A two-stage workflow for vitiligo diagnosis: clinical characteristic classification and large language model (LLM)–based report generation - Scorecard - MDSpire
Advertisement
A two-stage workflow for vitiligo diagnosis: clinical characteristic classification and large language model (LLM)–based report generation
Clinical Scorecard: A Dual-Phase Approach for Diagnosing Vitiligo: Classification of Clinical Features and Report Generation Using a Large Language Model (LLM)
At a Glance
Category
Detail
Condition
Vitiligo
Key Mechanisms
AI-assisted diagnostic system integrating a large language model for differentiating vitiligo from other hypopigmentary disorders.
Target Population
Patients diagnosed with vitiligo or similar hypopigmentary disorders.
Care Setting
Multicenter hospitals in China.
Key Highlights
Model achieved an AUC of 0.9906 for distinguishing vitiligo.
Sensitivity of 98.29% and specificity of 93.73%.
AI model significantly outperformed dermatologists in diagnostic sensitivity.
Structured clinical reports generated via DeepSeek LLM.
Accurate classification of eight key clinical characteristics.
Guideline-Based Recommendations
Diagnosis
Diagnosis relies on clinical assessment and Wood’s lamp examination.
Histopathological evaluation confirms diagnosis.
Management
Treatment strategies are influenced by accurate differential diagnosis.
Monitoring & Follow-up
Follow-up plans included in generated clinical reports.
Risks
Misdiagnosis linked to disparities in medical resources.
Patient & Prescribing Data
Patients with vitiligo or ten other hypopigmentary disorders.
