Table 1.
Methods of studies included in descriptive analysis.
| First Author, Year | Imaging Modality | Computer Software/Model | Reference Standard | Accuracy Measures |
|---|---|---|---|---|
| Maduskar et al., 2013 [12] | CXR | CAD4TB (v 1.08) | AFB smear, MTB culture | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Muyoyeta et al., 2014 [13] | CXR | CAD4TB (v 1.08) | Xpert MTB/RIF, human reader | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Steiner et al., 2015 [14] | CXR | CAD4TB (v 3.07) | Human reader | AUC |
| Melendez et al., 2018 [15] | CXR | CAD4TB (v 5) | Human reader | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Zaidi et al., 2018 [16] | CXR | CAD4TB (v 3.07) | Xpert MTB/RIF | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Qin et al., 2019 [17] | CXR | CAD4TB (v 6), qXR (v 2), Lunit INSIGHT CXR (v 4.7.2) |
Xpert MTB/RIF | TP, FP, TN, FN, AUC, ACC, Sn, Sp |
| Philipsen et al., 2019 [18] | CXR | CAD4TB (v 5) | Xpert MTB/RIF, human reader | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Murphy et al., 2020 [19] | CXR | CAD4TB (v 6) | Xpert MTB/RIF | TP, FP, TN, FN, AUC, Sn, Sp |
| Nash et al., 2020 [20] | CXR | qXR (v 2) | AFB smear, Xpert MTB/RIF or MTB culture | AUC, Sn, Sp |
| Soares et al., 2023 [21] | CXR | CAD4TB (v 6), Lunit INSIGHT CXR (v 3.1.0.0), qXR (v 3) |
Xpert MTB/RIF, MTB culture | AUC, Sn, Sp, PPV, NPV |
| Qin et al., 2021 [6] | CXR | CAD4TB (v 7), InferRead DR (v 2), Lunit INSIGHT CXR (v 4.9.0), JF CXR-1 (v 2), qXR, (v 3) |
Xpert MTB/RIF | AUC, Sn, Sp |
| Breuninger et al., 2014 [22] | CXR | CAD4TB (v 3.07) | AFB smear, MTB culture | Sn, Sp, PPV, NPV |
| Khan et al., 2020 [23] | CXR | qXR (v 2), CAD4TB (v 6) |
MTB culture | ACC, Sn, Sp, PPV, NPV |
| Young et al., 2020 [24] | CXR | Not named | Human reader | AUC, Sn, Sp |
| Liao et al., 2022 [25] | CXR | JF CXR-1 (v 2) | Human reader | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Codlin et al., 2021 [26] | CXR | qXR (v 3), CAD4TB (v 7), Genki (v 2), Lunit INSIGHT CXR (v 3.1.0.0), JF CXR-1 (v 3.0), InferRead DR Chest (v 1.0.0.0), ChestEye (v 1), T-Xnet (v 1), XrayAME (v 1), COTO (v 1), SemanticMD (v 1), Dr CADx (v 0.1) |
Xpert MTB/RIF | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Habib et al., 2020 [27] | CXR | CAD4TB (v 3.07) | Xpert MTB/RIF | AUC, Sn, Sp, PPV, NPV |
| Koesoemadinata et al., 2018 [28] | CXR | CAD4TB (v 5) | Composite reference standard(s) | AUC, Sn, Sp |
| Lee et al., 2020 [29] | CXR | Lunit INSIGHT CXR (v 4.7.2) | MTB culture, AFB smear, TB polymerase chain reaction, human reader | TP, FP, TN, FN, AUC, ACC, Sn, Sp, PPV, NPV |
| Gelaw et al., 2022 [30] | CXR | CAD4TB (v 6), Lunit INSIGHT CXR (v 4.9.0), qXR (v 2) |
Xpert MTB/RIF, Mycobacterium tuberculosis (MTB) culture | TP, FP, TN, FN, Sn, Sp |
| Ehrlich et al., 2022 [31] | CXR | CAD4TB (v 7) | Human reader | TP, FP, TN, FN, AUC, Sn, Sp |
| Kagujje et al., 2022 [32] | CXR | CAD4TB (v 7), qXR (v 3) |
Xpert MTB/RIF | TP, FP, TN, FN, AUC, Sn, Sp |
| Tavaziva et al., 2022 [33] | CXR | Lunit INSIGHT CXR (v 4.9.0) | Xpert MTB/RIF, Mycobacterium tuberculosis (MTB) culture | TP, FP, TN, FN, AUC, ACC, Sn, Sp |
| Shen et al., 2010 [34] | CXR | Bayesian classifier | Human reader | ACC |
| Melendez et al., 2015 [35] | CXR | si-miSVM+PEDD | Human reader | AUC |
| Pasa et al., 2019 [36] | CXR | CNN | Human reader | AUC, ACC |
| Xie et al., 2020 [37] | CXR | RCNN | Human reader | AUC, ACC, Sn, Sp |
| Ma et al., 2020 [38] | CT | U-Net | Sputum smear | AUC, ACC, Sn, Sp, PPV, NPV |
| Rajpurkar et al., 2020 [39] | CXR | DenseNet | Xpert MTB/RIF, MTB culture | ACC, Sn, Sp |
| Oloko-Oba et al., 2021 [40] | CXR | EfficientNets | Human reader | AUC, ACC, Sn, Sp |
| Mamalakis et al., 2021 [41] | CXR | DenseNet-121, ResNet-50 |
Human reader | AUC, F1, precision, recall |
| Rajakumar et al., 2021 [42] | CXR | VGG16, VGG19, KNN |
Human reader | ACC, Sn, Sp, NPV |
| Sharma et al., 2021 [43] | CXR | Tree, SVM, Naïve Bayes |
Composite reference standard(s) | AUC, F1, CA, precision, recall |
| Wang et al., 2021 [44] | CT | 3D-ResNet | AFB smear, MTB culture | AUC, Sn, Sp, ACC, F1 |
| Showkatian et al., 2022 [45] | CXR | ConvNet | Human reader | AUC, ACC, F1, precision, recall |
| Zhou et al., 2022 [46] | CXR | ResNet | Human reader | AUC, ACC, Sn, Sp, PPV, NPV |
| Rajaraman et al., 2021 [47] | CXR | ImageNet, VGG-16 |
Human reader | AUC, ACC, Sn, Sp, F1, precision |
| Yan et al., 2021 [48] | CT | SeNet-ResNet-18 | Human reader | ACC, precision, recall |
| Zhang et al., 2021 [49] | CT | CBIR-CSNN | Composite reference standard(s) | AUC, ACC |
| Arzhaeva et al., 2009 [50] | CXR | MVDB | Human reader | AUC |
| Jaeger et al., 2014 [51] | CXR | SVM | Human reader | AUC, ACC |
| Chauhan et al., 2014 [52] | CXR | SVM | Human reader | AUC, ACC, Sn, Sp, F1, precision |
| Hogeweg et al., 2015 [53] | CXR | RF50, GB50, LDA, KNN13 |
MTB culture, human reader | AUC |
| Lakhani et al., 2017 [54] | CXR | AlexNet, GoogLeNet |
Human reader | AUC, ACC, Sn, Sp |
| Han et al., 2021 [55] | CXR | VGG16 | Human reader | AUC, Sn, Sp |
| An et al., 2022 [56] | CXR | E-TBNet (ResNet) | Human reader | ACC, Sn, Sp, NPV, ppv, F1 |
| Lee et al., 2021 [57] | CXR | EfficientNet | Xpert MTB/RIF, MTB culture, human reader | AUC |
| Khatibi et al., 2021 [58] | CXR | CNN, CCNSE |
Human reader | AUC, ACC |
| Kim et al., 2020 [59] | CXR | DCNN | Human reader | AUC, Sn, Sp, NPV, PPV, F1 |
| Feng et al., 2020 [60] | CT | CNN | Composite reference standard(s) | AUC, ACC, Sn, Sp |
| Hwang et al., 2019 [61] | CXR | CNN | Human reader | AUC, Sn, Sp |
| Heo et al., 2019 [62] | CXR | I-CNN(VGG19), D-CNN(VGG19) |
Human reader | AUC |
| Aguiar et al., 2016 [63] | CXR | MLP | Human reader | AUC, Sn, Sp, PPV, NPV |
| Faruk et al., 2021 [64] | CT | Xception, InceptionV3, InceptionResNetV2, MobileNetV2 |
Human reader | Sn, precision, recall, F1 |
| Karki et al., 2021 [65] | CXR | InceptionV3, Xception |
Human reader | AUC |
| Dasanayaka et al., 2021 [66] | CXR | VGG16, InceptionV3, Ensemble |
Human reader | ACC, Sn, Sp |
| Govindarajan et al., 2021 [67] | CXR | ELM, OSELM |
Human reader | Sn, Sp, precision, F1 |
| Acharya et al., 2022 [68] | CXR | ImageNet fine-tuned normalization-free networks | Human reader | Sn, Sp, AUC, ACC, precision, recall |
| Kadry et al., 2022 [69] | CXR | VGG16, Fine Tree | Xpert MTB/RIF, Mycobacterium tuberculosis (MTB) culture, human reader | Sn, Sp, ACC, NPV |
| Kazemzadeh et al., 2023 [70] | CXR | NR | Human reader | Sn, Sp, AUC |
| Margarat et al., 2022 [71] | CXR | DBN-AMBO | Human reader | Sp, ACC, precision, recall, NPV |
Abbreviations: CXR, chest X-ray; CT, computed tomography; CAD, computer-aided detection; CNN, convolutional neural networks; RCNN, regions with CNN features; KNN, K-nearest neighbor; VGG, visual geometry group; SVM, support vector machine; HIV, human immunodeficiency virus; DLAD, deep-learning-based automatic detection; AFB, acid-fast bacilli; MTB, Mycobacterium tuberculosis; TP, true positive; FP, false positive; TN, true negative; FN, false negative; AUC, area under the receiver operating curve; ACC, accuracy; Sn, sensitivity; Sp, specificity; CA, cluster accuracy; DBN-AMBO, deep belief network with adaptive monarch butterfly optimization.