Fusion of CNN1 and CNN2-based magnetic resonance image diagnosis of knee meniscus injury and a comparative analysis with computed tomography. (November 2021)

• Record Type: Journal Article
• Authors: Qiu, Xubin; Liu, Zhiwei; Zhuang, Ming; Cheng, Dong; Zhu, Chenlei; Zhang, Xiaoying
• Is Part Of: Computer methods and programs in biomedicine. Volume 211(2021)
• Publication Date: 2021-11
• Subjects: Knee joint -- Meniscus injury -- Magnetic Resonance Imaging -- Convolutional neural network -- Diagnostic value; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2021.106297
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlight: We propose a meniscus detection method based on Fusion of CNN1 and CNN2 (CNNf), which uses Magnetic Resonance Imaging (MRI) and Computer tomography (CT) to compare the diagnosis results, verifies the proposed method through 2460 images collected from 205 patients in the hospital. We used accuracy, sensitivity, specificity, receiver operating characteristics (ROC), and damage total rate to evaluate performance. The accuracy of our model was 93.86%, the sensitivity was 91.35%, the specificity was 94.65%, and the area under the receiver operating characteristic curve was 96.78%. The total damage rate of MRI is 91.57%, which is far greater than the total damage rate of CT diagnosis of 80.13%. Abstract: Purpose: We used convolutional neural network (CNN) technology to improve the accuracy of diagnosis of knee meniscus injury and shorten the diagnosis time. Method: We propose a meniscus detection method based on Fusion of CNN1 and CNN2 (CNNf), which uses Magnetic Resonance Imaging (MRI) and Computer tomography (CT) to compare the diagnosis results, verifies the proposed method through 2460 images collected from 205 patients in the hospital. We used accuracy, sensitivity, specificity, receiver operating characteristics (ROC), and damage total rate to evaluate performance. Results: The accuracy of our model was 93.86%, the sensitivity was 91.35%, the specificity was 94.65%, and the area under the receiver operating characteristic curve was 96.78%. The total damage rate of MRI is 91.57%, which is far greater than the total damage rate of CT diagnosis of 80.13%. Conclusion: CNNf-based MRI technology of knee meniscus injury has high practical value in clinical practice. It can effectively improve the accuracy of diagnosis and reduce the rate of misdiagnosis.