arXiv subjects

Yasmeen George

Publications and source records attributed to Yasmeen George.

10 recordsLinked to original sources

arXiv2026

NeRD: Neuro-Symbolic Rule Distillation for Efficient Ontology-Grounded Chain-of-Thought in Medical Image Diagnosis

Interpretability is essential for trustworthy medical image diagnosis. However, existing concept-driven interpretable methods have key limitations: Concept Bottleneck Models (CBMs) require scoring all predefined concepts at inference time and for manual intervention, imposing a substantial burden on clinicians, while rationale-based generative approaches often select concepts by class discriminability, which can drift from diagnostic ontologies. To address these issues, we propose Neuro-Symbolic Rule Distillation (NeRD), a framework that produces efficient, ontology-grounded reasoning chains that are sufficient yet non-redundant, without manually crafting diagnostic rules. Experiments on two skin datasets demonstrate strong diagnostic performance and interpretability, and blinded expert evaluation confirms the clinical plausibility of NeRD rationales. Our method further enables a first expert-in-the-loop study for Multimodal Chain-of-Thought-based diagnosis, achieving efficient and effective concept-level intervention.

cs.CV↗

arXiv2026

Single-Rollout Hidden-State Dynamics for Training-Free RLVR Data Selection

Reinforcement learning with verifiable rewards (RLVR) can yield large reasoning gains from very few training instances, yet its strong sensitivity to which instances are used makes data selection a central bottleneck. Most existing selection pipelines rely on training-time optimization signals and/or require access to verifiable rewards or ground-truth answers over large candidate pools, which is costly and often infeasible in specialized domains. We study RLVR data selection in a setting where selection must be performed before any RL training and without labels or reward evaluation on the full pool. We propose SHIFT, a one-shot, training-free selector based solely on inference-time hidden-state dynamics. For each candidate instance, SHIFT runs a single deterministic reasoning rollout and computes a reasoning-induced representation shift (RIRS) as the start-to-end hidden-state delta. SHIFT uses the RIRS magnitude as a lightweight proxy for instance utility and enforces coverage via a quality-weighted farthest-first CoreSet procedure in an RIRS-augmented feature space, producing compact subsets that scale to large unlabeled pools. Across mathematical reasoning and medical QA benchmarks under ultra-low budgets, SHIFT consistently outperforms training-free diversity and difficulty/uncertainty baselines, improving both in-domain accuracy and transfer to harder evaluation settings. Ablations show that RIRS-based coverage and quality-weighting contribute complementary gains, and analyses indicate that RIRS is not explained by simple input/output length statistics. Code is available at github.com/JianghaoWu/SHIFT.

cs.LG↗

arXiv2026

VIHD: Visual Intervention-based Hallucination Detection for Medical Visual Question Answering

While medical Multimodal Large Language Models (MLLMs) have shown promise in assisting diagnosis, they still frequently generate hallucinated responses that appear linguistically plausible but lack visual evidence. Such hallucinations pose risks to clinical decision-making and necessitate effective detection. Existing introspective detection methods primarily perform uncertainty estimation or logical verification by analyzing model responses conditioned on original or perturbed inputs. However, such external perturbations are often heuristic and context-agnostic, which overlooks the internal cross-modal dependency between generated tokens and related visual tokens during decoding. To address this issue, we propose VIHD, a Visual Intervention-based Hallucination Detection method that leverages targeted visual token masking to calibrate semantic entropy for more effective hallucination detection. VIHD locates visually dominant decoder layers via Visual Dependency Probing (VDP), executes Visual Intervention Decoding (VID) via token masking to calibrate the semantic distribution, and quantifies the resulting Calibrated Semantic Entropy (CSE) as a reliable hallucination signal. Extensive experiments on three medical VQA benchmarks with two medical MLLMs demonstrate that VIHD consistently outperforms state-of-the-art methods, underscoring the importance of fine-grained visual dependency for hallucination detection. The code will be available at https://github.com/Jiayi-Chen-AU/VIHD

cs.CV↗

arXiv2026

EviATTA: Evidential Active Test-Time Adaptation for Medical Segment Anything Models

Deploying foundational medical Segment Anything Models (SAMs) via test-time adaptation (TTA) is challenging under large distribution shifts, where test-time supervision is often unreliable. While active test-time adaptation (ATTA) introduces limited expert feedback to improve reliability, existing ATTA methods still suffer from unreliable uncertainty estimation and inefficient utilization of sparse annotations. To address these issues, we propose Evidential Active Test-Time Adaptation (EviATTA), which is, to our knowledge, the first ATTA framework tailored for medical SAMs. Specifically, we adopt the Dirichlet-based Evidential Modeling to decompose overall predictive uncertainty into distribution uncertainty and data uncertainty. Building on this decomposition, we design a Hierarchical Evidential Sampling strategy, where image-wise distribution uncertainty is used to select informative shifted samples, while distance-aware data uncertainty guides sparse pixel annotations to resolve data ambiguities. We further introduce Dual Consistency Regularization, which enforces progressive prompt consistency on sparsely labeled samples to better exploit sparse supervision and applies variational feature consistency on unlabeled samples to stabilize adaptation. Extensive experiments on six medical image segmentation datasets demonstrate that EviATTA consistently improves adaptation reliability with minimal expert feedback under both batch-wise and instance-wise test-time adaptation settings.

cs.CV↗

arXiv2026

SPINE: Token-Selective Test-Time Reinforcement Learning with Entropy-Band Regularization

Large language models (LLMs) and multimodal LLMs (MLL-Ms) excel at chain-of-thought reasoning but face distribution shift at test-time and a lack of verifiable supervision. Recent test-time reinforcement learning (TTRL) methods derive label-free pseudo-rewards from self-consistency voting over sampled trajectories, yet they often collapse: the majority-vote reward prevails, responses shorten, and Pass@1 declines. We trace this to uniform sequence updates in which most tokens are low-entropy followers, while a small high-entropy subset determines the reasoning branches. Thus we propose \method, a token-selective test-time reinforcement learning framework that (i) performs distribution-aware forking-token selection to update only decision-critical branch points, and (ii) applies a robust entropy-band regularizer at those tokens to prevent premature collapse and suppress noisy drift. \method plugs into GRPO-style objectives (optionally with a KL anchor) and requires neither labels nor reward models. Across eight benchmarks spanning multimodal VQA, text-only reasoning, \method consistently improves Pass@1 over TTRL while avoiding response-length collapse and yielding more stable training dynamics on both LLM and MLLM backbones. These results indicate that aligning updates with chain-of-thought branch points is a simple and label-free mechanism for stable and effective test-time adaptation in reasoning models. Code will be released.

cs.CL↗

arXiv2026

SAM-aware Test-time Adaptation for Universal Medical Image Segmentation

Leveraging the Segment Anything Model (SAM) for medical image segmentation remains challenging due to its limited adaptability across diverse medical domains. Although fine-tuned variants, such as MedSAM, improve performance in scenarios similar to the training modalities or organs, they may lack generalizability to unseen data. To overcome this limitation, we propose SAM-aware Test-time Adaptation (SAM-TTA), a lightweight and flexible framework that preserves SAM's inherent generalization ability while enhancing segmentation accuracy for medical images. SAM-TTA tackles two major challenges: (1) input-level discrepancy caused by channel mismatches between natural and medical images, and (2) semantic-level discrepancy due to different object characteristics in natural versus medical images (e.g., with clear boundaries vs. ambiguous structures). To this end, we introduce two complementary components: a self-adaptive Bezier Curve-based Transformation (SBCT), which maps single-channel medical images into SAM-compatible three-channel images via a few learnable parameters to be optimized at test time; and IoU-guided Multi-scale Adaptation (IMA), which leverages SAM's intrinsic IoU scores to enforce high output confidence, dual-scale prediction consistency, and intermediate feature consistency, to improve semantic-level alignments. Extensive experiments on eight public medical image segmentation tasks, covering six grayscale and two color (endoscopic) tasks, demonstrate that SAM-TTA consistently outperforms state-of-the-art test-time adaptation methods. Notably, on six grayscale datasets, SAM-TTA even surpasses fully fine-tuned models, achieving significant Dice improvements (i.e., average 4.8% and 7.4% gains over MedSAM and SAM-Med2D) and establishing a new paradigm for universal medical image segmentation. Code is available at https://github.com/JianghaoWu/SAM-TTA.

cs.CV↗

arXiv2024

A Physics-based Generative Model to Synthesize Training Datasets for MRI-based Fat Quantification

Deep learning-based techniques have potential to optimize scan and post-processing times required for MRI-based fat quantification, but they are constrained by the lack of large training datasets. Generative models are a promising tool to perform data augmentation by synthesizing realistic datasets. However no previous methods have been specifically designed to generate datasets for quantitative MRI (q-MRI) tasks, where reference quantitative maps and large variability in scanning protocols are usually required. We propose a Physics-Informed Latent Diffusion Model (PI-LDM) to synthesize quantitative parameter maps jointly with customizable MR images by incorporating the signal generation model. We assessed the quality of PI-LDM's synthesized data using metrics such as the Fréchet Inception Distance (FID), obtaining comparable scores to state-of-the-art generative methods (FID: 0.0459). We also trained a U-Net for the MRI-based fat quantification task incorporating synthetic datasets. When we used a few real (10 subjects, $~200$ slices) and numerous synthetic samples ($>3000$), fat fraction at specific liver ROIs showed a low bias on data obtained using the same protocol than training data ($0.10\%$ at $\hbox{ROI}_1$, $0.12\%$ at $\hbox{ROI}_2$) and on data acquired with an alternative protocol ($0.14\%$ at $\hbox{ROI}_1$, $0.62\%$ at $\hbox{ROI}_2$). Future work will be to extend PI-LDM to other q-MRI applications.

eess.SP↗

arXiv2024

Deep kernel representations of latent space features for low-dose PET-MR imaging robust to variable dose reduction

Low-dose positron emission tomography (PET) image reconstruction methods have potential to significantly improve PET as an imaging modality. Deep learning provides a promising means of incorporating prior information into the image reconstruction problem to produce quantitatively accurate images from compromised signal. Deep learning-based methods for low-dose PET are generally poorly conditioned and perform unreliably on images with features not present in the training distribution. We present a method which explicitly models deep latent space features using a robust kernel representation, providing robust performance on previously unseen dose reduction factors. Additional constraints on the information content of deep latent features allow for tuning in-distribution accuracy and generalisability. Tests with out-of-distribution dose reduction factors ranging from $\times 10$ to $\times 1000$ and with both paired and unpaired MR, demonstrate significantly improved performance relative to conventional deep-learning methods trained using the same data. Code:https://github.com/cameronPain

cs.CV↗

arXiv2023

The KiTS21 Challenge: Automatic segmentation of kidneys, renal tumors, and renal cysts in corticomedullary-phase CT

This paper presents the challenge report for the 2021 Kidney and Kidney Tumor Segmentation Challenge (KiTS21) held in conjunction with the 2021 international conference on Medical Image Computing and Computer Assisted Interventions (MICCAI). KiTS21 is a sequel to its first edition in 2019, and it features a variety of innovations in how the challenge was designed, in addition to a larger dataset. A novel annotation method was used to collect three separate annotations for each region of interest, and these annotations were performed in a fully transparent setting using a web-based annotation tool. Further, the KiTS21 test set was collected from an outside institution, challenging participants to develop methods that generalize well to new populations. Nonetheless, the top-performing teams achieved a significant improvement over the state of the art set in 2019, and this performance is shown to inch ever closer to human-level performance. An in-depth meta-analysis is presented describing which methods were used and how they faired on the leaderboard, as well as the characteristics of which cases generally saw good performance, and which did not. Overall KiTS21 facilitated a significant advancement in the state of the art in kidney tumor segmentation, and provides useful insights that are applicable to the field of semantic segmentation as a whole.

cs.CV↗

arXiv2021

Real-time Spatio-temporal Event Detection on Geotagged Social Media

A key challenge in mining social media data streams is to identify events which are actively discussed by a group of people in a specific local or global area. Such events are useful for early warning for accident, protest, election or breaking news. However, neither the list of events nor the resolution of both event time and space is fixed or known beforehand. In this work, we propose an online spatio-temporal event detection system using social media that is able to detect events at different time and space resolutions. First, to address the challenge related to the unknown spatial resolution of events, a quad-tree method is exploited in order to split the geographical space into multiscale regions based on the density of social media data. Then, a statistical unsupervised approach is performed that involves Poisson distribution and a smoothing method for highlighting regions with unexpected density of social posts. Further, event duration is precisely estimated by merging events happening in the same region at consecutive time intervals. A post processing stage is introduced to filter out events that are spam, fake or wrong. Finally, we incorporate simple semantics by using social media entities to assess the integrity, and accuracy of detected events. The proposed method is evaluated using different social media datasets: Twitter and Flickr for different cities: Melbourne, London, Paris and New York. To verify the effectiveness of the proposed method, we compare our results with two baseline algorithms based on fixed split of geographical space and clustering method. For performance evaluation, we manually compute recall and precision. We also propose a new quality measure named strength index, which automatically measures how accurate the reported event is.

cs.SI↗