Intracranial Hemorrhage Classification From CT Scan Using Deep Learning and Bayesian Optimization

Intracranial hemorrhage is a medical condition characterized by bleeding within the skull or brain tissue. It has mainly five subtypes: epidural, subdural, subarachnoid, intraparenchymal, and intraventricular. To ensure a successful outcome for a patient, timely and accurate identification of intracranial hemorrhage is crucial. However, a shortage of radiologists, particularly in rural areas, can lead to a delay in diagnosis. In this work, we proposed an automatic way of identifying intracranial hemorrhage from a Computed Tomography (CT) scan. To classify intracranial hemorrhage accurately, we have optimized the Densely Connected Convolutional Network (DenseNet) using Bayesian Optimization (BO). We utilized Bayesian optimization (BO) to determine the optimal learning rate, optimizer, and the number of nodes in the dense layer for the DenseNet architecture. Our proposed model can analyze CT scans to detect the presence of hemorrhage and identify its subtype. The optimized DenseNet model showcased remarkable performance. By ensuring accurate and reliable diagnoses, our method will assist doctors in making better-informed decisions and providing better care for their patients.