uncertainty sets for synthetic CT, reinforcing trust in synthetic CTs. By addressing both anatomical fidelity and uncertainty quantification, STF-RUE marks a crucial step toward safer and more effective adaptive proton therapy. Code is available at https://anonymous.4open.science/r/cbct2ct_translation-B2D9/.

In-Model Merging for Enhancing the Robustness of Medical Imaging Classification Models

Model merging is an effective strategy to merge multiple models for enhancing model performances, and more efficient than ensemble learning as it will not introduce extra computation into inference. However, limited research explores if the merging process can occur within one model and enhance the model's robustness, which is particularly critical in the medical image domain. In the paper, we are the first to propose in-model merging (InMerge), a novel approach that enhances the model's robustness by selectively merging similar convolutional kernels in the deep layers of a single convolutional neural network (CNN) during the training process for classification. We also analytically reveal important characteristics that affect how in-model merging should be performed, serving as an insightful reference for the community. We demonstrate the feasibility and effectiveness of this technique for different CNN architectures on 4 prevalent datasets. The proposed InMerge-trained model surpasses the typically-trained model by a substantial margin. The code will be made public.

FetalCLIP: A Visual-Language Foundation Model for Fetal Ultrasound Image Analysis

Foundation models are becoming increasingly effective in the medical domain, offering pre-trained models on large datasets that can be readily adapted for downstream tasks. Despite progress, fetal ultrasound images remain a challenging domain for foundation models due to their inherent complexity, often requiring substantial additional training and facing limitations due to the scarcity of paired multimodal data. To overcome these challenges, here we introduce FetalCLIP, a vision-language foundation model capable of generating universal representation of fetal ultrasound images. FetalCLIP was pre-trained using a multimodal learning approach on a diverse dataset of 210,035 fetal ultrasound images paired with text. This represents the largest paired dataset of its kind used for foundation model development to date. This unique training approach allows FetalCLIP to effectively learn the intricate anatomical features present in fetal ultrasound images, resulting in robust representations that can be used for a variety of downstream applications. In extensive benchmarking across a range of key fetal ultrasound applications, including classification, gestational age estimation, congenital heart defect (CHD) detection, and fetal structure segmentation, FetalCLIP outperformed all baselines while demonstrating remarkable generalizability and strong performance even with limited labeled data. We plan to release the FetalCLIP model publicly for the benefit of the broader scientific community.

Noise is an Efficient Learner for Zero-Shot Vision-Language Models

Recently, test-time adaptation has garnered attention as a method for tuning models without labeled data. The conventional modus operandi for adapting pre-trained vision-language models (VLMs) during test-time primarily focuses on tuning learnable prompts; however, this approach overlooks potential distribution shifts in the visual representations themselves. In this work, we address this limitation by introducing Test-Time Noise Tuning (TNT), a novel method for handling unpredictable shifts in the visual space. TNT leverages, for the first time, a noise adaptation strategy that optimizes learnable noise directly in the visual input space, enabling adaptive feature learning from a single test sample. We further introduce a novel approach for inter-view representation alignment by explicitly enforcing coherence in embedding distances, ensuring consistent feature representations across views. Combined with scaled logits and confident view selection at inference, TNT substantially enhances VLM generalization and calibration, achieving average gains of +7.38% on natural distributions benchmark and +0.80% on cross-dataset evaluations over zero-shot CLIP. These improvements lay a strong foundation for adaptive out-of-distribution handling.

PulmoFusion: Advancing Pulmonary Health with Efficient Multi-Modal Fusion

Traditional remote spirometry lacks the precision required for effective pulmonary monitoring. We present a novel, non-invasive approach using multimodal predictive models that integrate RGB or thermal video data with patient metadata. Our method leverages energy-efficient Spiking Neural Networks (SNNs) for the regression of Peak Expiratory Flow (PEF) and classification of Forced Expiratory Volume (FEV1) and Forced Vital Capacity (FVC), using lightweight CNNs to overcome SNN limitations in regression tasks. Multimodal data integration is improved with a Multi-Head Attention Layer, and we employ K-Fold validation and ensemble learning to boost robustness. Using thermal data, our SNN models achieve 92% accuracy on a breathing-cycle basis and 99.5% patient-wise. PEF regression models attain Relative RMSEs of 0.11 (thermal) and 0.26 (RGB), with an MAE of 4.52% for FEV1/FVC predictions, establishing state-of-the-art performance. Code and dataset can be found on https://github.com/ahmed-sharshar/RespiroDynamics.git

Optimizing Brain Tumor Segmentation with MedNeXt: BraTS 2024 SSA and Pediatrics

Identifying key pathological features in brain MRIs is crucial for the long-term survival of glioma patients. However, manual segmentation is time-consuming, requiring expert intervention and is susceptible to human error. Therefore, significant research has been devoted to developing machine learning methods that can accurately segment tumors in 3D multimodal brain MRI scans. Despite their progress, state-of-the-art models are often limited by the data they are trained on, raising concerns about their reliability when applied to diverse populations that may introduce distribution shifts. Such shifts can stem from lower quality MRI technology (e.g., in sub-Saharan Africa) or variations in patient demographics (e.g., children). The BraTS-2024 challenge provides a platform to address these issues. This study presents our methodology for segmenting tumors in the BraTS-2024 SSA and Pediatric Tumors tasks using MedNeXt, comprehensive model ensembling, and thorough postprocessing. Our approach demonstrated strong performance on the unseen validation set, achieving an average Dice Similarity Coefficient (DSC) of 0.896 on the BraTS-2024 SSA dataset and an average DSC of 0.830 on the BraTS Pediatric Tumor dataset. Additionally, our method achieved an average Hausdorff Distance (HD95) of 14.682 on the BraTS-2024 SSA dataset and an average HD95 of 37.508 on the BraTS Pediatric dataset. Our GitHub repository can be accessed here: Project Repository : https://github.com/python-arch/BioMbz-Optimizing-Brain-Tumor-Segmentation-with-MedNeXt-BraTS-2024-SSA-and-Pediatrics

Rethinking Weight-Averaged Model-merging

Weight-averaged model-merging has emerged as a powerful approach in deep learning, capable of enhancing model performance without fine-tuning or retraining. However, the underlying mechanisms that explain its effectiveness remain largely unexplored. In this paper, we investigate this technique from three novel perspectives to provide deeper insights into how and why weight-averaged model-merging works: (1) we examine the intrinsic patterns captured by the learning of the model weights, through the visualizations of their patterns on several datasets, showing that these weights often encode structured and interpretable patterns; (2) we investigate model ensemble merging strategies based on averaging on weights versus averaging on features, providing detailed analyses across diverse architectures and datasets; and (3) we explore the impact on model-merging prediction stability in terms of changing the parameter magnitude, revealing insights into the way of weight averaging works as regularization by showing the robustness across different parameter scales. Our findings shed light on the "black box" of weight-averaged model-merging, offering valuable insights and practical recommendations that advance the model-merging process.