Table 2.

Overview of the AI algorithm and the ethical and legal concerns in the included studies.

Author (year)	AI technology	Application of AI technology	Categories of AI applications	Ethical principles related to the concern	Legal principles related to the concern
Vaidya et al. (2024) (45)	DL	To distinguish between lung adenocarcinoma and lung squamous cell carcinoma	Diagnosis	Algorithmic fairness and biases	-
Sangeetha et al. (2023) (40)	Multimodal Fusion Deep Neural Network	To integrate and process data from diverse sources, including medical images, genomic data, and clinical records; and to perform binary classification of lung cancer cases as cancerous or non-cancerous.	Diagnosis	Data privacy, informed consent to use, safety and transparency	Data protection and privacy, liability
Kumar et al. (2020) (38)	DL (RCNNa)	To detect lung cancer in radiological images and estimate the region of interest in the CT images.	Diagnosis	Data privacy	Data protection and privacy
Joshi et al. (2023) (41)	CNNb, SVMc	To identify various forms of lung cancer using radiological images.	Diagnosis	Data privacy, informed consent to use	-
Horry et al. (2021) (31)	DL, DTd	To stratify lung cancer patient CXR images from an independent dataset into benign/malignant categories.	Diagnosis	Data privacy	-
Field et al. (2021) (32)	Distributed learning approach, SVM	To extract and report on oncology data and validate an overall survival model in patients with unresectable Stage I–III NSCLC treated with radiotherapy	Prognosis	Data Privacy	Data protection and privacy
Fan et al. 2024 (39)	Federated learning algorithm, CNN	To solve the problem of small size and fragmentation of medical data, without exposing local private data by proposing federated learning for lung nodule detection	Screening	Data Privacy	Data protection and privacy, cybersecurity
Etienne et al. (2020) (43)	ML, DL (CNN)	To distinguish between benign and malignant nodules, detect nodules on chest radiographs, differentiate lung adenocarcinoma from squamous cell carcinoma using pathology slides, predict gene mutations, support decision-making for surgery patients by evaluating individual surgical risk factors, and adapt decision making individually, support Robotic-Assisted Surgery.	Screening, Diagnosis, Treatment, Prognosis	Data privacy, no harm to patients	Data protection and privacy, liability, safety and effectiveness
Cucchiara et al. (2021) (33)	ML, DL	To link patients’ clinical data with tumor molecular profiles and imaging characteristics; and to implement radiomics and liquid biopsy for integrated analysis	Diagnosis, Treatment, prognosis	-	Data protection and privacy, intellectual property law
Collmann et al. (1996) (46)	ANN	To distinguish true positives from false positives in the diagnosis of lung cancer	Diagnosis	No harm to patients	-
Bellini et al. (2021) (34)	DL(CNN), ML (XGBOOST, SVM, random forest, DT)	To diagnose and detect pulmonary nodules using CADx; to predict the risk of major complications and mortality following lung resection; to reduce hospital stay duration and postoperative complications through the use of surgical robotics; to distinguish between lung cancer types in pathological analysis; and to predict the risk of lung adenocarcinoma recurrence.	Diagnosis, Prognosis, Treatment	Informed consent to use, equity in access and use	Data protection and privacy
Adhikary et al. (2023) (42)	Deep neural network	To classify CT scanned images of three types of lung cancer	Diagnosis	Data privacy	-
Abbaker et al. (2024) (47)	DL (CNN, RNN, ANN)	To classify challenging cytological slide images from lung samples and predict lung cancer–related IHC phenotypes; to classify pulmonary nodules on CT scans and assist surgeons by identifying anatomical structures and aiding decision-making; to reduce delays in post-surgery diagnoses and estimate postoperative prognosis; to predict therapy responses, assess surgical risks, and support cancer staging; to predict genetic mutations such as ALK rearrangements and EGFR mutations; to estimate cardiorespiratory morbidity and postoperative outcomes; and to provide personalized drug treatment recommendations guiding targeted therapy selection and surgical planning.	Diagnosis, Treatment, Prognosis	Informed consent to use, safety and transparency, Algorithmic fairness and biases, Data Privacy, trust	Accountability
Zhang et al. (2021) (35)	DL (CNN), ML (SVM, DT), CDSS	To identify target sites in clinical images to assist imaging inspections using CADe and CADx systems; to analyze ambiguous morphology in histopathological images to support diagnosis; to detect minimal biomarker presence in liquid biopsy; to support clinical decision-making using a CDSS; to enhance surgical precision and reduce invasiveness via RATS; and to plan personalized treatment by regulating irradiation time, dose rate, and imaging in radiotherapy	Screening, Diagnosis, Treatment, Prognosis	Data privacy	Lack of regulation
Huang et al. (2021) (36)	SVMs, CNN, ANN, BN, Fuzzy C-means	To classify pulmonary nodules as benign or malignant	Diagnosis	Data privacy, No harm to patients	Data protection and privacy
Davri et al. (2023) (48)	ML, DL	To use histological data to assist in lung cancer diagnosis; to support prognosis estimation and mutational status assessment; to aid cytological interpretation; and to evaluate programmed cell death ligand 1 expression	Diagnosis, Prognosis	Data privacy	-
De Margerie-Mellon et al. (2022) (44)	CNN	To detect lung nodules using DL-based CADe algorithms in CXR and CT scans; to distinguish benign from malignant nodules using CADx; to assist in lung nodule segmentation; to predict mutations; to stratify patients into low- and high-mortality risk groups after radiotherapy and surgery; and to predict survival and cancer-specific outcomes	Screening, diagnosis, treatment, prognosis	Liability,	Liability
Kaliyugarasan et al. (2021) (37)	CNN	To classify pulmonary nodules as malignant or benign	Diagnosis	Safety and transparency	-
Kriegsmann et al. (2020) (49)	CNN	To differentiate the most common lung cancer subtypes	Diagnosis	No harm to patients	-
Rabbani et al. (2018) (50)	ML (DT, SVM), ANN	To detect solid, nonsolid, and cavitary nodules; to discriminate benign from malignant tumors; to identify genetic subtypes of NSCLC; to select the optimal radiation beam angle through dose–volume histogram predictions; to predict cancer subtype, tumor growth, metastatic potential, and patient survival; and to improve patient selection and prognostic models for predicting early mortality or treatment failure	Screening, Diagnosis, Prognosis, Treatment	Data privacy, data ownership	Lack of proper regulation, data protection and privacy, cybersecurity risks

^aRCNN, Region-based Convolutional Neural Network.

^bCNN, Convolutional Neural Network.

^cSVM, Support Vector Machine.

^dDT, Decision Tree.