RESPONSE TO LETTER TO THE EDITOR: “ARTIFICIAL INTELLIGENCE AND DECISION-MAKING FOR VESTIBULAR SCHWANNOMA SURGERY”

In Reply: We would like to thank the authors for their interest in our article “Predictors of Postoperative Complications in Vestibular Schwannoma Surgery – A Population Based Study” (1). The group noted our analysis of population data in identifying predictors of postoperative outcomes and suggested that an artificial neural network (ANN) could be used to overcome the limitations of our study. This group has recently published an excellent proof-of-concept study comparing a multivariable logistic regression (LR) model to an ANN for predicting recurrence of vestibular schwannomas (VS). Their ANN was impressively able to predict recurrence with higher sensitivity and specificity compared with a standard regression model (2). We appreciate the group's insight into alternative analyses that could be performed on the population datasets used in our study. Recently, several groups have demonstrated the ability of ANNs to predict postoperative outcomes following various surgical interventions (3–6). These studies show that ANNs can offer advantages over LRs in predicting specific outcomes depending on the dataset on which they are trained (4,6). Indeed, these flexible non-linear systems perform well on noisy input patterns, can detect complex implicit interactions between predictor variables, have a high fault tolerance, and easily generalize from input data (5). Despite the advantages of ANN, there are drawbacks that limit the feasibility of ANN in certain circumstances. The use of ANNs is more computationally burdensome compared with LR and they are more prone to overfitting (7). Moreover, due to the “black box” nature of ANNs, it can be difficult to identify important predictors, whereas with LR this task is much more expedient (7). For these reasons, logistic regression models continue to be employed for identifying predictors of VS surgery postoperative outcomes (8,9). We recognize the potential value in deploying an ANN to predict postoperative outcomes in patients who have undergone VS microsurgery using the datasets from our study. We agree with the authors that the complexity of patient risk-factors and the multifactorial causes of postoperative outcomes supports further analyses by machine learning techniques. We applaud the author's recent success in developing an ANN, and are eagerly awaiting the group's future work. We invite the authors to reach out to our group in the future for potential collaborations, including sharing of datasets and analytic expertise.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.