Machine Learning-Based Head Computerized Tomography Imaging in Diagnosis and Surgery Treatment of Hypertension Cerebral Hemorrhage

Abstract

The aim of this study was to explore the adoption of computerized tomography (CT) based on machine learning in the diagnosis and treatment of cerebral hemorrhage. A total of 150 patients with hypertensive intracerebral hemorrhage who underwent surgical treatment in hospital in the recent three years were selected. All patients underwent brain CT examination, and image data were collected. FCM, DRLSRE, FCRLS, and other algorithms were utilized to segment the patient's bleeding region, and the segmentation performance of each algorithm was compared. Based on the FCRLS algorithm segmentation, the bleeding area was calculated by using the function in the MATLAB toolbox. Based on the previous research, a computer-aided diagnosis (CAD) system was designed and verified. The results showed that FCRLS had the best segmentation effect and the fastest segmentation speed for discontinuous and irregular bleeding areas. The CAD system can obtain the information of bleeding site, bleeding amount, midline displacement, and ventricle compression comprehensively and accurately. The accuracy of this system was 94% and 86% for hematoma of patients with rapid bleeding and those with slow bleeding, respectively. The overall accuracy was 91.3%. To sum up, the CAD system based on machine learning and head CT image had good diagnostic effect for patients with hypertensive intracerebral hemorrhage and had high clinical adoption value.