1 | Chunzheng Zhu

MedSynapse-V: Bridging Visual Perception and Clinical Intuition via Latent Memory Evolution

Tue, 12 May 2026 00:00:00 +0000

MedSynapse-V bridges visual perception and clinical intuition via latent diagnostic memory evolution for medical vision-language models.

Anatomy-Anchored Self-Supervision: Distilling Vision Foundation Models for Invariant Ultrasound Representation

Mon, 11 May 2026 00:00:00 +0000

AnaUS discovers unlabeled anatomical structures with an LP-SAM driven latent prompt engine and combines anatomy disentanglement alignment with core-region prediction for robust ultrasound representation learning.

Subspace-Guided Semantic and Topological Invariant Registration for Annotation-Free Ultrasound Plane Quality Control

Sun, 10 May 2026 00:00:00 +0000

STRIQ is a registration-driven annotation-free ultrasound plane quality control framework. It uses a latent registration aligner and orthogonal knowledge subspaces to improve correlation with clinical quality scoring.

When Models Learn to Ask Why: Adaptive Causal Reasoning for Trustworthy Medical Vision-Language Models

Tue, 24 Mar 2026 00:00:00 +0000

MedCausalX explicitly models causal reasoning chains in medical VLMs to reduce spurious correlations and improve trustworthy diagnosis.

MedEyes: Learning Dynamic Visual Focus for Medical Progressive Diagnosis

Sat, 14 Mar 2026 00:00:00 +0000

MedEyes dynamically models clinician-style diagnostic reasoning by progressively attending to and interpreting relevant medical image regions.

Pathology-Aware Prototype Evolution via LLM-Driven Semantic Disambiguation for Multicenter Diabetic Retinopathy Diagnosis

Wed, 01 Jan 2025 00:00:00 +0000

This work integrates fine-grained pathological descriptions with visual prototypes to resolve semantic ambiguity in multicenter diabetic retinopathy grading.

Advancing Ultrasound Medical Continuous Learning with Task-Specific Generalization and Adaptability

Tue, 03 Dec 2024 00:00:00 +0000

This work addresses catastrophic forgetting in ultrasound medical image analysis with masked ultrasound image modeling and contrastive historical-current learning.