Applications of artificial intelligence in neurosurgical practice and residency education: a systematic review
https://doi.org/10.13181/mji.oa.268320
- Author:
Dewa Putu Wisnu Wardhana
1
;
Irawan Yusuf
2
;
Irwin Aras
3
Author Information
1. Department of Surgery, Faculty of Medicine, Universitas Udayana, Universitas Udayana Hospital, Bali, Indonesia
2. Department of Physiology, Universitas Hasanuddin, Makassar, Indonesia; Department of Medical Education, Universitas Hasanuddin, Makassar, Indonesia
3. Department of Medical Education, Universitas Hasanuddin, Makassar, Indonesia
- Publication Type:Journal Article
- Keywords:
artificial intelligence;
education;
internship and residency;
medicine;
neurosurgery;
simulation training
- MeSH:
artificial intelligence;
education;
internship and residency;
medicine;
neurosurgery;
simulation training
- From:
Medical Journal of Indonesia
2026;35(3):186-194
- CountryIndonesia
- Language:English
-
Abstract:
BACKGROUND :Artificial intelligence (AI) offers innovative and efficient training methods for neurosurgery residents, yet existing evidence remains fragmented, often limited to specific technological applications and failing to fully assess their impact on clinical competence. Therefore, this study aimed to map the current evidence on AI applications in neurosurgery and curriculum development.
METHODS :This study was conducted following the preferred reporting items for systematic reviews and meta-analyses guidelines by searching for observational or experimental studies on AI in neurosurgery residency education (2020–2025) in PubMed, ScienceDirect, and Google Scholar. Keywords included variations of “training,” “artificial intelligence,” and “neurosurgery”. After screening 719 initial records and assessing their eligibility, 10 studies were selected for narrative synthesis.
RESULTS :A review of 10 studies reveals 2 key roles for AI in neurosurgical training: improving psychomotor skills through machine learning-integrated simulators and reinforcing theoretical knowledge through large language models (LLMs). Virtual reality (VR) simulators such as NeuroVR and virtual operative assistant provide objective metrics and personalized feedback, with assessment accuracy of up to 80%. LLMs such as ChatGPT show 50.4–84.8% accuracy in answering exam questions. However, simulators cannot yet capture all surgical competencies, and LLMs are still weak in processing image-based data.
CONCLUSIONS :AI can improve neurosurgical training through personalized, objective learning tools, but AI simulators and LLMs remain supplements and cannot replace human expertise in clinical reasoning and complex image interpretation.
- Full text:202610301506561788435-3-8320-Dewa-Putu-Wisnu-Wardhana_xmp.pdf
