Application Progress of Machine Learning Assisted Clinical Decision-making in Dermatology
10.11714/jsysu.med.YX20260019
- VernacularTitle:机器学习辅助临床决策在皮肤病中的应用进展
- Author:
Shucheng ZHANG
1
;
Xiaoqing SI
1
Author Information
1. Department of Dermatology， The First Affiliated Hospital of Shandong First Medical University， Jinan 250013， China
- Publication Type:Review
- Keywords:
machine learning;
dermatology;
clinical decision-making;
artificial intelligence-assisted diagnosis;
multimodal data
- From:
Journal of Sun Yat-sen University(Medical Sciences)
2026;47(2):195-202
- CountryChina
- Language:Chinese
-
Abstract:
Machine learning （ML） is increasingly being integrated into clinical decision-making in dermatology， penetrating multiple critical stages including lesion identification， multimodal differential diagnosis， personalized treatment recommendation， efficacy prediction， and prognosis evaluation. By synthesizing multi-source data such as medical imaging， genomics， and clinical characteristics， ML models not only assist clinicians in improving diagnostic accuracy but also optimize therapeutic selection， enabling dynamic disease management and individualized intervention， thereby demonstrating substantial clinical application potential. For instance， convolutional neural network-based image analysis systems have exhibited performance comparable to or exceeding that of dermatology experts in the recognition of cutaneous neoplasms. Regarding treatment optimization， ML can recommend personalized medication regimens， predict treatment responses and adverse event risks by analyzing multidimensional patient data， thereby providing robust support for precision medicine. In the realm of prognosis assessment and long-term management， ML models incorporating patient-reported outcomes and time-series data facilitate dynamic monitoring of disease progression and individualized prediction of recurrence risk. However， despite rapid technological advances， the practical translation of ML into dermatological clinical decision-making still confronts multiple bottlenecks， including data bias， insufficient algorithmic generalization， lack of robust clinical validation， difficulties in system integration， and incomplete ethical supervision. This review systematically outlines the current applications and research progress of ML across various stages of dermatological clinical decision-making， thoroughly analyzes the key bottlenecks encountered， and comprehensively synthesizes advances in ML-assisted clinical decision support spanning image recognition， treatment optimization， prognosis evaluation， and ethical challenges， aiming to provide valuable insights for related research and clinical practice.
