AI-Driven Computational Models for Lung Cancer Diagnosis: A Systematic Review and Meta-Analysis
Preprint
- 26 March 2025
- preprint
- Published by MDPI AG in Preprints.org
Abstract
Lung cancer remains the leading cause of cancer-related mortality worldwide, with early detection being critical for improving survival rates. While artificial intelligence (AI) has shown promise in enhancing lung cancer diagnosis and prognosis, challenges such as model generalizability, dataset diversity, and clinical validation hinder its widespread adoption. In recent years, hybrid approaches integrating deep learning with radiomics have gained attention for their potential to improve accuracy, interpretability, and robustness in lung cancer prediction. This systematic review and meta-analysis examine studies published between 2015 and 2023, focusing on hybrid AI-radiomics models that combine handcrafted radiomic features with deep learning architectures such as CNN, U-Net, and VGG-16. Additionally, machine learning classifiers like XGBoost, Random Forest, and SVM are explored in the context of radiomic feature analysis. Beyond performance evaluation, this study investigates dataset diversity, clinical validation challenges, and regulatory concerns affecting the translation of hybrid models into clinical practice. To provide a quantitative synthesis of current evidence, we conduct a meta-analysis of existing studies, assessing the effectiveness and reliability of hybrid AI-radiomics approaches compared to standalone AI models. Furthermore, an independent benchmarking experiment on the LIDC-IDRI dataset is performed to empirically validate the findings, demonstrating the superior performance of hybrid models in lung cancer diagnosis. Unlike previous reviews, this study fills a critical gap by combining both a systematic review and meta-analysis, offering a comprehensive evaluation of hybrid AI-radiomics models. The meta-analysis provides quantitative validation of these models' effectiveness, ensuring a more rigorous assessment of their real-world applicability. By identifying key limitations and opportunities, this review aims to bridge the gap between research and clinical implementation, offering insights for the development of more explainable, generalizable, and ethically responsible AI-driven solutions for lung cancer diagnosis.Keywords
All Related Versions
This publication has 62 references indexed in Scilit:
- Enhanced lung cancer subtype classification using attention-integrated DeepCNN and radiomic features from CT images: a focus on feature reproducibilityDiscover Oncology, 2025
- Landscape of 2D Deep Learning Segmentation Networks Applied to CT Scan from Lung Cancer Patients: A Systematic ReviewJournal of Imaging Informatics in Medicine, 2025
- Diagnostic accuracy of radiomics and artificial intelligence models in diagnosing lymph node metastasis in head and neck cancers: a systematic review and meta-analysisNeuroradiology, 2024
- Machine learning for administrative health records: A systematic review of techniques and applicationsArtificial Intelligence in Medicine, 2023
- A comparative study for the evaluation of CT-based conventional, radiomic, combined conventional and radiomic, and delta-radiomic features, and the prediction of the invasiveness of lung adenocarcinoma manifesting as ground-glass nodulesClinical Radiology, 2022
- Harmony search: Current studies and uses on healthcare systemsArtificial Intelligence in Medicine, 2022
- Mapping of global, regional and national incidence, mortality and mortality-to-incidence ratio of lung cancer in 2020 and 2050International Journal of Clinical Oncology, 2022
- A Systematic Review of Artificial Intelligence Techniques in Cancer Prediction and DiagnosisArchives of Computational Methods in Engineering, 2021
- A survey on incorporating domain knowledge into deep learning for medical image analysisMedical Image Analysis, 2021
- A survey on deep learning in medical image analysisMedical Image Analysis, 2017