@article{SekuboyinaHusseiniBayatetal., author = {Sekuboyina, Anjany and Husseini, Malek E. and Bayat, Amirhossein and L{\"o}ffler, Maximilian and Liebl, Hans and Li, Hongwei and Tetteh, Giles and Kukačka, Jan and Payer, Christian and Štern, Darko and Urschler, Martin and Chen, Maodong and Cheng, Dalong and Lessmann, Nikolas and Hu, Yujin and Wang, Tianfu and Yang, Dong and Xu, Daguang and Ambellan, Felix and Amiranashvili, Tamaz and Ehlke, Moritz and Lamecker, Hans and Lehnert, Sebastian and Lirio, Marilia and de Olaguer, Nicol{\´a}s P{\´e}rez and Ramm, Heiko and Sahu, Manish and Tack, Alexander and Zachow, Stefan and Jiang, Tao and Ma, Xinjun and Angerman, Christoph and Wang, Xin and Brown, Kevin and Kirszenberg, Alexandre and Puybareau, {\´E}lodie and Chen, Di and Bai, Yiwei and Rapazzo, Brandon H. and Yeah, Timyoas and Zhang, Amber and Xu, Shangliang and Hou, Feng and He, Zhiqiang and Zeng, Chan and Xiangshang, Zheng and Liming, Xu and Netherton, Tucker J. and Mumme, Raymond P. and Court, Laurence E. and Huang, Zixun and He, Chenhang and Wang, Li-Wen and Ling, Sai Ho and Huynh, L{\^e} Duy and Boutry, Nicolas and Jakubicek, Roman and Chmelik, Jiri and Mulay, Supriti and Sivaprakasam, Mohanasankar and Paetzold, Johannes C. and Shit, Suprosanna and Ezhov, Ivan and Wiestler, Benedikt and Glocker, Ben and Valentinitsch, Alexander and Rempfler, Markus and Menze, Bj{\"o}rn H. and Kirschke, Jan S.}, title = {VerSe: A Vertebrae labelling and segmentation benchmark for multi-detector CT images}, series = {Medical Image Analysis}, volume = {73}, journal = {Medical Image Analysis}, doi = {10.1016/j.media.2021.102166}, abstract = {Vertebral labelling and segmentation are two fundamental tasks in an automated spine processing pipeline. Reliable and accurate processing of spine images is expected to benefit clinical decision support systems for diagnosis, surgery planning, and population-based analysis of spine and bone health. However, designing automated algorithms for spine processing is challenging predominantly due to considerable variations in anatomy and acquisition protocols and due to a severe shortage of publicly available data. Addressing these limitations, the Large Scale Vertebrae Segmentation Challenge (VerSe) was organised in conjunction with the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) in 2019 and 2020, with a call for algorithms tackling the labelling and segmentation of vertebrae. Two datasets containing a total of 374 multi-detector CT scans from 355 patients were prepared and 4505 vertebrae have individually been annotated at voxel level by a human-machine hybrid algorithm (https://osf.io/nqjyw/, https://osf.io/t98fz/). A total of 25 algorithms were benchmarked on these datasets. In this work, we present the results of this evaluation and further investigate the performance variation at the vertebra level, scan level, and different fields of view. We also evaluate the generalisability of the approaches to an implicit domain shift in data by evaluating the top-performing algorithms of one challenge iteration on data from the other iteration. The principal takeaway from VerSe: the performance of an algorithm in labelling and segmenting a spine scan hinges on its ability to correctly identify vertebrae in cases of rare anatomical variations. The VerSe content and code can be accessed at: https://github.com/anjany/verse.}, language = {en} } @inproceedings{EstacioEhlkeTacketal., author = {Estacio, Laura and Ehlke, Moritz and Tack, Alexander and Castro-Gutierrez, Eveling and Lamecker, Hans and Mora, Rensso and Zachow, Stefan}, title = {Unsupervised Detection of Disturbances in 2D Radiographs}, series = {2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI)}, booktitle = {2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI)}, doi = {10.1109/ISBI48211.2021.9434091}, pages = {367 -- 370}, abstract = {We present a method based on a generative model for detection of disturbances such as prosthesis, screws, zippers, and metals in 2D radiographs. The generative model is trained in an unsupervised fashion using clinical radiographs as well as simulated data, none of which contain disturbances. Our approach employs a latent space consistency loss which has the benefit of identifying similarities, and is enforced to reconstruct X-rays without disturbances. In order to detect images with disturbances, an anomaly score is computed also employing the Frechet distance between the input X-ray and the reconstructed one using our generative model. Validation was performed using clinical pelvis radiographs. We achieved an AUC of 0.77 and 0.83 with clinical and synthetic data, respectively. The results demonstrated a good accuracy of our method for detecting outliers as well as the advantage of utilizing synthetic data.}, language = {en} } @article{TackPreimZachow, author = {Tack, Alexander and Preim, Bernhard and Zachow, Stefan}, title = {Fully automated Assessment of Knee Alignment from Full-Leg X-Rays employing a "YOLOv4 And Resnet Landmark regression Algorithm" (YARLA): Data from the Osteoarthritis Initiative}, series = {Computer Methods and Programs in Biomedicine}, volume = {205}, journal = {Computer Methods and Programs in Biomedicine}, number = {106080}, doi = {https://doi.org/10.1016/j.cmpb.2021.106080}, abstract = {We present a method for the quantification of knee alignment from full-leg X-Rays. A state-of-the-art object detector, YOLOv4, was trained to locate regions of interests (ROIs) in full-leg X-Ray images for the hip joint, the knee, and the ankle. Residual neural networks (ResNets) were trained to regress landmark coordinates for each ROI.Based on the detected landmarks the knee alignment, i.e., the hip-knee-ankle (HKA) angle, was computed. The accuracy of landmark detection was evaluated by a comparison to manually placed landmarks for 360 legs in 180 X-Rays. The accuracy of HKA angle computations was assessed on the basis of 2,943 X-Rays. Results of YARLA were compared to the results of two independent image reading studies(Cooke; Duryea) both publicly accessible via the Osteoarthritis Initiative. The agreement was evaluated using Spearman's Rho, and weighted kappa as well as regarding the correspondence of the class assignment (varus/neutral/valgus). The average difference between YARLA and manually placed landmarks was less than 2.0+- 1.5 mm for all structures (hip, knee, ankle). The average mismatch between HKA angle determinations of Cooke and Duryea was 0.09 +- 0.63°; YARLA resulted in a mismatch of 0.10 +- 0.74° compared to Cooke and of 0.18 +- 0.64° compared to Duryea. Cooke and Duryea agreed almost perfectly with respect to a weighted kappa value of 0.86, and showed an excellent reliability as measured by a Spearman's Rho value of 0.99. Similar values were achieved by YARLA, i.e., a weighted kappa value of0.83 and 0.87 and a Spearman's Rho value of 0.98 and 0.99 to Cooke and Duryea,respectively. Cooke and Duryea agreed in 92\% of all class assignments and YARLA did so in 90\% against Cooke and 92\% against Duryea. In conclusion, YARLA achieved results comparable to those of human experts and thus provides a basis for an automated assessment of knee alignment in full-leg X-Rays.}, language = {de} } @misc{TackAmbellanZachow2021, author = {Tack, Alexander and Ambellan, Felix and Zachow, Stefan}, title = {Towards novel osteoarthritis biomarkers: Multi-criteria evaluation of 46,996 segmented knee MRI data from the Osteoarthritis Initiative (Supplementary Material)}, series = {PLOS One}, volume = {16}, journal = {PLOS One}, number = {10}, doi = {10.12752/8328}, year = {2021}, abstract = {Convolutional neural networks (CNNs) are the state-of-the-art for automated assessment of knee osteoarthritis (KOA) from medical image data. However, these methods lack interpretability, mainly focus on image texture, and cannot completely grasp the analyzed anatomies' shapes. In this study we assess the informative value of quantitative features derived from segmentations in order to assess their potential as an alternative or extension to CNN-based approaches regarding multiple aspects of KOA A fully automated method is employed to segment six anatomical structures around the knee (femoral and tibial bones, femoral and tibial cartilages, and both menisci) in 46,996 MRI scans. Based on these segmentations, quantitative features are computed, i.e., measurements such as cartilage volume, meniscal extrusion and tibial coverage, as well as geometric features based on a statistical shape encoding of the anatomies. The feature quality is assessed by investigating their association to the Kellgren-Lawrence grade (KLG), joint space narrowing (JSN), incident KOA, and total knee replacement (TKR). Using gold standard labels from the Osteoarthritis Initiative database the balanced accuracy (BA), the area under the Receiver Operating Characteristic curve (AUC), and weighted kappa statistics are evaluated. Features based on shape encodings of femur, tibia, and menisci plus the performed measurements showed most potential as KOA biomarkers. Differentiation between healthy and severely arthritic knees yielded BAs of up to 99\%, 84\% were achieved for diagnosis of early KOA. Substantial agreement with weighted kappa values of 0.73, 0.73, and 0.79 were achieved for classification of the grade of medial JSN, lateral JSN, and KLG, respectively. The AUC was 0.60 and 0.75 for prediction of incident KOA and TKR within 5 years, respectively. Quantitative features from automated segmentations yield excellent results for KLG and JSN classification and show potential for incident KOA and TKR prediction. The validity of these features as KOA biomarkers should be further evaluated, especially as extensions of CNN-based approaches. To foster such developments we make all segmentations publicly available together with this publication.}, language = {en} } @misc{TackShestakovLuedkeetal., author = {Tack, Alexander and Shestakov, Alexey and L{\"u}dke, David and Zachow, Stefan}, title = {A deep multi-task learning method for detection of meniscal tears in MRI data from the Osteoarthritis Initiative database}, issn = {1438-0064}, url = {http://nbn-resolving.de/urn:nbn:de:0297-zib-84415}, abstract = {We present a novel and computationally efficient method for the detection of meniscal tears in Magnetic Resonance Imaging (MRI) data. Our method is based on a Convolutional Neural Network (CNN) that operates on a complete 3D MRI scan. Our approach detects the presence of meniscal tears in three anatomical sub-regions (anterior horn, meniscal body, posterior horn) for both the Medial Meniscus (MM) and the Lateral Meniscus (LM) individually. For optimal performance of our method, we investigate how to preprocess the MRI data or how to train the CNN such that only relevant information within a Region of Interest (RoI) of the data volume is taken into account for meniscal tear detection. We propose meniscal tear detection combined with a bounding box regressor in a multi-task deep learning framework to let the CNN implicitly consider the corresponding RoIs of the menisci. We evaluate the accuracy of our CNN-based meniscal tear detection approach on 2,399 Double Echo Steady-State (DESS) MRI scans from the Osteoarthritis Initiative database. In addition, to show that our method is capable of generalizing to other MRI sequences, we also adapt our model to Intermediate-Weighted Turbo Spin-Echo (IW TSE) MRI scans. To judge the quality of our approaches, Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) values are evaluated for both MRI sequences. For the detection of tears in DESS MRI, our method reaches AUC values of 0.94, 0.93, 0.93 (anterior horn, body, posterior horn) in MM and 0.96, 0.94, 0.91 in LM. For the detection of tears in IW TSE MRI data, our method yields AUC values of 0.84, 0.88, 0.86 in MM and 0.95, 0.91, 0.90 in LM. In conclusion, the presented method achieves high accuracy for detecting meniscal tears in both DESS and IW TSE MRI data. Furthermore, our method can be easily trained and applied to other MRI sequences.}, language = {en} } @article{TackShestakovLuedkeetal., author = {Tack, Alexander and Shestakov, Alexey and L{\"u}dke, David and Zachow, Stefan}, title = {A deep multi-task learning method for detection of meniscal tears in MRI data from the Osteoarthritis Initiative database}, series = {Frontiers in Bioengineering and Biotechnology, section Biomechanics}, journal = {Frontiers in Bioengineering and Biotechnology, section Biomechanics}, doi = {10.3389/fbioe.2021.747217}, pages = {28 -- 41}, abstract = {We present a novel and computationally efficient method for the detection of meniscal tears in Magnetic Resonance Imaging (MRI) data. Our method is based on a Convolutional Neural Network (CNN) that operates on a complete 3D MRI scan. Our approach detects the presence of meniscal tears in three anatomical sub-regions (anterior horn, meniscal body, posterior horn) for both the Medial Meniscus (MM) and the Lateral Meniscus (LM) individually. For optimal performance of our method, we investigate how to preprocess the MRI data or how to train the CNN such that only relevant information within a Region of Interest (RoI) of the data volume is taken into account for meniscal tear detection. We propose meniscal tear detection combined with a bounding box regressor in a multi-task deep learning framework to let the CNN implicitly consider the corresponding RoIs of the menisci. We evaluate the accuracy of our CNN-based meniscal tear detection approach on 2,399 Double Echo Steady-State (DESS) MRI scans from the Osteoarthritis Initiative database. In addition, to show that our method is capable of generalizing to other MRI sequences, we also adapt our model to Intermediate-Weighted Turbo Spin-Echo (IW TSE) MRI scans. To judge the quality of our approaches, Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) values are evaluated for both MRI sequences. For the detection of tears in DESS MRI, our method reaches AUC values of 0.94, 0.93, 0.93 (anterior horn, body, posterior horn) in MM and 0.96, 0.94, 0.91 in LM. For the detection of tears in IW TSE MRI data, our method yields AUC values of 0.84, 0.88, 0.86 in MM and 0.95, 0.91, 0.90 in LM. In conclusion, the presented method achieves high accuracy for detecting meniscal tears in both DESS and IW TSE MRI data. Furthermore, our method can be easily trained and applied to other MRI sequences.}, language = {en} } @article{TackAmbellanZachow, author = {Tack, Alexander and Ambellan, Felix and Zachow, Stefan}, title = {Towards novel osteoarthritis biomarkers: Multi-criteria evaluation of 46,996 segmented knee MRI data from the Osteoarthritis Initiative}, series = {PLOS One}, volume = {16}, journal = {PLOS One}, number = {10}, doi = {10.1371/journal.pone.0258855}, abstract = {Convolutional neural networks (CNNs) are the state-of-the-art for automated assessment of knee osteoarthritis (KOA) from medical image data. However, these methods lack interpretability, mainly focus on image texture, and cannot completely grasp the analyzed anatomies' shapes. In this study we assess the informative value of quantitative features derived from segmentations in order to assess their potential as an alternative or extension to CNN-based approaches regarding multiple aspects of KOA. Six anatomical structures around the knee (femoral and tibial bones, femoral and tibial cartilages, and both menisci) are segmented in 46,996 MRI scans. Based on these segmentations, quantitative features are computed, i.e., measurements such as cartilage volume, meniscal extrusion and tibial coverage, as well as geometric features based on a statistical shape encoding of the anatomies. The feature quality is assessed by investigating their association to the Kellgren-Lawrence grade (KLG), joint space narrowing (JSN), incident KOA, and total knee replacement (TKR). Using gold standard labels from the Osteoarthritis Initiative database the balanced accuracy (BA), the area under the Receiver Operating Characteristic curve (AUC), and weighted kappa statistics are evaluated. Features based on shape encodings of femur, tibia, and menisci plus the performed measurements showed most potential as KOA biomarkers. Differentiation between non-arthritic and severely arthritic knees yielded BAs of up to 99\%, 84\% were achieved for diagnosis of early KOA. Weighted kappa values of 0.73, 0.72, and 0.78 were achieved for classification of the grade of medial JSN, lateral JSN, and KLG, respectively. The AUC was 0.61 and 0.76 for prediction of incident KOA and TKR within one year, respectively. Quantitative features from automated segmentations provide novel biomarkers for KLG and JSN classification and show potential for incident KOA and TKR prediction. The validity of these features should be further evaluated, especially as extensions of CNN- based approaches. To foster such developments we make all segmentations publicly available together with this publication.}, language = {en} }