Artificial intelligence technologies in predicting outcomes and adopting therapeutic tactics in patients with recurrent intracranial meningiomas

INTRODUCTION. Meningiomas are common primary intracranial and spinal tumors in adults. Aggressive meningiomas can recur after surgical removal and radiotherapy. There are no treatment standards for such tumors. Modern artificial intelligence (AI) technologies can help a neurosurgeon predict the behavior of a neoplastic process in the central nervous system. AIM. Development and evaluation of the effectiveness of a neural network algorithm predicting the further development of a neoplastic process in recurrent intracranial meningiomas. MATERIALS AND METHODS. To solve the problem, an Excel database was used with patient information obtained from the analysis of medical records. More than 160 multimodal features grouped into sections were used. According to the results of statistical analysis, correlated, uninformative, and features with uneven data distribution within classes were removed. Finally, two of the most appropriate classification models were chosen: decision tree and random forest algorithms. RESULTS. Four models were built and evaluated based on the results of the work. The basic random forest model showed the best classification accuracy (about 90 %). It also helped to assess the significance of the studied features. CONCLUSION. Due to the constant growth of multimodal data in neuro-oncology, it is difficult for doctors to analyze them using traditional approaches and predict the behavior of a neoplastic process. Therefore, neurosurgeons need to turn to modern artificial intelligence (AI) technologies for help.