A New Deep-Learning Method for Classification Intracranial Aneurysm

Nowadays, the problem of detecting IA (intracranial aneurysms) is coming to the fore. IA is a bulge on the blood vessel, which is caused by the weakening of its wall. Such a bulge is very dangerous as it can burst and cause fatal consequences up to death. Therefore, the development of an automated system that would be able to detect IA could significantly help in early diagnosis and reduce mortality from this disease. For this reason, we aimed to classify IA from an image using DNN (deep neural networks). In our work, we propose a 2DCNN (2D Convolutional Neural Network) and compare the results with available ResNets (ResNet50, ResNet101, ResNet152). The network proposed by us obtained an overall accuracy of 98.21%. The ResNet networks on the test data achieved lower accuracy values, but the best results were achieved by the ResNet50 network with an overall accuracy of 98.55%.