Journal: Diagnostics

Article Title: A Systematic Review: The Role of Artificial Intelligence in Lung Cancer Screening in Detecting Lung Nodules on Chest X-Rays

doi: 10.3390/diagnostics15030246

Figure Lengend Snippet: AI versus radiologist performance in lung nodule detection across study parameters and outcomes.

Article Snippet: Hamanaka et al., 2024 [ ] , AI detection rate: 51.7% of all patients (90/174) 57.7% excluding adenocarcinoma in situ cases Software: EIRL X-ray Lung Nodule version 1.12 (LPIXEL Inc., Tokyo, Japan) , N/A , N/A , AI sensitivity by lung area: - Lung apices: 60% - Upper lung field: 63.6% - Middle lung field: 78.6% - Lower lung field: 43.8% - Hilar area: 30% AI sensitivity by tumor diameter: - ≤10 mm: 0 - 11–15 mm: 0.38 - 16–20 mm: 0.52 - ≥21 mm: >83% Lower lung fields were less frequently detected , N/A , AI challenges include overlapping anatomical structures, small nodules (<0.7 cm), oversight, and specific tumors such as adenocarcinoma in situ (AIS). Larger total histopathological tumor size improves AI detection Physician challenges include detecting smaller tumors (<1 cm) and GGNs AI is unable to make comparative diagnoses with past radiographs , AI missed all 18 cases of AIS AI missed 3 cases due to flat-appearing tumors on radiographs.

Techniques: Software, Biomarker Discovery, Imaging, Diagnostic Assay, In Situ