Anthony A Gatti1,2, Monica R Maly3,4. 1. School of Rehabilitation Sciences, McMaster University, 1280 Main St. W., Hamilton, ON, L8S 4L8, Canada. anthony@neuralseg.com. 2. NeuralSeg Ltd., Hamilton, ON, Canada. anthony@neuralseg.com. 3. School of Rehabilitation Sciences, McMaster University, 1280 Main St. W., Hamilton, ON, L8S 4L8, Canada. 4. Department of Kinesiology, University of Waterloo, Waterloo, Canada.
Abstract
OBJECTIVES: Accurate and efficient knee cartilage and bone segmentation are necessary for basic science, clinical trial, and clinical applications. This work tested a multi-stage convolutional neural network framework for MRI image segmentation. MATERIALS AND METHODS: Stage 1 of the framework coarsely segments images outputting probabilities of each voxel belonging to the classes of interest: 4 cartilage tissues, 3 bones, 1 background. Stage 2 segments overlapping sub-volumes that include Stage 1 probability maps concatenated to raw image data. Using six fold cross-validation, this framework was tested on two datasets comprising 176 images [88 individuals in the Osteoarthritis Initiative (OAI)] and 60 images (15 healthy young men), respectively. RESULTS: On the OAI segmentation dataset, the framework produces cartilage segmentation accuracies (Dice similarity coefficient) of 0.907 (femoral), 0.876 (medial tibial), 0.913 (lateral tibial), and 0.840 (patellar). Healthy cartilage accuracies are excellent (femoral = 0.938, medial tibial = 0.911, lateral tibial = 0.930, patellar = 0.955). Average surface distances are less than in-plane resolution. Segmentations take 91 ± 11 s per knee. DISCUSSION: The framework learns to automatically segment knee cartilage tissues and bones from MR images acquired with two sequences, producing efficient, accurate quantifications at varying disease severities.
OBJECTIVES: Accurate and efficient knee cartilage and bone segmentation are necessary for basic science, clinical trial, and clinical applications. This work tested a multi-stage convolutional neural network framework for MRI image segmentation. MATERIALS AND METHODS: Stage 1 of the framework coarsely segments images outputting probabilities of each voxel belonging to the classes of interest: 4 cartilage tissues, 3 bones, 1 background. Stage 2 segments overlapping sub-volumes that include Stage 1 probability maps concatenated to raw image data. Using six fold cross-validation, this framework was tested on two datasets comprising 176 images [88 individuals in the Osteoarthritis Initiative (OAI)] and 60 images (15 healthy young men), respectively. RESULTS: On the OAI segmentation dataset, the framework produces cartilage segmentation accuracies (Dice similarity coefficient) of 0.907 (femoral), 0.876 (medial tibial), 0.913 (lateral tibial), and 0.840 (patellar). Healthy cartilage accuracies are excellent (femoral = 0.938, medial tibial = 0.911, lateral tibial = 0.930, patellar = 0.955). Average surface distances are less than in-plane resolution. Segmentations take 91 ± 11 s per knee. DISCUSSION: The framework learns to automatically segment knee cartilage tissues and bones from MR images acquired with two sequences, producing efficient, accurate quantifications at varying disease severities.
Entities:
Keywords:
Cartilage; Deep learning; Image processing; Magnetic resonance imaging; Osteoarthritis
Authors: Zhiliang Wei; Jiadi Xu; Peiying Liu; Lin Chen; Wenbo Li; Peter van Zijl; Hanzhang Lu Journal: Magn Reson Med Date: 2017-12-21 Impact factor: 4.668
Authors: Andrew J Metcalfe; Maria L E Andersson; Rhian Goodfellow; Carina A Thorstensson Journal: BMC Musculoskelet Disord Date: 2012-08-22 Impact factor: 2.362