Finally, a confidence level for a tear of the medial and lateral meniscus is computed by a softmax layer within the second dense layer. The sagittal and coronal images are processed by two distinct convolution blocks and the results are concatenated before being processed by the dense layers. Both sequences are the input for the deep convolutional neural network (CNN), represented by the middle box.
![difference between deep fritz 12 and fritz 15 difference between deep fritz 12 and fritz 15](https://i1.rgstatic.net/publication/336535300_Language_and_perception_Introduction_to_the_Special_Issue_Speakers_and_Listeners_in_the_Visual_World/links/5fc497b6a6fdcc6cc684cda1/largepreview.png)
#DIFFERENCE BETWEEN DEEP FRITZ 12 AND FRITZ 15 FULL#
Out of a full set of sequences of a knee MR examination, the algorithm selects a coronal and a sagittal fluid-sensitive fat-suppressed sequence with subsequent rescaling and cropping around the menisci. The top box represents the initial preprocessing step. Schematic illustration of the deep learning-based software.
![difference between deep fritz 12 and fritz 15 difference between deep fritz 12 and fritz 15](https://new.uschess.org/sites/default/files/styles/1080px_wide_scale/public/wp-thumbnails/Screenshot-2019-11-13-14.26.03.png)
Inter-reader agreement was very good for the medial (kappa = 0.876) and good for the lateral meniscus (kappa = 0.741).ĭCNN-based meniscus tear detection can be performed in a fully automated manner with a similar specificity but a lower sensitivity in comparison with musculoskeletal radiologists.Īrtificial intelligence Data accuracy Magnetic resonance imaging Neural networks (computer) Tibial meniscus injuries. The AUC-ROC of the DCNN was 0.882, 0.781, and 0.961 for detection of medial, lateral, and overall meniscus tear. Sensitivity for medial meniscus tears was significantly different between reader 2 and the DCNN (p = 0.039), and no significant differences existed for all other comparisons (all p ≥ 0.092). For lateral meniscus tear detection, sensitivity, specificity, and accuracy were for reader 1: 71%, 95%, and 89%, for reader 2: 67%, 99%, and 91%, and for the DCNN: 58%, 92%, and 84%. For medial meniscus tear detection, sensitivity, specificity, and accuracy were for reader 1: 93%, 91%, and 92%, for reader 2: 96%, 86%, and 92%, and for the DCNN: 84%, 88%, and 86%. Statistics included sensitivity, specificity, accuracy, ROC curve analysis, and kappa statistics.įifty-seven percent (57/100) of patients had a tear of the medial and 24% (24/100) of the lateral meniscus, including 12% (12/100) with a tear of both menisci.
![difference between deep fritz 12 and fritz 15 difference between deep fritz 12 and fritz 15](https://steamcdn-a.akamaihd.net/steam/apps/763270/ss_a0974b6cae810d866c8d268bdd43ce8c29e0df6b.1920x1080.jpg)
Surgical reports served as the standard of reference. Included patients were not part of the training set of the DCNN. All MRI were evaluated for medial and lateral meniscus tears by two musculoskeletal radiologists independently and by DCNN. One hundred consecutive patients were retrospectively included, who underwent knee MRI and knee arthroscopy in our institution. To clinically validate a fully automated deep convolutional neural network (DCNN) for detection of surgically proven meniscus tears.