Fully Automated Spleen Segmentation in Patients using Convolutional Neural Network on CT Images

A deep learning network was created for the purpose of segmenting the spleen on thorax-abdomen CT images for the purpose of this retrospective investigation. The purpose of this research was to create a model that is based on deep learning and is capable of fully automated spleen fragmentation utilizing computerized tomography images. Additionally, the effectiveness of this algorithm was going to be evaluated either directly or indirectly affect the spleen. In order to do this, UNet segmentation is used to train the dataset that had illnesses with and without splenic involvement, as well as a dataset obtained from the Medical Segmentation Decathlon that did not contain splenic abnormalities. The Dice Similarity Coefficient was one of the four developed metrics that were used in the evaluation of the two models’ respective segmentation abilities. A modified deep learning neural network is used in this work by blending convolutional neural network and XGBoost classifier (XGNN). The training database of a Deep learning model should contain factors that impact the spleen in order to achieve best precision, completely automated spleen fragmentation in clinical practice.