皮肤外科诊疗多模态标注高质量数据集

本数据集聚焦皮肤外科诊疗场景，围绕皮肤病的临床鉴别、诊断与健康指导等核心需求构建，原始数据均从合作三甲医院合规采集，经专业团队完成数据脱敏、文件校对及标准化标注全流程处理，最终形成高质量、结构化的多模态医疗数据集，数据总量级达 500GB，可直接对接皮肤疾病 AI 辅助诊断模型训练、临床教学研究、智能问诊系统开发等各类应用场景，为皮肤外科领域的人工智能技术落地、临床诊疗优化与医疗产品创新提供高价值、高可靠性的数据支撑。
数据集为 “影像 + 文本” 多模态医疗数据，单条数据完整覆盖皮肤疾病诊疗全流程信息，核心包含皮肤皮损影像类数据与临床诊疗标注类数据两大类别。影像类数据为皮肤皮损实拍图像，完整保留临床采集的皮损形态、颜色、分布等关键视觉特征；临床诊疗标注类数据涵盖患者主诉、皮损重点描述、可能疾病及分型、问诊对话、最终诊断及诊断依据、健康建议等核心诊疗信息，字段信息完整覆盖皮肤疾病从症状采集、初步鉴别、问诊交互到确诊与健康指导的全链路流程，标注内容精准规范，与影像数据一一匹配，真实还原皮肤外科临床诊疗的实际场景与数据特征。

This dataset focuses on the dermatologic surgery clinical scenario, and is constructed around the core needs of clinical differential diagnosis, diagnosis and health guidance for skin diseases. The original data was collected compliantly from cooperative tertiary hospitals, and has undergone the entire process of data de-identification, document proofreading and standardized annotation by a professional team, finally forming a high-quality, structured multimodal medical dataset. With a total data volume of 500 gigabytes, it can be directly connected to various application scenarios such as training AI-assisted diagnostic models for skin diseases, clinical teaching and research, and development of intelligent consultation systems, providing high-value and highly reliable data support for the implementation of AI technology in the field of dermatologic surgery, optimization of clinical diagnosis and treatment, and innovation of medical products. This is an "image + text" multimodal medical dataset. Each complete data entry covers the entire workflow information of skin disease diagnosis and treatment, and mainly includes two categories: skin lesion image data and clinical diagnosis and treatment annotation data. The image data consists of real-shot images of skin lesions, which fully retain the key visual features collected clinically, such as lesion morphology, color and distribution. The clinical diagnosis and treatment annotation data covers core clinical information including patient's chief complaint, key description of lesions, possible diseases and their subtypes, consultation dialogues, final diagnosis and diagnostic basis, health advice and so on. The field information completely covers the end-to-end workflow of skin diseases from symptom collection, preliminary differential diagnosis, consultation interaction, to definitive diagnosis and health guidance. The annotation content is accurate and standardized, and is one-to-one matched with the image data, truly restoring the actual clinical scenario and data characteristics of dermatologic surgery.

搜集汇总

背景与挑战

背景概述

该数据集是针对皮肤外科诊疗场景构建的多模态医疗数据，包含皮肤皮损影像和临床诊疗标注信息，完整覆盖疾病鉴别、诊断与健康指导全流程。数据源自合作三甲医院，经脱敏和标准化处理，总量达500GB，适用于AI辅助诊断模型训练、临床教学研究等应用。