Promises and Perils of Artificial Intelligence in Neurosurgery

Abstract

Artificial intelligence (AI)-facilitated clinical automation is expected to become increasingly prevalent in the near future. AI techniques may permit rapid and detailed analysis of the large quantities of clinical data generated in modern healthcare settings, at a level that is otherwise impossible by humans. Subsequently, AI may enhance clinical practice by pushing the limits of diagnostics, clinical decision making, and prognostication. Moreover, if combined with surgical robotics and other surgical adjuncts such as image guidance, AI may find its way into the operating room and permit more accurate interventions, with fewer errors. Despite the considerable hype surrounding the impending medical AI revolution, little has been written about potential downsides to increasing clinical automation. These may include both direct and indirect consequences. Directly, faulty, inadequately trained, or poorly understood algorithms may produce erroneous results, which may have wide-scale impact. Indirectly, increasing use of automation may exacerbate de-skilling of human physicians due to over-reliance, poor understanding, overconfidence, and lack of necessary vigilance of an automated clinical workflow. Many of these negative phenomena have already been witnessed in other industries that have already undergone, or are undergoing “automation revolutions,” namely commercial aviation and the automotive industry. This narrative review explores the potential benefits and consequences of the anticipated medical AI revolution from a neurosurgical perspective.

Artificial Intelligence in Brain Tumour Surgery—An Emerging Paradigm

“…In ML, only the input and output are provided, while the algorithm “learns” the relevant patterns and trends [ 3 , 5 ]. Alternatively, ML may be “unsupervised”, in which the program analyses data without pre-defined labels, resulting in the ML program identifying similarities between datasets and clustering the data to identify the trends and patterns [ 2 , 6 , 7 ]. A worked example is the field of radiomics, in which AI programs analyse unlabelled scan images to identify clusters and patterns associated with certain grades of glioma, or by clustering GBM patients who have particularly good outcomes and then identifying the common patterns between these patients [ 6 ].…”

“…Alternatively, ML may be “unsupervised”, in which the program analyses data without pre-defined labels, resulting in the ML program identifying similarities between datasets and clustering the data to identify the trends and patterns [ 2 , 6 , 7 ]. A worked example is the field of radiomics, in which AI programs analyse unlabelled scan images to identify clusters and patterns associated with certain grades of glioma, or by clustering GBM patients who have particularly good outcomes and then identifying the common patterns between these patients [ 6 ]. Finally, reinforcement ML is the process through which algorithms are honed based on reward and punishment, whereby actions that increase the likelihood of achieving an end goal are rewarded, and actions distancing the program from the desired goal are punished [ 6 ].…”

“…A worked example is the field of radiomics, in which AI programs analyse unlabelled scan images to identify clusters and patterns associated with certain grades of glioma, or by clustering GBM patients who have particularly good outcomes and then identifying the common patterns between these patients [ 6 ]. Finally, reinforcement ML is the process through which algorithms are honed based on reward and punishment, whereby actions that increase the likelihood of achieving an end goal are rewarded, and actions distancing the program from the desired goal are punished [ 6 ]. The machine subsequently learns the optimal strategy for a given task [ 8 ].…”

“…The machine subsequently learns the optimal strategy for a given task [ 8 ]. With all forms of ML, large volumes of data can be processed leading to the detection of patterns and subtleties indiscernible to clinicians [ 6 ].…”

DeepNavNet: Automated Landmark Localization for Neuronavigation

“…Thus, a novel system that provides more accurate and reliable AC-PC localization would improve functional neurosurgery. With the revival of artificial intelligence (AI), sparked by deep learning methods demonstrating humanlevel performance in computer vision applications, it is now possible to conceive an AI-powered DBS surgical targeting system that augments surgical planning by the neurosurgery team (Panesar et al, 2020).…”