MedPixel: A Unified Pixel-Language Model for Medical Reasoning and Segmentation

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Haoyu Yang, Meixing Shi, Zengjie Chen, Haoran Sun, Haitao Leng, Xiaoming Shi, Yuxiang Cai, Yankai Jiang
• Venue: arXiv: cs.CV
• Published: 2026-08-10
• Collected: 2026-08-11
• URL: http://arxiv.org/abs/2608.09818v1
• PDF: https://arxiv.org/pdf/2608.09818v1
• DOI: UNKNOWN
• Registry group: academic_apis_and_preprints
• Evidence role: canonical_academic_index
• Language: en
• Country / region: UNKNOWN

Plain-Language Brief

Researchers present MedPixel, a unified medical pixel-language model that bridges the gap between clinical language and visual reasoning in medical image understanding. The model is trained using MedPLG-440K, a large dataset of pixel-language task samples created through a synthesis process. MedPixel performs well across a range of tasks and shows strong performance in both pixel-level prediction and response generation.

What the Document Says

• MedPixel is a unified medical pixel-language model for medical image understanding.
• It addresses the gap between medical vision-language models and segmenters by providing a shared language--mask interface.
• MedPLG-440K, comprising approximately 440K pixel-language task samples, is introduced for scalable supervision.
• The model is trained using joint multi-task supervised fine-tuning followed by Pixel-Level Preference Optimization.
• MedPixel supports tasks spanning explicit grounding, implicit reasoning, spatial interaction, grounded explanation, and medical VQA.
• It achieves strong performance in both pixel-level prediction and response generation.
• The model shows effective zero-shot transfer to external grounding benchmarks and robustness to imperfect spatial prompts.
• Code and model checkpoints will be released at https://github.com/yhy-whu/Medpixel.

Key Concepts Mentioned

• Medical image understanding
• Pixel-language model
• MedPLG-440K
• Multi-task supervised fine-tuning
• Pixel-Level Preference Optimization
• Medical VQA

Claims Made by the Document

• MedPixel bridges the gap between clinical language and visual reasoning in medical image understanding.
• MedPixel achieves strong performance in both pixel-level prediction and response generation.
• MedPixel shows effective zero-shot transfer to external grounding benchmarks.
• MedPixel is robust to imperfect spatial prompts.

Evidence Used

• The text explains the purpose and functions of MedPixel and its unified approach.
• The text provides details about MedPLG-440K and its role in providing scalable supervision for MedPixel.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Reliable medical image understanding requires models to connect clinical language and visual reasoning with pixel-level grounding. Yet medical vision-language models often lack precise localization, whereas medical segmenters typically rely on explicit target categories or precise spatial prompts. This divide is reinforced by a supervision mismatch: segmentation datasets provide precise masks but little language supervision, whereas medical vision-language data rarely pair language with dense spatial annotations. To address this gap, we present MedPixel, a unified medical pixel-language model built around a shared language--mask interface. To provide scalable supervision, we introduce MedPLG-440K, comprising approximately 440K pixel-language task samples constructed through a clinically motivated synthesis process without external LLM annotation. MedPixel is trained with joint multi-task supervised fine-tuning followed by Pixel-Level Preference Optimization, which uses ground-truth masks as offline verifiers to derive response preferences from mask quality. MedPixel supports a broad spectrum of tasks spanning explicit grounding, implicit reasoning, spatial interaction, grounded explanation, and medical VQA. Across this task spectrum, MedPixel achieves strong performance in both pixel-level prediction and response generation, together with effective zero-shot transfer to external grounding benchmarks and robustness to imperfect spatial prompts. Code and model checkpoints will be released at https://github.com/yhy-whu/Medpixel.