Imran Razzak

Compact, automatically selected anatomical regions can drastically reduce hallucinations in medical VLMs without requiring expert annotations.

AsyTO achieves state-of-the-art forecasting accuracy while keeping model complexity linear, challenging the notion that more parameters always lead to better performance.

TopoBrick achieves superior zero-shot forecasting by intelligently sampling exogenous variables based on building topology, outperforming traditional models without the need for extensive training.

Retinal graph phenotypes can prioritize systemic pathways in diabetic retinopathy, revealing critical mediators like glycaemic–renal interactions that traditional methods overlook.

Hallucination rates drop to just 8.1% with TAVR-VLM, revolutionizing the reliability of AI-generated surgical reports.

RetiSEM achieves superior causal accuracy in fragmented biomedical data, revealing hidden indirect effects that traditional methods miss.

The dual-edge spatial-Jacobian image graph reveals intricate relationships between retinal lesions and vascular biomarkers, achieving up to 0.9055 accuracy in referable diabetic retinopathy grading.

NLICV not only speeds up LLM personalization evaluation by up to 2100 times but also offers a clearer understanding of model behaviors beyond simple binary scoring.

Tailoring medical image segmentation models to dataset-specific requirements can boost performance by over 10% compared to traditional architecture-first approaches.

Forget choosing just one vision encoder – fusing CLIP and DINO representations unlocks a significant performance boost in vision-language tasks.

Achieve state-of-the-art video polyp segmentation by adaptively selecting informative reference frames and aggregating multi-scale historical features with causal attention.

Medical VLMs get a calibration boost without training or labels: LATA sharpens predictions by smoothing over a k-NN graph, shrinking prediction sets and balancing class coverage.