Detection of Biliary Artesia using Sonographic Gallbladder Images with the help of Deep Learning approaches

Abstract

BA (Biliary Atresia) is the major cause of both chronic liver illness and the high collective sign of liver transplantation. Connected techniques continue to progress to support the diagnosis of BA and the use of ultrasonography to support projected results after treatment with the KP (Kasai Portoenterostomy). The triangle cord mark, gallbladder anomalies, hilar lymphadenopathy, and the presence of hepatic subcapsular flow are all symptoms that are consistent with BA. However, there are no definite ultrasonography findings for BA. Ultrasound reports, on the other hand, offer a low cost and a real-time evaluation of intra-abdominal tissues. Researchers believe it is difficult to diagnose BA using sonographic gallbladder images without the appropriate skills, especially in rural locations where often experienced sonographers are scarce. To support the diagnosis of BA based on sonographic gallbladder images, a DL (Deep Learning) framework is built. We have applied four types of DL models i.e.VGG16, InceptionV3, ResNet152, and MobileNet, out of which MobileNet performs better with an accuracy of 97.87%, specificity of 97.51%, and a sensitivity of 98.18%. The DL framework in this paper provides a clarification to enable radiologists in advancing the diagnosis of BA in different clinical machine scenarios, particularly in underdeveloped countries and rural areas with limited specialists.