Lung cancer is the second most common cancer in both males and females, with the highest mortality worldwide, causing 21% of total cancer-related deaths. The notion of artificial intelligence (AI) was initially proposed by John McCarthy in 1956. It involves using computer systems and technology to replicate human-like intelligent behavior and critical thinking abilities. In the realm of medicine, AI is divided into two main categories: virtual and physical. The virtual branch is further categorized into machine learning (ML) and deep learning (DL).
Author, Year | Dataset | AI Algorithm | Outcomes | Results |
---|---|---|---|---|
Ardila et al., 2019 [26][7] | Low-dose CT scan | Deep learning algorithm | Diagnosis of lung cancer | AUC = 0.94 |
Delzell et al., 2019 [47][11] | CT scan of 200 lung nodules | Radiomics | Verify nodules as benign or malignant | AUC = 0.72 |
Schwyzer et al., 2018 [48][12] | FDG-PET imaging | Deep machine learning | Diagnosis of lung cancer using ultra-low-dose PET scans | Sensitivity = 95.9% Specificity = 98.1% |
Liu et al., 2023 [44][13] | Images and radiological features of 5251 patients from 14 studies | ANN SVM |
Diagnosis of lung cancer | Sensitivity = 87% Specificity = 87% |
Zheng et al., 2022 [49][14] | CT images of 9 NSCLC studies | Radiomics Deep learning |
To diagnose whether patient had NSCLC | AUROC = 0.78 |
Sun et al., 2020 [50][15] | Pure ground glass nodules of 385 patients | Radiomics | Invasiveness prediction | AUC = 0.77 |
Feng et al., 2019 [51][16] | Sub-solid nodules of 100 patients | Radiomics | Differentiate minimally invasive and invasive adenocarcinoma | AUC = 0.912 |
Avanzo et al., 2020 [52][17] | Nodules of low-dose CT scan | SVM | Differentiate adenocarcinoma from focal pneumonia | Accuracy = 87.6% |
Aydin et al., 2021 [53][18] | 301 lung cancer CT scans | CNN | Differentiate into squamous cell, adenocarcinoma, and small cell carcinoma | Sensitivity = 90% Specificity = 44% |
Chen et al., 2020 [54][19] | CT radiomics of 69 lung cancer patients | Radiomics | Differentiate NSCLC from SCLC | AUC = 0.93 |
Yu et al., 2016 [7][20] | 2480 histopathological images of lung adenocarcinoma | SVM Random forest |
Distinguish malignant tumors from healthy tissue | AUC = 0.81 |
Teramoto et al., 2017 [55][21] | 298 histopathological images | Conventional deep neural networks | Classified adenocarcinoma, squamous cell carcinoma, and small cell carcinoma | Accuracy = 89%, 60%, 70% respectively |
Coudray et al., 2018 [56][22] | Pathological images of adenocarcinoma | Conventional deep neural networks | Predicted 10 most prevalent genes in adenocarcinoma | Accuracy = 73.3%–85.6% |
Flores-Fernandez et al., 2012 [57][23] | Serum biomarkers of 63 lung cancer patients | Artificial neural network modeling | Correctly classifying lung cancer patients based on biomarker panel | Correct classification rate = 93.3% |