Synergic deep learning model–based automated detection and classification of brain intracranial hemorrhage images in wearable networks

Abstract

With an intention of improving healthcare performance, wearable technology products utilize several digital health sensors which are classically linked into sensor networks, including body-worn and ambient sensors. On the other hand, intracerebral hemorrhage (ICH) defines the injury of blood vessels in the brain regions, which is accountable for 10–15% of strokes. X-ray computed tomography (CT) scans are commonly employed to determine the position and size of the hemorrhages. Manual segmentation of the CT scans by planimetry using a radiologist is effective; however, it consumes more time. Therefore, this paper develops deep learning (DL)–based ICH diagnosis using GrabCut-based segmentation with synergic deep learning (SDL), named GC-SDL model. The proposed method make use of Gabor filtering for noise removal, thereby the image quality can be raised. In addition, GrabCut-based segmentation technique is applied to identify the diseased portions effectively in the image. To perform the feature extraction process, SDL model is utilized and finally, softmax (SM) layer is employed as a classifier. In order to investigate the performance of the GC-SDL model, an extensive set of experimentation takes place using a benchmark ICH dataset, and the results are examined under different evaluation metrics. The experimental outcome stated that the GC-SDL model has reached a higher sensitivity of 94.01%, specificity of 97.78%, precision of 95.79%, and accuracy of 95.73%.