An original deep learning model using limited data for COVID‐19 discrimination: A multicenter study

Abstract

Objectives Artificial intelligence (AI) has been proved to be a highly efficient tool for COVID‐19 diagnosis, but the large data size and heavy label force required for algorithm development and the poor generalizability of AI algorithms, to some extent, limit the application of AI technology in clinical practice. The aim of this study is to develop an AI algorithm with high robustness using limited chest CT data for COVID‐19 discrimination. Methods A three dimensional algorithm that combined multi‐instance learning with the LSTM architecture (3DMTM) was developed for differentiating COVID‐19 from community acquired pneumonia (CAP) while logistic regression (LR), k‐nearest neighbor (KNN), support vector machine (SVM), and a three dimensional convolutional neural network set for comparison. Totally, 515 patients with or without COVID‐19 between December 2019 and March 2020 from five different hospitals were recruited and divided into relatively large (150 COVID‐19 and 183 CAP cases) and relatively small datasets (17 COVID‐19 and 35 CAP cases) for either training or validation and another independent dataset (37 COVID‐19 and 93 CAP cases) for external test. Area under the receiver operating characteristic curve (AUC), sensitivity, specificity, precision, accuracy, F1 score, and G‐mean were utilized for performance evaluation. Results In the external test cohort, the relatively large data‐based 3DMTM‐LD achieved an AUC of 0.956 (95% confidence interval, 95% CI, 0.929∼0.982) with 86.2% and 98.0% for its sensitivity and specificity. 3DMTM‐SD got an AUC of 0.937 (95% CI, 0.909∼0.965), while the AUC of 3DCM‐SD decreased dramatically to 0.714 (95% CI, 0.649∼0.780) with training data reduction. KNN‐MMSD, LR‐MMSD, SVM‐MMSD, and 3DCM‐MMSD benefited significantly from the inclusion of clinical information while models trained with relatively large dataset got slight performance improvement in COVID‐19 discrimination. 3DMTM, trained with either CT or multi‐modal data, presented comparably excellent performance in COVID‐19 discrimination. Conclusions The 3DMTM algorithm presented excellent robustness for COVID‐19 discrimination with limited CT data. 3DMTM based on CT data performed comparably in COVID‐19 discrimination with that trained with multi‐modal information. Clinical information could improve the performance of KNN, LR, SVM, and 3DCM in COVID‐19 discrimination, especially in the scenario with limited data for training.

Classification of COVID-19 from community-acquired pneumonia: Boosting the performance with capsule network and maximum intensity projection image of CT scans

“…The results in Table 6 show that our proposed method achieves an accuracy of 0.970, outperforming all state-of-the-art methods except for our previous approach. The sensitivity of our method is higher than or comparable to that of the pipeline mimicking radiologist [ 13 ], a combination of CNN and SVM [ 55 ], multi-instance learning and a long short-term memory (LSTM) network [ 56 ], weakly supervised multi-scale learning [ 4 ], [ 12 ], a 2D CNN [ 57 ], a semi-supervised learning strategy with multi-view fusion [ 58 ], the BigBiGAN framework [ 59 ], the pretrained EfficientNet-b7 [ 60 ], and 3D ResNet-34 with attention modules [ 23 ].…”

“… Sen. Spe. AUC Our proposed method 156 patients (56 COVID-19 and 100 CAP) Lung segmentation MIP 0.970 0.971 0.968 0.986 HU et al, 2022 [ 4 ] 450 patient scans (150 of COVID-19, CAP and NP) Lung segmentation Weakly supervised multi-scale learning 0.891 0.870 0.862 0.906 Qi et al, 2022 [ 13 ] 157 patients (57 COVID-19 and 100 CAP) Lung segmentation Selection of slices with lesions Slice-level prediction Patient-level prediction 0.971 0.959 0.981 0.992 Ibrahim et al, 2022 [ 29 ] 2984 patients (COVID-19: 1396; non-COVID-19: 1588) VGGNet Convolutional deep belief network High-resolution network 0.95 0.95 0.96 Erdal et al, 2022 [ 55 ] 2496 CT scans (1428 COVID‐19 and 1068 CAP) Deep CNN for feature extraction SVM classification 0.932 0.858 0.993 0.984 Xu et al, 2022 [ 56 ] 515 patients (204 COVID-19 and 311 CAP) Multi-instance learning LSTM – …”