Machine Learning in Neuro-Oncology: Can Data Analysis from 5,346 Patients Change Decision Making Paradigms?

BACKGROUND: Machine learning (ML) is an application of artificial intelligence (AI) giving computer systems the ability to learn data, without being explicitly programmed. ML is currently successfully used for optical character recognition, spam filtering, and face recognition. The aim of this study is to review its current application in the field of neuro-oncology.

METHODS: We conducted a systematic literature review on PubMed and Cochrane Database using a keyword search for the period January 30, 2000-March 31, 2018. Data were clustered for neuro-oncology scope of ML into three categories: patient outcome predictors, imaging analysis, and gene expression.

RESULTS: Data from 5,346 patients in 29 studies has been used to develop ML based algorithms (MLBA) in neuro-oncology. MLBA were used to predict outcome in 2,483 patients with a sensitivity range of 78-98% and specificity range of 76-95%. In all studies, MLBA had higher accuracy than conventional ones. MLBA for image analysis showed accuracy diagnosing low grade versus high grade gliomas (HGG) ranging from 80 to 93% and 90% diagnosing HGG versus lymphoma. Seven studies used MLBA to analyze gene expression in neuro-oncology.

CONCLUSIONS: MLBA in neuro-oncology have shown to predict patients' outcome more accurately than conventional parameters in retrospective analysis. If their high diagnostic accuracy in imaging analysis and detection of somatic mutations is corroborated in prospective studies, tissue diagnosis or liquid biopsy might curtail. Finally, MLBA are promising to help guide targeted therapy, lead to personalized medicine, and open areas of study in the cancer cellular signaling system, not otherwise known.