Abstract
Intracranial hemorrhages (ICHs) pose a critical medical challenge with a high mortality rate, necessitating timely and accurate diagnosis. This study focuses on enhancing the performance of deep learning models for classifying brain hemorrhages in CT scan images. A grid search methodology was employed to tune hyperparameters, particularly addressing the imbalance between ICH and healthy slices. The study utilized ResNet50 for hyperparameter tuning, achieving significant improvements in sensitivity and overall performance through various techniques. The model demonstrated remarkable performance at both slice-level and patient-level scopes, outperforming previous literature.
Introducing the Hemorica Dataset
This research introduces the Hemorica Dataset, a curated collection of non-contrast CT scans for Intracranial Hemorrhage detection and classification. It comprises 82 patient scans, with 36 diagnosed with ICH, and includes detailed slice-level annotations reviewed by expert radiologists. The dataset is specifically designed to tackle challenges like class imbalance and is ideal for developing and benchmarking deep learning models in neuroradiology.