Determination of the Rupture Risk of Cerebral Aneurysms via the Application of a Narrow Neural Network Classifier

Abstract

Determining the risk of rupture inAcases of unruptured cerebral aneurysms is critical for the treatment process. Certain morphological and hemodynamic parameters have been investigated to discover their effects on the risk of rupture. The present study aimed to determine the rupture risk rate of unruptured cerebral aneurysms by examining age, sex, location, width, length, type, and comorbidity parameters. For this purpose, the data were categorized using different classifiers with supervised machine learning for 220 patients diagnosed with cerebral aneurysms between the years 2011 and 2022. Of the patients, 127 had unruptured aneurysms, while 93 had experienced ruptured aneurysms. The training was conducted by applying the Narrow Neural Network classifier algorithm. This algorithm was preferred for its capability to deliver satisfactory performance in learning and classification tasks using limited datasets. Based on the results of the analysis, the rupture risk for the training network was classified with 75% accuracy, and the validation accuracy in the testing process was calculated with 81.8% success. The close values of the validation rates in the training and test scenarios indicate the high level of success at which this study was conducted. Although no ruptures were noted in the radiology reports for three of the patients, the machine learning classification algorithm predicted that ruptures would occur in these patients over time. Upon examination of the disease history of these patients, which included the results recorded in radiology reports completed at later dates, the aneurysms of all three were observed to have ruptured. This model revealed with great accuracy that for ruptured cerebral aneurysms, middle cerebral artery location and a size of 2-10 mm constitute the primary morphological factors, while heart and hypertension comorbidities represent the major hemodynamic parameters. These findings highlight the potential of the NNN algorithm model to evaluate aneurysms of all sizes with high accuracy, without the need for exclusion criteria. This is the first known study to incorporate such a diverse range of comorbidity parameters, along with detailed information on the aneurysm's laterality (right-left) and segment location, in determining rupture status.