Deep learning methods allow fully automated segmentation of metacarpal bones to quantify volumetric bone mineral density

Abstract

Arthritis patients develop hand bone loss, which leads to destruction and functional impairment of the affected joints. High resolution peripheral quantitative computed tomography (HR-pQCT) allows the quantification of volumetric bone mineral density (vBMD) and bone microstructure in vivo with an isotropic voxel size of 82 micrometres. However, image-processing to obtain bone characteristics is a time-consuming process as it requires semi-automatic segmentation of the bone. In this work, a fully automatic vBMD measurement pipeline for the metacarpal (MC) bone using deep learning methods is introduced. Based on a dataset of HR-pQCT volumes with MC measurements for 541 patients with arthritis, a segmentation network is trained. The best network achieves an intersection over union as high as 0.94 and a Dice similarity coefficient of 0.97 while taking only 33 s to process a whole patient yielding a speedup between 2.5 and 4.0 for the whole workflow. Strong correlation between the vBMD measurements of the expert and of the automatic pipeline are achieved for the average bone density with 0.999 (Pearson) and 0.996 (Spearman’s rank) with $$p < 0.001$$ p < 0.001 for all correlations. A qualitative assessment of the network predictions and the manual annotations yields a 65.9% probability that the expert favors the network predictions. Further, the steps to integrate the pipeline into the clinical workflow are shown. In order to make these workflow improvements available to others, we openly share the code of this work.

DeepNAPSI multi-reader nail psoriasis prediction using deep learning

“…By adequately optimizing the internal parameters of the networks and with an appropriate amount of data, they can learn arbitrary mappings from the input data to the output, e.g. classify CT scans for the presence of a disease 7 . Lately, neural networks also were applied to challenges in the rheumatology, such as image-based classification of rheumatic diseases via data from magnetic resonance imaging or computed tomography 8 , 9 .…”

Deep Learning-Based Classification of Inflammatory Arthritis by Identification of Joint Shape Patterns—How Neural Networks Can Tell Us Where to “Deep Dive” Clinically

“…In our previous work, the second metacarpal bone head was segmented yielding a binary mask indicating the location of this region ( 11 ). This mask is a dense prediction of the bone region and does not preserve the internal microstructures of the bone.…”

“…These pre-processing steps yield two volumes of equal extent: The sub-region of the HR-pQCT scan, precisely, the second metacarpal bone head, and the corresponding bone mask indicating the region of the bone voxel-wise. For the classification task, in addition to the HR-pQCT sub-region inputs, we also investigated the performance of the model using only the bone mask extracted by the segmentation model described in reference ( 11 ). This way the model prediction is entirely based on the volumetric shape of the bone.…”

“…In arthritis patients, the technology is sensitive in detecting and quantifying surface changes that influence the shape of bone, such as erosions and bony spurs ( 6 ). Based on such HR-pQCT datasets from arthritis patients we have recently set up a convolutional supervised auto-encoder (CSAE) network that could reliably define the form of the articular bone without the need of human intervention ( 11 ).…”

DeepQCT: Predicting fragility fracture from high-resolution peripheral quantitative CT using deep learning

“…Using images directly as input, DL has showed supreme diagnostic performance in pathology 28,29 , dermatology 30,31 , and radiology 32,33 . Specifically, it was also able to diagnose rheumatoid and psoriatic arthritis from hand HR-pQCT 34,35 , making it a promising solution also for limb HR-pQCT.…”