A global survey of machine learning applications in neurosurgery

Machine Learning (ML) has the potential for predictive analytics, integrating all the complex data of patients, or mining deep-seated features from imaging, histology, and genomics.
, more and more neurosurgeons are replacing traditional statistical models with complex ML models to improve predictive abilities, such as postoperative satisfaction and early complications.
E. Staartjes of Neurosurgery at the Centre for Clinical Neuroscience at the University Hospital Zurich, Switzerland, and others conducted a global survey on the use of ML by neurosurgeons to assess the factors associated with the use of ML in neurosurgery clinical and scientific research, published online in August 2020 in Acta Neurochir.
study authors sent online questionnaires containing 9-10 questions to 7,280 members of the European Society of Neurosurgery (EANS) and the Association of Neurosurgeons (CNS) in February and March 2019.
received 362 responses.
respondents were aged 30-40 (32.6 per cent), 89.2 per cent were men and the majority were in spinal cord surgery (36.2 per cent).
103 (28.5%) of these neurosurgeons used ML in clinical practice, and 31.1% used ML in scientific research.
The global distribution of ML usage is relatively uniform (p-0.125), mainly in Europe and North America, 30.9 per cent in Europe, 25.6 per cent in North America, 33.3 per cent in Latin America and the Middle East, 44.4 per cent in Asia-Pacific and 100 per cent in Africa (figure 1).
ML is most commonly used to predict prognosis (60.2%) and complications (51.5%), or to interpret or quantify medical imaging (50.5%).
, ML can play an advantageous role in patients' knowledge (38.8%), disease severity (37.9%) and diagnosis (19.4%).
1. 362 neurosurgeons who responded to the questionnaire using ML were distributed around the world.
the results of the multi-logistic regression analysis shows that the use of ML in the world has been more extensive and evenly distributed.
ML applied to nerve tumors (OR=2.76; 95% CI, 1.28-6.05; p=0.010), function (OR=2.79;95% CI, 1.03-7.47; p=0.040), external Studies such as injuries (OR=3.8; 95% CI, 1.44-10.02; p=0.007) and epilepsy (OR=3.8; 95% CI, 1.14-12.9; p=0.030).
neurosurgeons use ML clinically, most commonly in preoperative surgical decision-making or treatment planning (3.27±0.86), objective assessment of diagnosis/grading/risk (3.22±0.84), predictive complications (3.13±0.92) and improved treatment decision-making/patient information (3.07±0.9).
the difficulties faced by ML usage are the lack of resources to develop ML models, including personnel and equipment (3.11±0.98), time limits (2.85±0.96), lack of observation indicators (2.84±1), and lack of data for developing ML models (2.67±0.99).
The study analyzed an overview of ML's use in global neurosurgery, which can better interpret imaging data, predict prognosis and complications, improve patients' quality of life and surgical satisfaction, and improve diagnostic and surgical decision-making.
of the neurosurgeons interviewed in the study had extensive clinical experience in using ML.
follow-up study to clarify the role and value of ML in neurosurgery.
: The intellectual property rights of the content published by the Brain Medical Exchange's Outside Information, God's Information and Brain Medicine Consulting are owned by the Brain Medical Exchange and the organizers, original authors and other relevant rights persons.
, editing, copying, cutting, recording, etc. without permission.
be licensed for use, the source must also be indicated.
welcome to forward and share.