Automatic knee cartilage and menisci segmentation from 3D-DESS MRI using deep semi-supervised learning

Panfilov, Egor; Tiulpin, Aleksei; Juntunen, Mikael; Casula, Victor; Nieminen, Miika T.; Saarakkala, Simo

doi:10.1016/j.joca.2019.02.391

Cited by 3 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Atlas-based segmentation approaches are one of the most robust image-based segmentation techniques in the field of medical simulations, which perform classification and segmentation in one go, 42 although deep learning-based techniques also look promising. 35 In terms of generation of the model geometry, we compared earlier the results of the FE models obtained from the atlas-based approach and manually segmented knee joint models with experimental data. 32 In that study, we showed that the models based on the atlas-based approach were able to predict the progression of OA similar to the models based on manual segmentation.…”

Section: Discussionmentioning

confidence: 99%

Rapid X-Ray-Based 3-D Finite Element Modeling of Medial Knee Joint Cartilage Biomechanics During Walking

et al. 2022

View full text Add to dashboard Cite

Finite element (FE) modeling is becoming an increasingly popular method for analyzing knee joint mechanics and biomechanical mechanisms leading to osteoarthritis (OA). The most common and widely available imaging method for knee OA diagnostics is planar X-ray imaging, while more sophisticated imaging methods, e.g., magnetic resonance imaging (MRI) and computed tomography (CT), are seldom used. Hence, the capability to produce accurate biomechanical knee joint models directly from X-ray imaging would bring FE modeling closer to clinical use. Here, we extend our atlas-based framework by generating FE knee models from X-ray images (N = 28). Based on measured anatomical landmarks from X-ray and MRI, knee joint templates were selected from the atlas library. The cartilage stresses and strains of the X-ray-based model were then compared with the MRI-based model during the stance phase of the gait. The biomechanical responses were statistically not different between MRI- vs. X-ray-based models when the template obtained from X-ray imaging was the same as the MRI template. However, if this was not the case, the peak values of biomechanical responses were statistically different between X-ray and MRI models. The developed X-ray-based framework may pave the way for a clinically feasible approach for knee joint FE modeling.

show abstract

Section: Discussionmentioning

confidence: 99%

Rapid X-Ray-Based 3-D Finite Element Modeling of Medial Knee Joint Cartilage Biomechanics During Walking

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Panfilov et al. [94] used the GAN model based on U‐Net for the cartilage and meniscus segmentation. The U‐net architecture was modified to work as a generator which performed slice‐wise segmentation of 3D‐DESS MRIs.…”

Section: Knee Image Segmentation Approachesmentioning

confidence: 99%

Review of automated segmentation approaches for knee images

Ridhma

Kaur

Sofat

et al. 2020

IET Image Processing

View full text Add to dashboard Cite

Knee disorders are common among the human population. Knee osteoarthritis (OA) is the most widespread knee joint disorder, which may require surgical treatment. The detection and diagnosis of knee joint disorders from medical images demand enormous human effort and time. The development of a computer‐aided diagnosis (CAD) system can notably minimise the burden of medical experts and remove the intra‐observer and inter‐observer variations. To achieve the goal, the highly challenging research problem of knee image segmentation has been frequently paid attention in past years, which can be efficiently applied in the development of the CAD system. Knee image segmentation is a challenging task owing to the image contrasts, intensity variations, shape irregularities, and the presence of thin cartilage structures. Therefore, this paper presents a literature review of automated segmentation approaches mainly focused on the segmentation of knee cartilage and bone, with respect to the underlying technical aspects, datasets used, and the performance reported. The paper also presents the growth from classical segmentation approaches towards the deep learning approaches in the knee image segmentation. Owing to the varying quality and complexity of different knee image datasets, this paper abstains from doing a rigorous comparative evaluation of image segmentation approaches.

show abstract

Нейросетевые решения на основе архитектуры U-Net для автоматической сегментации хрящевой ткани лучезапястного сустава на МР изображениях

Бруй,

Владимиров

2021

Моделирование

View full text Add to dashboard Cite

Сегментация хрящевой ткани на трехмерных магнитно-резонансных (МР) изображениях используется в задачах определения стадии дегенеративных и воспалительных заболеваний суставов. Для лучезапястного сустава ручная сегментация является крайне трудоемкой задачей в связи с его сложным устройством. Это обусловливает актуальность разработки полностью автоматических методов сегментации. Единственный предложенный ранее автоматический метод основан на глубоком обучении. Он обеспечивает неоднородную точность сегментации в зависимости от положения среза внутри трехмерного изображения. Целью данной работы является повышение точности автоматической сегментации хрящевой ткани на боковых срезах МР изображений лучезапястного сустава при помощи применения глубоких сверточных нейронных сетей (СНС). В работе рассматривались две архитектуры СНС: классическая U-Net и усеченная версия U-Net, в которой был удален последний блок сверток в пути уменьшения размерности. Точность сегментации была оценена при помощи трехмерного и двумерного коэффициентов Соренсена-Дайса (DSC), а также при помощи расчета площади под кривой точность-полнота (AUC-PR). Результаты сравнивались с опубликованными ранее данными для автоматического метода сегментации хрящевой ткани лучезапястного сустава при помощи СНС, обученной на основе патчей, а также с опубликованными результатами для процедуры ручной сегментации. Использование архитектур на основе U-Net позволило значительно повысить точность автоматической сегментации. Усеченная архитектура U-Net показала наилучшую производительность в терминах времени (0.05 с на срез) и самую высокую точность сегментации (2D DSC=0.77, AUC-PR=0.844) среди рассмотренных СНС для тестовой выборки изображений. Для срезов, не содержащих хрящевую ткань, при использовании данной архитектуры коэффициент DSC повысился от 0.21 до 0.75. Таким образом, переход к использованию архитектуры на основе U-Net обеспечил более однородную сегментацию трехмерных изображений, чем метод с применением сверточной нейронной сети, обучавшейся на основе патчей. Segmentation of cartilage tissue in 3D magnetic resonance (MR) images is used to determine the stage of degenerative and inflammatory joints diseases. For the wrist joint, manual segmentation is of immense complexity task due to its complex structure. It determines the relevance of the development of fully automatic segmentation methods. The only automated method previously proposed is based on deep learning. It provided non-uniform segmentation accuracy depending on the slice position within the 3D image. This work aims to improve the accuracy of automatic segmentation of cartilage tissue in lateral slices of wrist joint MR images using deep convolutional neural networks (CNN). Two CNN architectures were considered: a classical U-Net architecture and a truncated version of U-Net, in which the last block of convolutions was removed in the way of decreasing the dimension. The segmentation accuracy was assessed using 3D and 2D Sørensen–Dice coefficients (DSC), as well as by calculating the area under the precision-recall curve (AUC-PR). The results were compared with previously published data for an automated method of cartilage segmentation of the wrist joint using a patch-based CNN, as well as with published results for a manual segmentation procedure. The use of U-Net-based architectures has significantly improved the automatic segmentation accuracy. The truncated U-Net architecture showed the best performance in terms of time (0.05 s per slice) and the highest segmentation accuracy (2D DSC = 0.77, AUC-PR = 0.844) among the reviewed CNNs for the test dataset of images. For sections without cartilage, the DSC increased from 0.21 to 0.75 using this architecture. Thus, the use of the U-Net architecture provided more uniform segmentation of 3D images than the method using the patch-based convolutional neural network.

show abstract

Automatic knee cartilage and menisci segmentation from 3D-DESS MRI using deep semi-supervised learning

Cited by 3 publications

References 0 publications

Rapid X-Ray-Based 3-D Finite Element Modeling of Medial Knee Joint Cartilage Biomechanics During Walking

Rapid X-Ray-Based 3-D Finite Element Modeling of Medial Knee Joint Cartilage Biomechanics During Walking

Review of automated segmentation approaches for knee images

Нейросетевые решения на основе архитектуры U-Net для автоматической сегментации хрящевой ткани лучезапястного сустава на МР изображениях

Contact Info

Product

Resources

About