Machine learning directed sentinel lymph node biopsy in cutaneous head and neck melanoma

Area of Science:

• Oncology
• Medical Informatics
• Surgical Oncology

Background:

• Sentinel lymph node biopsy (SLNB) has limited specificity for detecting lymph node metastasis in head and neck melanoma (HNM) under current National Comprehensive Cancer Network (NCCN) guidelines.
• Accurate staging is crucial for HNM treatment decisions.

Purpose of the Study:

• To develop and validate machine learning (ML) algorithms for identifying HNM patients at very low risk of occult nodal metastasis.
• To compare the performance of ML algorithm recommendations against NCCN guidelines for SLNB.

Main Methods:

• Developed multiple ML algorithms using National Cancer Database (NCDB) data from 8466 clinically node-negative HNM patients who underwent SLNB.
• Compared SLNB performance under NCCN guidelines and ML recommendations using independent test data from NCDB (n=2117) and an academic medical center (n=96).

Main Results:

• The top-performing ML algorithm achieved an AUC of 0.734.
• ML recommendations demonstrated significantly higher specificity than NCCN guidelines in both internal (25.8% vs. 11.3%) and external (30.1% vs. 7.1%) test populations.
• ML algorithms maintained high sensitivity (>97%) in identifying patients who would benefit from SLNB.

Conclusions:

• Machine learning effectively identifies clinically node-negative HNM patients at very low risk for nodal metastasis.
• These ML models can help refine SLNB decision-making, potentially sparing select patients from the procedure.