Advancing 3D Medical Image Segmentation: Unleashing the Potential of Planarian Neural Networks in Artificial Intelligence

Author Note Correspondence concerning this article should be addressed to Ziyuan Huang, University of Massachusetts Chan Medical School, 368 Plantation Street, Worcester, MA 01605 . Advancing 3D Medical Image Segmentation: Unleashing the Potential of Planarian Neural Networks in Artificial Intelligence Abstract Our study presents PNN - UNet as a method fo r constructing deep neural networks that replicate the planarian neural network (PNN) structure in the context of 3D medical image data. Planarians typically have a cerebral structure comprising two neural cords, where the cerebrum acts as a coordinator, and the neural cords serve slightly different purposes within the organism's neurological system. Accordingly, PNN - UNet comprises a D eep - UNet and a W ide - UNet as the nerve cords, with a densely connected autoencoder performing the role of the brain. This dist inct architecture offers advantages over both monolithic (UNet) and modular networks (Ensemble - UNet). Our outcomes on a 3D MRI hippocampus dataset, with and without data augmentation, demonstrate that PNN - UNet outperforms the baseline UNet and several othe r UNet variants in image segmentation. Introduction Medical image segmentation using deep learning techniques plays an increasingly crucial role in assisting clinical diagnosis. Every day, hospitals capture exponentially more medical images, making it increasingly difficult to process big data efficiently and effectively. Medical imaging segmentation can be classified into three major categories: 2D, 2.5D, and 3D (Minaee et al., 2021; Zhang et al., 2022) . The 2D method is to segment 3D images slice - by - slice, utilizing 2D slices as training and testing data. For the 2.5D category, segmentation algorithms usually segment 3D images slice - by - slice, adding neighboring slices as additional inputs. Lastly, 3D im ages are cropped and segmented into small cubic images for training and testing. It is important to note that different methods have their advantages and disadvantages in 3D medical image segmentation.