Artificial intelligence and machine learning in neurosurgery: A review of diagnostic significance and treatment planning efficiency

Abstract

This review analyzes the significance of artificial intelligence (AI) and deep learning (DL) approaches used in radiology in neurosurgery patients and compares AI applications with human models to determine the applicability of AI in disease diagnosis, decision-making, and outcome prediction. A systematic review was conducted from 1997 to 2020 from the PubMed (MEDLINE) database. The search strategy adhered to guidelines outlined by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. The keywords used for the literature search included “Deep learning,” “Neurosurgery,” “Artificial Intelligence,” “Brain,” “Magnetic resonance imaging-MRI Brain,” and “Machine learning.” The studies focusing on the significance of DL and comparing AI applications with radiologists or clinical experts to enhance diagnostic protocols were included, whereas non-English articles, animal studies, articles lacking full text, and publications such as commentaries, technical notes, abstracts, editorials, opinions, and letters were excluded. A total of 24 articles were included in the review. The P value was observed in 44 out of 63 outcome measures (70%), out of which in 26 out of 63 outturn measures, artificial application subset machine learning (ML) has a significant edge over clinical diagnosis (P < 0.05). The review highlights the potential impact of AI-driven advancements in clinical radiology on enhancing treatment plans for neurosurgery patients, emphasizing the benefits of early intervention, cost reduction, time-saving approaches, and judicious health-care resource utilization. The study’s limitations include potential constraints in identifying relevant literature due to the selected search scope and inclusion criteria, not including studies published outside the specified timeframe and database, and a small number of included studies. Consequently, there is a risk of overlooking innovative methodologies or ground-breaking studies contributing to a more comprehensive understanding of AI applications in neurosurgery. Furthermore, the exclusion of certain publication types, such as commentaries, and conference papers may limit the diversity of different perspectives. However, the study highlights the potential of ML in neurosurgery and the importance of addressing variability in study design, patient populations, and outcome measures in future research to enhance the applicability of AI-driven approaches in clinical practice. It is imperative to recognize and address these challenges to understand the opportunities and limitations inherent in the integration of AI in neurosurgical practice.