Systematic review of emerging artificial intelligence research in glioma diagnosis and therapy

Brain tumor management is one of the most challenging domains in neurosurgery, where timely diagnosis and intervention are critical, yet differentiating between diagnoses often presents difficulties. Despite advances, treatment options, especially for malignant tumors, have largely remained unchanged. Artificial Intelligence (AI) is increasingly being explored in neurosurgery, with its automation, data analysis capabilities, pattern recognition, and ability to detect previously imperceptible features offering potential for enhanced clinical management. This study provides significant insights for biomedical research by identifying promising algorithms for standardization in model healthcare settings and underscoring the interdisciplinary collaboration needed for their successful integration. A comprehensive search in PubMed, Scopus, and Web of Science yielded 4,362 results, of which 58 studies met the inclusion criteria for qualitative analysis. Studies were categorized by five clinical management domains: (1) Tumor Diagnosis, Imaging, Grading, (2) Surgical Planning and Intra-Operative Decision Making (3) Decision-Making AI in Prognostication and Follow-Up, and (4) AI in Histopathological Classification. Data extraction included study characteristics, metrics of model’s efficacy, and study limitations. The most studied domain was Prognostication and Follow-Up, with 20 studies. Most algorithms utilized institutional or publicly available datasets which were retrospectively evaluated. Various models yielded promising results in achieving their designated task. However, standardized validation methods and large, balanced databases are needed to ensure reliability and generalizability. Literature limitations and ethical considerations should also be addressed.