It is suggested that ten annotated cases are sufficient for clinically useful segmentation, effectively reducing bottlenecks for both image annotation and training time.
Abstract
Medical image segmentation is essential for clinical workflows such as treatment planning and disease assessment. While specialist tools like TotalSegmentator and MRSegmentator achieve strong performance, they require large annotated datasets for training. Medical foundation models offer a promising alternative through large-scale pretraining that reduces the annotation burden for new tasks, but zero-shot performance remains limited. Parameter-efficient adaptation via Low-Rank Adaptation (LoRA) enables efficient specialization with few trainable parameters, but a key question remains: how many expert-annotated cases are needed to achieve clinically useful segmentation performance? We address this by adapting MedSAM3 with LoRA for five abdominal organs (liver, kidneys, spleen, gallbladder, and pancreas) in CT and MRI using only 1, 2, 5, and 10 annotated cases, evaluating on AMOS22 dataset. With just 10 cases, models achieve performance competitive with specialist systems trained on orders of magnitude more data. Notably, this includes reliable gallbladder segmentation (Dice 0.68 CT, 0.59 MRI) where existing tools fail almost completely (Dice 0.0004), while remaining within 5--10% of MRSegmentator for liver, kidneys, and spleen using over 100 times fewer annotations. Furthermore, external validation on the Whole Heart Segmentation dataset shows that the approach extends to cardiac segmentation, a use case beyond the scope of TotalSegmentator (MRI) and MRSegmentator, achieving competitive left ventricle (LV) performance with only 10 annotated cases. Training requires only3--5,hours per organ on a single GPU, approximately 2--3 times faster than nnU-Net. These findings suggest that ten annotated cases are sufficient for clinically useful segmentation, effectively reducing bottlenecks for both image annotation and training time.
This survey extends beyond traditional and deep learning-based augmentation techniques or deep semi-supervised approaches, by explicitly focusing on medical/clinical imaging modalities, by explicitly focusing on CT, MRI, and X-ray, offering a broader perspective.
RegAL is proposed, a unified active semi-supervised framework governed by a shared topology-aware Pareto optimization that couples sample acquisition with unlabeled data utilization and consistently outperforms state-of-the-art AL, SSL, and active semi-supervised baselines across Dice and boundary-distance metrics under extreme annotation scarcity.
Bahram Jafrasteh, Cheng Wan, Heejong Kim et al.· 0 citations
The development of generalist artificial intelligence (AI) models for radiology is hindered by a lack of large-scale, three-dimensional (3D) imaging datasets with precise spatial annotations. While numerous datasets provide image-level labels for chest computed tomography (CT), these are insufficient for training models that can accurately localize findings. To address this gap, we present PatchChestCT, a large-scale, publicly available dataset for multi-abnormality localization in non-contrast chest CT. Using CT-RATE as the source cohort, PatchChestCT provides 3D patch-level annotations for nine clinically significant abnormalities across 2,201 physician-reviewed CT studies, with one reconstructed volume labeled per study. We introduce a token-aligned annotation scheme that is efficient, scalable, and naturally integrates with modern deep learning architectures like Transformers. To validate the dataset’s utility, we show that models trained with our patch-level labels achieved higher localization performance than weakly supervised baselines trained on image-level labels alone, across multiple architectures. PatchChestCT provides a public resource for training localization-aware models in chest CT.
Deploying unsupervised domain adaptation (UDA) in clinical practice requires choosing which algorithm to use and which of its trained models to ship. However, the deployment (target) domain is unlabeled, so models cannot be evaluated directly on it, leaving it unclear which to select. We address this by evaluating the complete UDA pipeline, considering both adaptation and label-free selection together. Our study covers eleven clinically relevant cross-domain scenarios from nine medical imaging datasets, with ten UDA algorithms and 13 label-free selection methods (validators), evaluating over 80,000 trained models in total. By this, we find that a capable adapted model usually exists, but identifying it without target labels is difficult: the validator-selected models leave a large and structural target performance gap to the best available one, with no evaluated validator consistently reliable. Towards closing it, we explore two strategies, ensembling and a small target-labeling budget; both narrow this gap but do not close it entirely. Overall, deployable UDA depends on the complete pipeline; addressing the less explored selection step could bring much of current UDA closer to clinical use.
The exponential growth in medical imaging volumes necessitates scalable, reliable diagnostic support systems capable of augmenting clinical workflows. This article presents a systematic quantitative evaluation of state-of-the-art Multimodal Large Language Models (MLLMs) for radiology Visual Question Answering (VQA), a task requiring integrated visual perception and clinical reasoning. We benchmark five leading models — GPT5-Nano, Gemini 3 Flash, Qwen3-VL-8B, LLaVA Next, and Llama 3.2 Vision — on the VQA-RAD dataset under a rigorous zero-shot protocol with standardized prompts and comprehensive precision–recall–F1 evaluation. Our empirical analysis reveals that Gemini 3 Flash achieves superior balanced performance (F1 = 0.78, Accuracy = 0.78, Recall = 0.83), while Qwen3-VL-8B attains the highest precision (0.78) while also maintaining competitive recall. These outcomes demonstrate that general-purpose MLLMs can perform competitively with specialized medical models in tasks such as modality and organ recognition, but still struggle with abnormality detection and complex clinical reasoning. The findings reinforce that MLLMs currently serve best as assistive decisionsupport tools rather than autonomous diagnostic agents, and highlight the potential of retrieval-augmented and context-aware strategies for improving clinical reliability and interpretability.
Cristovão Pessoa Cândido Neto, Matheus Alves de Oliveira Lima, Cláudio de Souza Baptista et al.· International Journal of Sem...· 0 citations
Lung ultrasound (LUS), valued for its portability and AI-driven analysis, has become an essential di agnostic tool in emergency medicine, critical care, and the screening of infectious diseases. However, the scarcity of annotated data and limited expert availability remain major barriers to large-scale deployment. In this study, we pro pose a Similarity-Guided Multi-Source Few-Shot Learning (SM-FSL) paradigm that shifts the focus from domain alignment to domain selection, enabling robust feature learning from multiple semantically relevant domains under limited data conditions. Based on this paradigm, we develop the Global-Local Multi-Source Domain Network (GLMD-Net) for LUS analysis in data-scarce scenarios, which first employs the proposed Average Nearest Neighbor Set Distance (ANNSD) to prioritize relevant source domains for pre-training. A similarity-guided optimization mechanism then harmonizes multi-source gradient updates, while a local feature enhancement module and a self-supervised auxiliary task improve robustness against noise and artifacts. Finally, a few-shot adaptation strategy fine-tunes only the classification head for efficient and stable knowledge transfer. Extensive experiments on two public COVID-19 ultrasound datasets demonstrate that our method surpasses state-of-the-art approaches, validating its effectiveness in enhancing generalization and robustness under few-shot settings.
Enguang Wang, Juxiang Wu, Shuo Gao et al.· IEEE journal of biomedical a...· 0 citations
Known for his clear and elegant writing style, Bertsekas shaped fields from control and optimization to large-scale computation and artificial intelligence.
