Abstract
In recent years, significant progress has been made in developing moreaccurate and efficient machine learning algorithms for segmentation of medicaland natural images. In this review article, we highlight the imperative role ofmachine learning algorithms in enabling efficient and accurate segmentation inthe field of medical imaging. We specifically focus on several key studiespertaining to the application of machine learning methods to biomedical imagesegmentation. We review classical machine learning algorithms such as Markovrandom fields, k-means clustering, random forest, etc. Although such classicallearning models are often less accurate compared to the deep learningtechniques, they are often more sample efficient and have a less complexstructure. We also review different deep learning architectures, such as theartificial neural networks (ANNs), the convolutional neural networks (CNNs),and the recurrent neural networks (RNNs), and present the segmentation resultsattained by those learning models that were published in the past three years.We highlight the successes and limitations of each machine learning paradigm.In addition, we discuss several challenges related to the training of differentmachine learning models, and we present some heuristics to address thosechallenges.