A Literature Review on Application of Machine Learning Techniques in Pancreas Segmentation

Abstract

Pancreas segmentation provide radiologically and clinically significant information for diagnosing diseases related to pancreas. Moreover, it also helps to plan pancreatic surgery. Pancreas segmentation is a challenging task due to very high shape and volume variability of the pancreas among different patients. In this paper, we have reviewed machine learning and deep learning-based approaches applied by various researchers for pancreas segmentation. These algorithms have a mean Dice Similarity Coefficient (DSC) of 78.35% with maximum and minimum DSC as 89.7% and 56.46% respectively. The availability of better computation power and large datasets will further enhance the performance of pancreas segmentation and related tasks. The approaches presented in this paper can be easily modified to the multi organ segmentation methods by changing the loss function. The study revealed the challenge of quantizing network while maintaining performance accuracy.