Transfer Learning for Image Segmentation by Combining Image Weighting and Kernel Learning.

Many medical image segmentation methods are based on the supervised classification of voxels. Such methods generally perform well when provided with a training set that is representative of the test images to the segment. However, problems may arise when training and test data follow different distributions, for example, due to differences in scanners, scanning protocols, or patient groups. Under such conditions, weighting training images according to distribution similarity have been shown to greatly improve performance. However, this assumes that a part of the training data is representative of the test data; it does not make unrepresentative data more similar. We, therefore, investigate kernel learning as a way to reduce differences between training and test data and explore the added value of kernel learning for image weighting. We also propose a new image weighting method that minimizes maximum mean discrepancy (MMD) between training and test data, which enables the joint optimization of image weights and kernel. Experiments on brain tissue, white matter lesion, and hippocampus segmentation show that both kernel learning and image weighting, when used separately, greatly improve performance on heterogeneous data. Here, MMD weighting obtains similar performance to previously proposed image weighting methods. Combining image weighting and kernel learning, optimized either individually or jointly, can give a small additional improvement in performance.

Aortic stiffness and brain integrity in elderly patients with cognitive and functional complaints.

PURPOSE: Cerebral white matter lesions (WML) and brain atrophy are frequent in older persons and are associated with adverse outcomes. It has been suggested that aortic stiffness plays a role in the pathogenesis of WML and gray matter (GM) loss. There is, however, little evidence on the association between aortic stiffness and brain integrity in older patients. In this study, we investigated whether aortic stiffness is associated with WML and GM volume in older patients with cognitive and functional complaints. PATIENTS AND METHODS: Fazekas score was used to analyze WML on brain imaging data of 84 persons; in a subanalysis on 42 MRI scans, the exact volume of white matter hyperintensities (WMH) and GM was determined using a brain-tissue and WMH tool. Aortic stiffness, measured as aortic pulse wave velocity (aPWV) and central pulse pressure (cPP), and blood pressure levels were non-invasively measured by the Mobil-O-Graph. RESULTS: Mean age was 76.6 (±6.8) years. Age was correlated with cPP (Spearman's ρ =0.296, P=0.008), aPWV (r 2=0.785, P<0.001) and WMH volume (r 2=0.297, P<0.001). cPP did not differ between categories of Fazekas, whereas aPWV increased with increasing Fazekas score (P for trend <0.001). After additional adjustment for age, levels of aPWV did not differ between categories. Both cPP and aPWV were associated with WMH volumes (lnB 0.025, P=0.055 and lnB 0.405, P<0.001, respectively); after additional adjustment for age, estimates were less consistent. Both cPP and aPWV were negatively associated with GM volumes in multivariate analysis (B=2.805, P=0.094 and B=111.052, P=0.032). CONCLUSION: Higher aortic stiffness was partly associated with increased volume of WMH and decreased volume of GM and slightly influenced by blood pressure. Age also plays a role in this association in older patients.

Transfer learning by feature-space transformation: A method for Hippocampus segmentation across scanners.

Many successful approaches in MR brain segmentation use supervised voxel classification, which requires manually labeled training images that are representative of the test images to segment. However, the performance of such methods often deteriorates if training and test images are acquired with different scanners or scanning parameters, since this leads to differences in feature representations between training and test data. In this paper we propose a feature-space transformation (FST) to overcome such differences in feature representations. The proposed FST is derived from unlabeled images of a subject that was scanned with both the source and the target scan protocol. After an affine registration, these images give a mapping between source and target voxels in the feature space. This mapping is then used to map all training samples to the feature representation of the test samples. We evaluated the benefit of the proposed FST on hippocampus segmentation. Experiments were performed on two datasets: one with relatively small differences between training and test images and one with large differences. In both cases, the FST significantly improved the performance compared to using only image normalization. Additionally, we showed that our FST can be used to improve the performance of a state-of-the-art patch-based-atlas-fusion technique in case of large differences between scanners.

Transfer learning improves supervised image segmentation across imaging protocols.

The variation between images obtained with different scanners or different imaging protocols presents a major challenge in automatic segmentation of biomedical images. This variation especially hampers the application of otherwise successful supervised-learning techniques which, in order to perform well, often require a large amount of labeled training data that is exactly representative of the target data. We therefore propose to use transfer learning for image segmentation. Transfer-learning techniques can cope with differences in distributions between training and target data, and therefore may improve performance over supervised learning for segmentation across scanners and scan protocols. We present four transfer classifiers that can train a classification scheme with only a small amount of representative training data, in addition to a larger amount of other training data with slightly different characteristics. The performance of the four transfer classifiers was compared to that of standard supervised classification on two magnetic resonance imaging brain-segmentation tasks with multi-site data: white matter, gray matter, and cerebrospinal fluid segmentation; and white-matter-/MS-lesion segmentation. The experiments showed that when there is only a small amount of representative training data available, transfer learning can greatly outperform common supervised-learning approaches, minimizing classification errors by up to 60%.

MRBrainS Challenge: Online Evaluation Framework for Brain Image Segmentation in 3T MRI Scans.

Many methods have been proposed for tissue segmentation in brain MRI scans. The multitude of methods proposed complicates the choice of one method above others. We have therefore established the MRBrainS online evaluation framework for evaluating (semi)automatic algorithms that segment gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) on 3T brain MRI scans of elderly subjects (65-80 y). Participants apply their algorithms to the provided data, after which their results are evaluated and ranked. Full manual segmentations of GM, WM, and CSF are available for all scans and used as the reference standard. Five datasets are provided for training and fifteen for testing. The evaluated methods are ranked based on their overall performance to segment GM, WM, and CSF and evaluated using three evaluation metrics (Dice, H95, and AVD) and the results are published on the MRBrainS13 website. We present the results of eleven segmentation algorithms that participated in the MRBrainS13 challenge workshop at MICCAI, where the framework was launched, and three commonly used freeware packages: FreeSurfer, FSL, and SPM. The MRBrainS evaluation framework provides an objective and direct comparison of all evaluated algorithms and can aid in selecting the best performing method for the segmentation goal at hand.

Multi-Center MRI Carotid Plaque Component Segmentation Using Feature Normalization and Transfer Learning.

Automated segmentation of plaque components in carotid artery magnetic resonance imaging (MRI) is important to enable large studies on plaque vulnerability, and for incorporating plaque composition as an imaging biomarker in clinical practice. Especially supervised classification techniques, which learn from labeled examples, have shown good performance. However, a disadvantage of supervised methods is their reduced performance on data different from the training data, for example on images acquired with different scanners. Reducing the amount of manual annotations required for each new dataset will facilitate widespread implementation of supervised methods. In this paper we segment carotid plaque components of clinical interest (fibrous tissue, lipid tissue, calcification and intraplaque hemorrhage) in a multi-center MRI study. We perform voxelwise tissue classification by traditional same-center training, and compare results with two approaches that use little or no annotated same-center data. These approaches additionally use an annotated set of different-center data. We evaluate 1) a nonlinear feature normalization approach, and 2) two transfer-learning algorithms that use same and different-center data with different weights. Results showed that the best results were obtained for a combination of feature normalization and transfer learning. While for the other approaches significant differences in voxelwise or mean volume errors were found compared with the reference same-center training, the proposed approach did not yield significant differences from that reference. We conclude that both extensive feature normalization and transfer learning can be valuable for the development of supervised methods that perform well on different types of datasets.