Evaluation of the diagnostic value of image processing algorithms and artificial intelligence in diagnosing and locating venous sinus thrombosis compared to 3D, MRV, T2WI sequences

Abstract

Background: Dural sinus thrombosis (DST) is a severe condition that should be diagnosed with magnetic resonance imaging (MRI) and treated as soon as possible. We aimed to assess the diagnostic value of image processing algorithms and artificial intelligence in diagnosing and locating venous sinus thrombosis. Methods: This is an analytic-descriptive, retrospective cross-sectional study performed from 2020-2022 on medical records of patients suspected of cerebral vein thrombosis who underwent 3DMRV and T2WI sequences imaging. T2WI coronal and sagittal images of patients were obtained from Kashani Hospital PACS system. In smartening the diagnosis of cerebral venous thrombosis, the areas marked in sagittal, coronal T2 images, and the patient MRV 3D image were used for the presence of venous sinus thrombosis to teach the intelligent algorithm. In the first stage, images of 100 standard sagittal and coronal MRI samples from the brain's venous sinuses were entered into the PYTHON software for software training. In the thrombosis evaluation stage (test), the patient's sagittal and coronal T2 images were given to the algorithm, and the algorithm intelligently segments the areas suspected of thrombosis. Then we evaluated the patient's MRV 3D image to confirm the result. Using this training, the program should identify the thrombosis area in the patient's MRV 3D image with appropriate accuracy. Results: The accuracy of artificial intelligence in diagnosing superior sagittal thrombosis was 99.1, and its accuracy for the right sigmoid sinus, left sigmoid sinus, and right and left transverse sinus were 75.8, 73.6, 71.5, and 70.4 respectively. Conclusion: We showed that artificial intelligence has high accuracy rates in image processing algorithms and artificial intelligence in diagnosing different areas of venous sinus thrombosis. This data could have critical importance in clinical practice.