|
Authors | Methodology | Database used | Performance metrics observed accuracy (Acc) and area under the curve (AUC) | Inference |
|
[15] | 3D CNN-AlexNet detection algorithm | LUNA | Acc—89% | The drawback of the AlexNet model is tested with 10% of data which is not efficient in medical real-time analysis |
|
[16] | CNN predictor | LUAD | Acc—71% | The limitation of this model is not focused on preprocessing and segmentation which improves the detection accuracy |
|
[9] | SVM, KNN classifiers | LIDC-IDRI, LUNA 16 | Acc—91% | Elapse time complexity is high |
|
[17] | DenseNet classifier | LIDC-IDRI | Acc—90.85% | Insignificant database only used |
|
[18] | Unsupervised learning algorithms | LIDC-IDRI | Acc—94.3% | Tissue-based classification model has been focused |
|
[19] | CNN and RNN | LUNA16, ANODE09, and LIDC-IDRI | Acc—91%, AUC—0.78 | Modelled as binary classifier and the distinct stages are not focused |
|
[20] | 3D-CNN model | | Acc—92.65% | These labelled models are detected as benign or malignant |
|
[21] | BPSO-DT | LUNA | Acc—88.25% | Accuracy can be improved with other trending algorithms |
|
[22] | Supervised learning algorithms | LIDC-IDRI | Acc—89.5% | The proposed model focused on cell classification not for segmentation of detection of lung nodules |
|
[23] | Tobacco Exposure Pattern classification model | LUAD | Acc—94.6% | TEP focused on pattern recognition of tobacco as genetic problems |
|
Proposed method | Multistage classifier with cloud connectivity is the new work in the lung tumor detection problem which has been modelled and achieved better results compared to the existing works | Two different databases are used for evaluation LIDC-DICOM and PET scans | Acc—98.9% | The proposed model can be extended in terms of security perspective on the cloud service usages |
|