Deep convolutional neural network for segmentation of knee joint anatomy

Self Cite

Purpose To describe and evaluate a segmentation method using joint adversarial and segmentation convolutional neural network to achieve accurate segmentation using unannotated MR image datasets. Theory and Methods A segmentation pipeline was built using joint adversarial and segmentation network. A convolutional neural network technique called cycle‐consistent generative adversarial network (CycleGAN) was applied as the core of the method to perform unpaired image‐to‐image translation between different MR image datasets. A joint segmentation network was incorporated into the adversarial network to obtain additional functionality for semantic segmentation. The fully automated segmentation method termed as SUSAN was tested for segmenting bone and cartilage on 2 clinical knee MR image datasets using images and annotated segmentation masks from an online publicly available knee MR image dataset. The segmentation results were compared using quantitative segmentation metrics with the results from a supervised U‐Net segmentation method and 2 registration methods. The Wilcoxon signed‐rank test was used to evaluate the value difference of quantitative metrics between different methods. Results The proposed method SUSAN provided high segmentation accuracy with results comparable to the supervised U‐Net segmentation method (most quantitative metrics having P > 0.05) and significantly better than a multiatlas registration method (all quantitative metrics having P < 0.001) and a direct registration method (all quantitative metrics having P< 0.0001) for the clinical knee image datasets. SUSAN also demonstrated the applicability for segmenting knee MR images with different tissue contrasts. Conclusion SUSAN performed rapid and accurate tissue segmentation for multiple MR image datasets without the need for sequence specific segmentation annotation. The joint adversarial and segmentation network and training strategy have promising potential applications in medical image segmentation.

mentioning

confidence: 99%

mentioning

confidence: 99%

mentioning

confidence: 99%

See 1 more Smart Citation

SUSAN: segment unannotated image structure using adversarial network

Liu

2018

Self Cite

“…Deep learning was used to provide automatic detection of artery regions in the reconstructed images, from which an artery contrast enhancement curve can be obtained to guide the selection of desired contrast phases for clinical use. Specifically, a convolutional encoder‐decoder (CED) network (Supporting Information Figure ), previously shown for accurate and efficient multi‐tissue segmentation, was trained to segment the femoral arteries in the prostate images. The images and segmentation masks used for training were obtained on 7 additional prostate data sets (different from the 22 data used for image reconstruction, with IRB approval for waived written informed consent) acquired with the same protocol.…”

Section: Methodsmentioning

confidence: 99%

GRASP‐Pro: imProving GRASP DCE‐MRI through self‐calibrating subspace‐modeling and contrast phase automation

Feng

Huang

Tong

et al. 2019

Self Cite

Purpose To propose a highly accelerated, high‐resolution dynamic contrast‐enhanced MRI (DCE‐MRI) technique called GRASP‐Pro (golden‐angle radial sparse parallel imaging with imProved performance) through a joint sparsity and self‐calibrating subspace constraint with automated selection of contrast phases. Methods GRASP‐Pro reconstruction enforces a combination of an explicit low‐rank subspace‐constraint and a temporal sparsity constraint. The temporal basis used to construct the subspace is learned from an intermediate reconstruction step using the low‐resolution portion of radial k‐space, which eliminates the need for generating the basis using auxiliary data or a physical signal model. A convolutional neural network was trained to generate the contrast enhancement curve in the artery, from which clinically relevant contrast phases are automatically selected for evaluation. The performance of GRASP‐Pro was demonstrated for high spatiotemporal resolution DCE‐MRI of the prostate and was compared against standard GRASP in terms of overall image quality, image sharpness, and residual streaks and/or noise level. Results Compared to GRASP, GRASP‐Pro reconstructed dynamic images with enhanced sharpness, less residual streaks and/or noise, and finer delineation of the prostate without prolonging reconstruction time. The image quality improvement reached statistical significance (P < 0.05) in all the assessment categories. The neural network successfully generated contrast enhancement curves in the artery, and corresponding peak enhancement indexes correlated well with that from the manual selection. Conclusion GRASP‐Pro is a promising method for rapid and continuous DCE‐MRI. It enables superior reconstruction performance over standard GRASP and allows reliable generation of artery enhancement curve to guide the selection of desired contrast phases for improving the efficiency of GRASP MRI workflow.

“…Recently, a few convolutional neural network (CNN) based deep learning models, such as SegNet 10 and UNet, 11 have been applied for automatic cartilage and other tissue segmentation in knee MR images. [12][13][14][15][16] In these deep learning based semantic segmentation models, the UNet 12,14 has shown the state-of-the-art performance for knee cartilage semantic segmentation tasks as it maps the feature from earlier encoder layers to later decoder layers to improve segmentation performance. In Norman et al, 12 2D UNet with long skip connection has been used for semantic segmentation of knee cartilages from 2D slices.…”

mentioning

confidence: 99%

Automated cartilage and meniscus segmentation of knee MRI with conditional generative adversarial networks

Gaj

Yang

Nakamura

et al. 2019

Purpose Fully automatic tissue segmentation is an essential step to translate quantitative MRI techniques to clinical setting. The goal of this study was to develop a novel approach based on the generative adversarial networks for fully automatic segmentation of knee cartilage and meniscus. Theory and Methods Defining proper loss function for semantic segmentation to enforce the learning of multiscale spatial constraints in an end‐to‐end training process is an open problem. In this work, we have used the conditional generative adversarial networks to improve segmentation performance of convolutional neural network, such as UNet alone by overcoming the problems caused by pixel‐wise mapping based objective functions, and to capture cartilage features during the training of the network. Furthermore, the Dice coefficient and cross entropy losses were incorporated to the loss functions to improve the model performance. The model was trained and tested on 176, 3D DESS (double‐echo steady‐state) knee images from the Osteoarthritis Initiative data set. Results The proposed model provided excellent segmentation performance for cartilages with Dice coefficients ranging from 0.84 in patellar cartilage to 0.91 in lateral tibial cartilage, with an average Dice coefficient of 0.88. For meniscus segmentation, the model achieves 0.89 Dice coefficient for lateral meniscus and 0.87 Dice coefficient for medial meniscus. The results are superior to previously published automatic cartilage and meniscus segmentation methods based on deep learning models such as convolutional neural network. Conclusion The proposed UNet‐conditional generative adversarial networks based model demonstrated a fully automated segmentation method with high accuracy for knee cartilage and meniscus.