Sheethal Bhat

Florian Kittler, Sheethal Bhat, Andreas Maier

Zero-shot vision-language models (VLMs) have shown promise for chest radiograph classification, but their performance is often limited by confounding label co-occurrence, long-tail class imbalance, and transfer instability under domain shift. We propose ProtoCLIP, a refinement strategy for CLIP-style VLMs that improves zero-shot discrimination through targeted data curation and distilled anchor alignment. Specifically, we construct pathology-focused training subsets with curated negative samples to reduce co-occurrence bias. We also introduce a representation-preserving distillation objective to stabilize adaptation while maintaining semantic structure and improving discrimination of clinically relevant co-occurring pathologies. Evaluated on an unseen dataset VinDr-CXR, ProtoCLIP improves AUC by 2-10 percentage points over a strong CLIP-based baseline across multiple findings. For pneumothorax specifically, ProtoCLIP achieves a state-of-the-art AUC of 0.94. These results demonstrate that anchor-guided refinement, coupled with curated supervision and controlled adaptation, can mitigate common zero-shot transfer failures in medical VLMs without requiring large-scale retraining.

Abdelrahman Zaian, Sheethal Bhat, Mohamed Abdalkader et al.

Diabetic Retinopathy (DR) is a leading cause of preventable blindness among working-age adults worldwide, yet most automated screening systems are limited to image-level classification and lack clinically structured reporting. We propose Retina-RAG, a low-cost modular framework that jointly performs DR severity grading, macular edema (ME) detection, and report generation. The architecture decouples a high-performance retinal classifier and a parameter-efficient vision-language model (Qwen2.5-VL-7B-Instruct) adapted via Low-Rank Adaptation (LoRA), enabling flexible component integration. A retrieval-augmented generation (RAG) module injects curated ophthalmic knowledge together with structured classifier outputs at inference time to improve diagnostic consistency and reduce hallucinations. Retina-RAG achieves an F1-score of 0.731 for DR grading and 0.948 for ME detection, substantially outperforming zero-shot Qwen (0.096, 0.732) and MMed-RAG (0.541, 0.641) on a retinal disease detection dataset with captions. For report generation, Retina-RAG attains ROUGE-L 0.429 and SBERT similarity 0.884, exceeding all baselines. The full framework operates on a single consumer-grade GPU, demonstrating that clinically structured retinal AI can be achieved with modest computational resources.

Mohammad Robaitul Islam Bhuiyan, Sheethal Bhat, Melika Qahqaie et al.

Precise localization and delineation of brain tumors using Magnetic Resonance Imaging (MRI) are essential for planning therapy and guiding surgical decisions. However, most existing approaches rely on task-specific supervised models and are constrained by the limited availability of annotated data. To address this, we propose LoGSAM, a parameter-efficient, detection-driven framework that transforms radiologist dictation into text prompts for foundation-model-based localization and segmentation. Radiologist speech is first transcribed and translated using a pretrained Whisper ASR model, followed by negation-aware clinical NLP to extract tumor-specific textual prompts. These prompts guide text-conditioned tumor localization via a LoRA-adapted vision-language detection model, Grounding DINO (GDINO). The LoRA adaptation updates using 5% of the model parameters, thereby enabling computationally efficient domain adaptation while preserving pretrained cross-modal knowledge. The predicted bounding boxes are used as prompts for MedSAM to generate pixel-level tumor masks without any additional fine-tuning. Conditioning the frozen MedSAM on LoGSAM-derived priors yields a state-of-the-art dice score of 80.32% on BRISC 2025. In addition, we evaluate the full pipeline using German dictations from a board-certified radiologist on 12 unseen MRI scans, achieving 91.7% case-level accuracy. These results highlight the feasibility of constructing a modular, speech-to-segmentation pipeline by intelligently leveraging pretrained foundation models with minimal parameter updates.

Lukas Bayer, Sheethal Bhat, Andreas Maier

Accurate multi-organ segmentation in abdominal CT scans is essential for computer-aided diagnosis and treatment. While convolutional neural networks (CNNs) have long been the standard approach in medical image segmentation, transformer-based architectures have recently gained attention due to their ability to model long-range dependencies. In this study, we systematically benchmark the three hybrid transformer-based models UNETR, SwinUNETR, and UNETR++ against a strong CNN baseline, SegResNet, for volumetric multi-organ segmentation on the heterogeneous RATIC dataset. The dataset comprises 206 annotated CT scans from 23 institutions worldwide, covering five abdominal organs. All models were trained and evaluated under identical preprocessing and training conditions using the Dice Similarity Coefficient (DSC) as the primary metric. The results show that the CNN-based SegResNet achieves the highest overall performance, outperforming all hybrid transformer-based models across all organs. Among the transformer-based approaches, UNETR++ delivers the most competitive results, while UNETR demonstrates notably faster convergence with fewer training iterations. These findings suggest that, for small- to medium-sized heterogeneous datasets, well-optimized CNN architectures remain highly competitive and may outperform hybrid transformer-based designs.

Adarsh Bhandary Panambur, Hui Yu, Sheethal Bhat et al.

The assessment of breast density is crucial in the context of breast cancer screening, especially in populations with a higher percentage of dense breast tissues. This study introduces a novel data augmentation technique termed Attention-Guided Erasing (AGE), devised to enhance the downstream classification of four distinct breast density categories in mammography following the BI-RADS recommendation in the Vietnamese cohort. The proposed method integrates supplementary information during transfer learning, utilizing visual attention maps derived from a vision transformer backbone trained using the self-supervised DINO method. These maps are utilized to erase background regions in the mammogram images, unveiling only the potential areas of dense breast tissues to the network. Through the incorporation of AGE during transfer learning with varying random probabilities, we consistently surpass classification performance compared to scenarios without AGE and the traditional random erasing transformation. We validate our methodology using the publicly available VinDr-Mammo dataset. Specifically, we attain a mean F1-score of 0.5910, outperforming values of 0.5594 and 0.5691 corresponding to scenarios without AGE and with random erasing (RE), respectively. This superiority is further substantiated by t-tests, revealing a p-value of p<0.0001, underscoring the statistical significance of our approach.

Shivam Chaudhary, Sheethal Bhat, Andreas Maier

Accurate detection and localization of traumatic injuries in abdominal CT scans remains a critical challenge in emergency radiology, primarily due to severe scarcity of annotated medical data. This paper presents a label-efficient approach combining self-supervised pre-training with semi-supervised detection for 3D medical image analysis. We employ patch-based Masked Image Modeling (MIM) to pre-train a 3D U-Net encoder on 1,206 CT volumes without annotations, learning robust anatomical representations. The pretrained encoder enables two downstream clinical tasks: 3D injury detection using VDETR with Vertex Relative Position Encoding, and multi-label injury classification. For detection, semi-supervised learning with 2,000 unlabeled volumes and consistency regularization achieves 56.57% validation mAP@0.50 and 45.30% test mAP@0.50 with only 144 labeled training samples, representing a 115% improvement over supervised-only training. For classification, expanding to 2,244 labeled samples yields 94.07% test accuracy across seven injury categories using only a frozen encoder, demonstrating immediately transferable self-supervised features. Our results validate that self-supervised pre-training combined with semi-supervised learning effectively addresses label scarcity in medical imaging, enabling robust 3D object detection with limited annotations.

Sheethal Bhat, Bogdan Georgescu, Adarsh Bhandary Panambur et al.

Detecting abnormalities in medical images poses unique challenges due to differences in feature representations and the intricate relationship between anatomical structures and abnormalities. This is especially evident in mammography, where dense breast tissue can obscure lesions, complicating radiological interpretation. Despite leveraging anatomical and semantic context, existing detection methods struggle to learn effective class-specific features, limiting their applicability across different tasks and imaging modalities. In this work, we introduce Exemplar Med-DETR, a novel multi-modal contrastive detector that enables feature-based detection. It employs cross-attention with inherently derived, intuitive class-specific exemplar features and is trained with an iterative strategy. We achieve state-of-the-art performance across three distinct imaging modalities from four public datasets. On Vietnamese dense breast mammograms, we attain an mAP of 0.7 for mass detection and 0.55 for calcifications, yielding an absolute improvement of 16 percentage points. Additionally, a radiologist-supported evaluation of 100 mammograms from an out-of-distribution Chinese cohort demonstrates a twofold gain in lesion detection performance. For chest X-rays and angiography, we achieve an mAP of 0.25 for mass and 0.37 for stenosis detection, improving results by 4 and 7 percentage points, respectively. These results highlight the potential of our approach to advance robust and generalizable detection systems for medical imaging.

Rajesh Madhipati, Sheethal Bhat, Lukas Buess et al.

Chest radiography (CXR) plays a crucial role in the diagnosis of various diseases. However, the inherent class imbalance in the distribution of clinical findings presents a significant challenge for current self-supervised deep learning models. These models often fail to accurately classify long-tailed classes. Current Vision-Language models such as Contrastive Language Image Pre-training (CLIP) models effectively model the manifold distribution of the latent space, enabling high zero-shot classification accuracies. Although CLIP performs well on most of the primary classes in the dataset, our work reveals that its effectiveness decreases significantly for classes with a long-tailed distribution. Our approach employs a class-weighting mechanism that directly aligns with the distribution of classes within the latent space. This method ensures a substantial improvement in overall classification performance, with particular emphasis on enhancing the recognition and accuracy of rarely observed classes. We accomplish this by applying Gaussian Mixture Model (GMM) clustering to the latent space. The subsequent clusters are further refined by Student t-distribution, followed by a metric loss that utilizes the altered embeddings. Our approach facilitates stable and adaptive clustering of the features. This results in a notable average improvement of 7\% points in zero-shot AUC scores across 40 classes in the MIMIC-CXR-JPG dataset from previous SOTA models.

Prakhar Bhardwaj, Sheethal Bhat, Andreas Maier

Due to the large volume of medical imaging data, advanced AI methodologies are needed to assist radiologists in diagnosing thoracic diseases from chest X-rays (CXRs). Existing deep learning models often require large, labeled datasets, which are scarce in medical imaging due to the time-consuming and expert-driven annotation process. In this paper, we extend the existing approach to enhance zero-shot learning in medical imaging by integrating Contrastive Language-Image Pre-training (CLIP) with Momentum Contrast (MoCo), resulting in our proposed model, MoCoCLIP. Our method addresses challenges posed by class-imbalanced and unlabeled datasets, enabling improved detection of pulmonary pathologies. Experimental results on the NIH ChestXray14 dataset demonstrate that MoCoCLIP outperforms the state-of-the-art CheXZero model, achieving relative improvement of approximately 6.5%. Furthermore, on the CheXpert dataset, MoCoCLIP demonstrates superior zero-shot performance, achieving an average AUC of 0.750 compared to CheXZero with 0.746 AUC, highlighting its enhanced generalization capabilities on unseen data.

Maged Badawi, Mohammedyahia Abushanab, Sheethal Bhat et al.

In this paper, different techniques of few-shot, zero-shot, and regular object detection have been investigated. The need for few-shot learning and zero-shot learning techniques is crucial and arises from the limitations and challenges in traditional machine learning, deep learning, and computer vision methods where they require large amounts of data, plus the poor generalization of those traditional methods. Those techniques can give us prominent results by using only a few training sets reducing the required amounts of data and improving the generalization. This survey will highlight the recent papers of the last three years that introduce the usage of few-shot learning and zero-shot learning techniques in addressing the challenges mentioned earlier. In this paper we reviewed the Zero-shot, few-shot and regular object detection methods and categorized them in an understandable manner. Based on the comparison made within each category. It been found that the approaches are quite impressive. This integrated review of diverse papers on few-shot, zero-shot, and regular object detection reveals a shared focus on advancing the field through novel frameworks and techniques. A noteworthy observation is the scarcity of detailed discussions regarding the difficulties encountered during the development phase. Contributions include the introduction of innovative models, such as ZSD-YOLO and GTNet, often showcasing improvements with various metrics such as mean average precision (mAP),Recall@100 (RE@100), the area under the receiver operating characteristic curve (AUROC) and precision. These findings underscore a collective move towards leveraging vision-language models for versatile applications, with potential areas for future research including a more thorough exploration of limitations and domain-specific adaptations.