Intracranial Hemorrhage Classification using CBAM Attention Module and Convolutional Neural Networks

Intra-Cerebral Hemorrhage (ICH) is a life-threatening condition characterized by internal bleeding in the brain. The conventional methods for detecting and diagnosing ICH are often time-consuming and complex. However, recent advancements in deep learning techniques have revealed encouraging outcomes by enhancing the performance and efficiency of ICH detection and classification. In this study, we present the CBAM-NeXt model, a fusion of the Convolutional Block Attention Module (CBAM) and the ResNeXt-101 architecture. The CBAM module incorporates both channel and spatial attention mechanisms, making the model selectively emphasize more informative parts of input images. Furthermore, we employ the Extreme Gradient Boosting (XGBoost) algorithm to categorize the extracted features. Our proposed method was trained and evaluated on the RSNA Intracranial Hemorrhage CT challenge (2019) dataset. The results illustrate that our proposed method obtained the average F1 score and average accuracy of 70% and 97.9%, respectively. Our results outperform contemporary baselines in the RSNA challenge and benchmark.