Machine Learning for Predicting Complications in Head and Neck Microvascular Free Tissue Transfer

Area of Science:

• Surgical Oncology
• Artificial Intelligence in Medicine
• Head and Neck Surgery

Background:

• Head and neck free tissue transfer is complex, with potential for surgical complications.
• Predicting these complications is crucial for optimizing patient care and outcomes.
• Traditional statistical models have limitations in identifying all significant predictive factors.

Purpose of the Study:

• To utilize a machine learning (ML) platform to identify key factors predicting surgical complications in head and neck free tissue transfer.
• To compare the predictive factors identified by ML with those from traditional logistic regression models.

Main Methods:

• A retrospective cohort study analyzed 364 patients undergoing head and neck free tissue transfer.
• A supervised ML algorithm (ensemble decision trees) was employed to predict surgical complications.
• Clinicopathologic characteristics were analyzed, and ML variable importance was compared to logistic regression.

Main Results:

• ML models achieved 65% to 75% accuracy in classifying outcomes.
• Key ML-identified predictors included institutional experience, flap ischemia time, age, and smoking pack-years.
• Traditional logistic regression primarily identified patient age, flap type, and primary reconstruction site.

Conclusions:

• Machine learning algorithms identified distinct predictors of complications after free tissue transfer compared to traditional regression models.
• This suggests ML offers a complementary approach to understanding and predicting surgical risks in this patient population.