Table 2.
Performance of studies exploring classification of pulmonary nodules.
| Classification | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Author | Year | Deep Learning Architecture | Dataset for Training | Dataset for Testing | Categories for Testing | Sensitivity | Specificity | AUC | Accuracy |
| Alakwaa, Wafaa et al. [27] | 2017 | CNN | LUNA16 and DSB17 | DSB17 | Cancer vs. no cancer | N/A | N/A | N/A | 86.6 |
| Chen, Sihang et al. [22] | 2019 | CNN | Independent dataset | Independent dataset | Adenocarcinoma vs. benign | N/A | N/A | N/A | 87.5 |
| Ciompi, Francesco et al. [28] | 2015 | CNN | ImageNet and NELSON | NELSON | Peri-fissural nodules (PFN) vs. non-PFN | N/A | N/A | 84.7 | N/A |
| Ciompi, Francesco et al. *[29] | 2017 | CNN | MILD | DLCST | Multiple categories (overall) | N/A | N/A | N/A | 79.5 |
| Jakimovski, Goran et al. [30] | 2019 | CDNN | LONI database | LONI database | Cancer vs. no cancer | 99.9 | 98.7 | N/A | 99.6 |
| Lakshmanaprabu, S.K. et al. [31] | 2018 | ODNN | ELCAP | ELCAP | Abnormal vs. normal | 96.2 | 94.2 | N/A | 94.5 |
| Li, Li et al. * [17] | 2018 | CNN | LIDC-IDRI and NLST | Independent dataset | Multiple categories (overall) | N/A | N/A | N/A | N/A |
| Liao, Fangzhou et al. [23] | 2019 | CNN | LUNA16 and DSB17 | DSB17 | Cancer vs. no-cancer (scale) | N/A | N/A | 87 | 81.4 |
| Liu, Shuang et al. [32] | 2017 | CNN | NLST and ELCAP | NLST and ELCAP | Malign vs. benign | N/A | N/A | 78 | N/A |
| Liu, Xinglong et al. * [33] | 2017 | CNN | LIDC-IDRI | ELCAP | Multiple categories (overall) | N/A | N/A | N/A | 90.3 |
| Masood, Anum et al. [21] | 2018 | FCNN | LIDC-IDRI, RIDER, LungCT-Diagnosis, LUNA16, LISS, SPIE challenge dataset and Independent dataset | Independent dataset | Four stage categories (overall) | 83.7 | 96.2 | N/A | 96.3 |
| Nishio, Mizuho et al. [34] | 2018 | CNN | Independent dataset | Independent dataset | Benign, primary and metastic cancer (overall) | N/A | N/A | N/A | 68 |
| Onishi, Yuya et al. [35] | 2018 | DCNN | Independent dataset | Independent dataset | Malign vs. benign | N/A | N/A | 84.1 | 81.7 |
| Polat, Huseyin et al. [36] | 2019 | CNN | DSB17 | DSB17 | Cancer vs. no cancer | 88.5 | 94.2 | N/A | 91.8 |
| Qiang, Yan et al. [37] | 2017 | Deep SDAE-ELM | Independent dataset | Independent dataset | Malign vs. benign | 84.4 | 81.3 | N/A | 82.8 |
| Rangaswamy et al. [38] | 2019 | CNN | ILD | ILD | Malign vs. benign | 98 | 94 | N/A | 96 |
| Sori, Worku Jifara et al. [39] | 2018 | CNN | LUNA16 and DSB17 | DSB17 | Cancer vs. no cancer | 87.4 | 89.1 | N/A | 87.8 |
| Suzuki, Kenji * [19] | 2009 | MTANN | Independent dataset A | Independent dataset B | Malign vs. benign | 96 | N/A | N/A | N/A |
| Tajbakhsh, Nima et al. [20] | 2017 | CNN | Independent dataset | Independent dataset | Malign vs. benign | N/A | N/A | 77.6 | N/A |
| MTANN | Independent dataset | Independent dataset | Malign vs. benign | N/A | N/A | 88.1 | N/A | ||
| Wang, Shengping et al. [40] | 2018 | CNN | Independent dataset | Independent dataset | PIL vs. IAC | 88.5 | 80.1 | 89.2 | 84 |
| Wang, Yang et al. [25] | 2019 | RCNN | Independent dataset | Independent dataset | Malign vs. benign | 76.5 | 89.1 | 90.6 | 87.3 |
| Yuan, Jingjing et al. * [41] | 2017 | CNN | LIDC-IDRI | ELCAP | Multiple categories (overall) | N/A | N/A | N/A | 93.9 |
| Zhang, Chao et al. * [42] | 2019 | CNN | LUNA16, DSB17 and Independent dataset(A) | Independent dataset(B) | Malign vs. benign | 96 | 88 | N/A | 92 |
Studies marked with * are studies where test dataset was different from training dataset. Abbreviations: massive training artificial neural network (MTANN), convolutional neural network (CNN), deep neural network (DNN), lung image database consortium and image database resource initiative (LIDC-IDRI), the Dutch–Belgian randomized lung cancer screening trial (Dutch acronym; NELSON), multicentric Italian lung detection (MILD), laboratory of neuro imaging (LONI), early lung cancer action program (ELCAP), reference image database to evaluate therapy response (RIDER), Society of Photo-Optical Instrumentation Engineers (SPIE), lung nodule analysis 2016 (LUNA16), lung CT imaging signs (LISS), Kaggle data science bowl 2017 (DSB17), interstitial lung disease (ILD), Danish lung cancer screening trial (DLCST), automatic nodule detection 2009 (ANODE09), pre-invasive lesions (PIL), invasive adenocarcinomas (IAC).