Abstract

Segmentation of brain images from Magnetic Resonance Images (MRI) is an indispensable step in clinical practice. Morphological changes of sub-cortical brain structures and quantification of brain lesions are considered biomarkers of neurological and neurodegenerative disorders and used for diagnosis, treatment planning, and monitoring disease progression. In recent years, deep learning methods showed an outstanding performance in medical image segmentation. However, these methods suffer from generalisability problem due to inter-centre and inter-scanner variabilities of the MRI images. The main objective of the study is to develop an automated deep learning segmentation approach that is accurate and robust to the variabilities in scanner and acquisition protocols. In this paper, we propose a transductive transfer learning approach for domain adaptation to reduce the domain-shift effect in brain MRI segmentation. The transductive scenario assumes that there are sets of images from two different domains: (1) source—images with manually annotated labels; and (2) target—images without expert annotations. Then, the network is jointly optimised integrating both source and target images into the transductive training process to segment the regions of interest and to minimise the domain-shift effect. We proposed to use a histogram loss in the feature level to carry out the latter optimisation problem. In order to demonstrate the benefit of the proposed approach, the method has been tested in two different brain MRI image segmentation problems using multi-centre and multi-scanner databases for: (1) sub-cortical brain structure segmentation; and (2) white matter hyperintensities segmentation. The experiments showed that the segmentation performance of a pre-trained model could be significantly improved by up to 10%. For the first segmentation problem it was possible to achieve a maximum improvement from 0.680 to 0.799 in average Dice Similarity Coefficient (DSC) metric and for the second problem the average DSC improved from 0.504 to 0.602. Moreover, the improvements after domain adaptation were on par or showed better performance compared to the commonly used traditional unsupervised segmentation methods (FIRST and LST), also achieving faster execution time. Taking this into account, this work presents one more step toward the practical implementation of deep learning algorithms into the clinical routine.

Highlights

  • Medical image segmentation is a pivotal task in diagnosis, treatment, and surgical planning, and monitoring disease progression over time

  • The first set, where Internet Brain Segmentation Repository1 (IBSR) was source and MICCAI 2012 was target, significant improvement in the overall result was observed after applying the domain adaptation, reaching a Dice Similarity Coefficient (DSC) of 0.799 compared to the baseline segmentation with the DSC score of 0.680 (p = 2.8 × 10−27)

  • Aside from the putamen structure, our method was effective in improving the performance of the Convolutional Neural Networks (CNNs) for all other structures and significantly improved overall average DSC from the baseline

Read more

Summary

Introduction

Medical image segmentation is a pivotal task in diagnosis, treatment, and surgical planning, and monitoring disease progression over time. Manually annotating MRI images is a timeconsuming and a laborious task, which has to be done by experts with knowledge in disease-specific aspects and anatomy. Deep learning methods—in particular, Convolutional Neural Networks (CNNs)—have shown a remarkable advance in the field of brain MRI segmentation for many different applications (Akkus et al, 2017; Bernal et al, 2019). The network has to be retrained using the images from this new domain, requiring expert annotated labels. This commonly faced issue is known as the domain-shift problem, which hinders the applicability of deep learning methods in practice. The data-driven nature, which demands a vast amount of expert annotated images, often makes fully retraining a CNN impossible

Objectives
Methods
Results
Discussion
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call