Title: A modified UNet-based semantic segmentation architecture for pancreas tumour detection

Authors: Banavathu Sridevi; B. John Jaidhan

Addresses: Department of CSE, GIT, GITAM (Deemed to be University), Visakhaptnam, Andhrapradesh, India ' Department of CSE, GIT, GITAM (Deemed to be University), Visakhaptnam, Andhrapradesh, India

Abstract: For computer aided diagnosis, computerised organ segmentation is a crucial but complicated task. The anatomy of the pancreas varies greatly and it is an abdominal organ. Especially when compared to other organs like the liver, heart, or kidneys, this prevents earlier segmentation approaches from obtaining high accuracy levels. To address this issue, we proposed a modification in UNet architecture called DAH-UNet that combines residual blocks, a rebuilt atrous spatial pyramid pooling (ASPP), and depth-wise convolutions. Also, a hybrid loss function which is explicitly aware of the boundaries is another thing we suggest. Experiments were conducted on two publicly available dataset and proved better in some metrics as compare to existing semantic segmentation models.

Keywords: pancreas tumour detection; UNet architecture; atrous spatial pyramid pooling; ASPP; depth-wise convolutions; semantic segmentation.

DOI: 10.1504/IJBRA.2024.137372

International Journal of Bioinformatics Research and Applications, 2024 Vol.20 No.1, pp.1 - 20

Received: 22 Jun 2023
Accepted: 19 Jul 2023
Published online: 14 Mar 2024 *