Prediction of ameloblastoma recurrence using random forest-a machine learning algorithm

The purpose of this study was to investigate whether ameloblastoma with a high likelihood of recurrence can be predicted using random forest model, a machine learning algorithm. Data were collected from patients treated for ameloblastoma between 1999 and 2019 at the University of Hong Kong. Fourteen clinical parameters were used to grow the decision trees to classify patients with or without ameloblastoma recurrence in the follow-up period. The random forest algorithm was computed 100 times in the training cohort (n = 100) and verified in the testing cohort (n = 50). The receiver operating characteristic curve (ROC) and area under the curve (AUC) were used as the performance measurement of separability. One hundred and fifty patients (76 female, 74 male) were recruited, with a mean follow-up time of 103 months. Recurrence occurred in a total of 25 cases (16.7%) over the 20-year period. The AUC were calculated for the median and mean ROC curves; these were 0.777 and 0.825, respectively. The results showed that random forest model was able to predict recurrence of ameloblastoma with reliable accuracy. The four most important variables influencing ameloblastoma recurrence were the time elapsed from treatment, initial surgical treatment, tumour size, and radiographic presentation. This study provides insights into the detection of high-risk patient groups to monitor recurrence. Further application of random forest to other diseases could greatly benefit clinical decisions.

Automated summary

The study aimed to investigate the use of a machine learning algorithm, random forest model, to predict the recurrence of ameloblastoma. By utilizing 14 clinical parameters, decision trees were developed to classify patients. The results demonstrated that the random forest model could effectively predict the recurrence of ameloblastoma and identified the key factors influencing recurrence.