Table 1.
Summary of papers for tuberculosis detection using deep learning.
| Authors | Deep Learning Technique | Features | Dataset |
|---|---|---|---|
| [74] | CNN with transfer learning and data augmentation | Features extracted from CNN | Montgomery |
| [38] | K-nearest neighbour, Simple Linear Regression and Sequential Minimal Optimisation (SMO) Classification | Area, major axis, minor axis, eccentricity, mean, kurtosis, skewness and entropy | Shenzhen |
| [84] | ViDi | Features extracted from CNN | Unspecified |
| [64] | CNN | Gabor, LBP, SIFT, PHOG and Features extracted from CNN | Private dataset |
| [24] | CNN | Features extracted from CNN | ImageCLEF 2018 dataset |
| [62] | CNN with transfer learning, with demographic information | Features extracted from CNN + demographic information | Private dataset |
| [79] | CNN with data augmentation, and ensemble by weighted averages of probability scores | Features extracted from CNN | Montgomery, Shenzhen, Belarus, JSRT |
| [70] | CNN with transfer learning and data augmentation | Features extracted from CNN | Private dataset, Montgomery, Shenzhen |
| [69] | CNN | Features extracted from CNN | Private datasets, Montgomery, Shenzhen |
| [71] | CNN with transfer learning and ensemble by simple linear probabilities averaging | Features extracted from CNN + rule-based features | Indiana, JSRT, Shenzhen |
| [29] | CNN | HoG features | ZiehlNeelsen Sputum smear Microscopy image DataBase |
| [75] | CNN and shuffle sampling | Features extracted from CNN | Private datasets |
| [81] | CNN with transfer learning and ensemble by averaging | CNN extracted features from edge images | Montgomery, Shenzhen |
| [57] | CNN with transfer learning, data augmentation and ensemble by weighted probability scores average | Features extracted from CNN | Private dataset, Montgomery, Shenzhen, Belarus |
| [85] | AutoEncoder-CNN | Features extracted from CNN | Private dataset |
| [76] | CNN with transfer learning and shuffle sampling | Features extracted from CNN | Private dataset |
| [65] | End-to-end CNN | Features extracted from CNN | Montgomery, Shenzhen |
| [88] | Optical flow model | Activity Description Vector on optical flow of video sequences | ImageCLEF 2019 dataset |
| [28] | CNN | Colours | TBimages dataset |
| [83] | Modified maximum pattern margin support vector machine (modified miSVM) | First four moments of the intensity distributions | Private datasets |
| [61] | CAD4TB with clinical information | Features extracted from CNN + clinical features | Private dataset |
| [31] | DBN | LoH + SURF features | ZiehlNeelsen Sputum smear Microscopy image DataBase |
| [60] | CAD4TB | Features extracted from CNN | Private dataset |
| [72] | CNN with transfer learning and data augmentation | Features extracted from CNN | Montgomery, Shenzhen, NIH-14 dataset |
| [30] | CNN | Features extracted from CNN | TBimages dataset |
| [63] | CNN from scratch and data augmentation | Features extracted from CNN | Montgomery, Shenzhen, Belarus |
| [86] | 3D CNN | Features extracted from CNN + lung volume + patient attribute metadata | ImageCLEF 2019 dataset |
| [12] | CNN with transfer learning and ensemble by stacking | local and global feature descriptors + features extracted from CNN | Private dataset, Montgomery, Shenzhen, India |
| [80] | CNN with transfer learning and feature level ensemble | Features extracted from CNN | Shenzhen |
| [15] | CNN with transfer learning and ensemble by averaging | CNN extracted features from edge images | Montgomery, Shenzhen |
| [32] | CNN with transfer learning | Features extracted from CNN | ZiehlNeelsen Sputum smear Microscopy image DataBase |
| [66] | CNN with data augmentation | Features extracted from CNN | Shenzhen |
| [73] | CNN with transfer learning and data augmentation | Features extracted from CNN | NIH-14, Montgomery, Shenzhen |
| [19] | CNN with transfer learning, Bag of CNN Features and ensemble by a simple soft-voting scheme | Features extracted from CNN + BOW | Private dataset, Montgomery, Shenzhen |
| [36] | Neural network | Shape, curvature descriptor histograms, eigenvalues of Hessian matrix | Montgomery, Shenzhen |
| [77] | CNN with transfer learning and data augmentation | Features extracted from CNN | Montgomery, Shenzhen, NIH-14 |
| [87] | 3D CNN | Features extracted from CNN | ImageCLEF 2019 dataset |
| [78] | CNN and Artificial Ecosystem-based Optimisation algorithm | Features extracted from CNN | Shenzhen |
| [67] | CNN | Features extracted from CNN | Shenzhen |
| [68] | Bayesian based CNN | Features extracted from CNN | Montgomery, Shenzhen |
| [82] | CNN with transfer learning, and ensemble by majority voting, simple averaging, weighted averaging, and stacking | Features extracted from CNN | Montgomery, Shenzhen, LDOCTCXR, 2018 RSNA pneumonia challenge dataset, Indiana dataset |