Integrasi Deep Learning dalam Skrining CT Dosis Rendah untuk Penilaian Multiorgan dan Deteksi Dini Kanker Paru: Telaah Literatur

Abstract

Low-dose Computed Tomography (CT) screening lowers lung cancer mortality, but a focus on nodules can overlook multiorgan comorbidities. This literature review evaluates the integration of deep learning for early detection of lung cancer and risk assessment. The current literature is searched through PubMed, Scopus, and Web of Science regarding the validation of deep learning algorithms on low-dose chest CT. The CNN model supports the detection, segmentation, measurement, and characterization of nodules, while radiomics strengthens malignant classifications. The system assesses coronary artery calcification, emphysema, and thoracic abnormalities. Deep learning equals or surpasses radiologist-related capabilities and has the potential to reduce workload and false-positive results. Nonetheless, clinical implementation still faces challenges, including algorithm generalization, the need for large amounts of annotated data, and prospective validation to ensure interpretability and reliability in real practice. Challenges include generalization, annotation, interpretability, and prospective validation. Deep learning integration promises efficient, accurate, multiorgan chest screening; clinical implementation requires external standardization and validation. In addition, the utilization of this technique allows for the simultaneous evaluation of coronary artery calcification loads, which serves as an important predictor of cardiovascular disease risk in patients undergoing lung cancer screening.