2024 World Conference on Lung Cancer (WCLC) - Posters-P4.04C.08 Sensitivity of Artificial Intelligence in Low-dose Computed Tomography Screening for Lung Cancer

P4.04C.08 Sensitivity of Artificial Intelligence in Low-dose Computed Tomography Screening for Lung Cancer

Pdf Summary

The study examines the potential of artificial intelligence (AI) to enhance the efficiency of low-dose computed tomography (LDCT) screening for lung cancer, focusing on alleviating the burden on radiologists. The study seeks to determine if AI systems can optimize the screening workflow by acting as primary or supporting CT readers.<br /><br />The research involved 1,000 participants from the Norwegian Lung Cancer Screening Trial (TIDL), identifying 23 cases of lung cancer in the first round. The sensitivity of three commercially available AI systems was compared against an experienced radiologist. The AI systems demonstrated varying sensitivity levels: 82.6%, 91.3%, and 78.3% for AI software #1, #2, and #3, respectively. The radiologist had a sensitivity of 95.7%. AI software #2 successfully detected a cancer missed by the radiologist, while AI software #1 and #3 did not. Only one cancer was missed by all AI systems.<br /><br />The study found that AI systems assist in reliably detecting lung cancer in LDCT screenings but lack the sensitivity to completely replace radiologists in autonomously reading CT scans. Some lesions, particularly those between or within emphysematous bullae, larger than 3 cm, or semisolid, were consistently missed by the AI systems, indicating the need for further development and training to improve AI sensitivity.<br /><br />Ultimately, while AI can support radiologists in detecting lung cancer, its current capability limits it to a supplementary role. The study calls for more in-depth prospective research to verify AI's performance as a standalone reader, especially in larger clinical settings.

Asset Subtitle

Albin Mahovkic

Meta Tag

Speaker Albin Mahovkic

Topic Screening & Early Detection

Keywords

artificial intelligence

low-dose computed tomography

lung cancer screening

radiologists

sensitivity comparison

AI systems

Norwegian Lung Cancer Screening Trial

CT scans

AI sensitivity

prospective research