Abstract
Magnetic resonance imaging (MRI) has been proposed as a complimentary method to measure bone quality and assess fracture risk. However, manual segmentation of MR images of bone is time-consuming, limiting the use of MRI measurements in the clinical practice. The purpose of this paper is to present an automatic proximal femur segmentation method that is based on deep convolutional neural networks (CNNs). This study had institutional review board approval and written informed consent was obtained from all subjects. A dataset of volumetric structural MR images of the proximal femur from 86 subjects were manually-segmented by an expert. We performed experiments by training two different CNN architectures with multiple number of initial feature maps, layers and dilation rates, and tested their segmentation performance against the gold standard of manual segmentations using four-fold cross-validation. Automatic segmentation of the proximal femur using CNNs achieved a high dice similarity score of 0.95 ± 0.02 with precision = 0.95 ± 0.02, and recall = 0.95 ± 0.03. The high segmentation accuracy provided by CNNs has the potential to help bring the use of structural MRI measurements of bone quality into clinical practice for management of osteoporosis.
Highlights
Magnetic resonance imaging (MRI) has been successfully performed in vivo for structural imaging of trabecular bone architecture within the proximal femur[16,17,18]
Two supervised deep convolutional neural networks (CNNs) architectures based on 2D convolution (2D CNN) and 3D convolution (3D CNN) were used and evaluated for automatic proximal femur segmentation on MR images
The 3D CNN-dilated with 32 initial feature maps and 4 layers each in the contracting/expanding paths and concatenation of feature maps obtained with dilation rates r = 1, 2, 4, 8 outperformed the other CNNs with area under the Receiver operating characteristics (ROC) curve (AUC) = 0.999 ± 0.0 and area under the PRC (AP) = 0.990 ± 0.002
Summary
MRI has been successfully performed in vivo for structural imaging of trabecular bone architecture within the proximal femur[16,17,18]. Bone quality metrics derived from FE analysis of MR images are shown to correlate with high resolution qCT imaging, and may reveal different information about bone quality than that provided by DXA18 These technical developments overlay the significance of image analysis tools to determine osteoporosis related hip fracture risk. In a study using structural MRIs, Hallyburton et al used pyramidal CNN architectures for segmenting the proximal femur to achieve moderate segmentation results with dice similarity coefficient (DSC) ~0.7035. These approaches are limited by the size of the receptive field of the networks and by the time required for CNN training and inference, especially for volumetric datasets
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
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
