AI-Based Lung Cancer Screening: Molly Li, MBBS, Shares Insights From the LC-SHIELD Study

Molly Li, MBBS, of the Chinese University of Hong Kong, joined Lung Cancers Today at the 2025 American Society of Clinical Oncology (ASCO) Annual Meeting to discuss the LC-SHIELD lung cancer screening study.

The prospective study, which is being conducted in Hong Kong, is designed to evaluate artificial intelligence (AI)–based lung cancer screening in people who have never smoked but are at a high risk for lung cancer. Dr. Li explained that this is particularly important in East Asia, where there is a high prevalence of lung cancer in people who have never smoked, with nearly half of lung cancer cases being detected in those with a light- or never-smoking history.

“Lung cancer in nonsmokers is a big health problem in East Asia,” Dr. Li said. “Therefore, we think that it’s important to create more lung cancer screening tools, guidelines, and evidence in this population.”

The LC-SHIELD study enrolled people aged 50 to 75 years old who had a lifetime exposure to fewer than 100 cigarettes and had at least one first-degree relative diagnosed with lung cancer. The study participants underwent low-dose CT screening of the thorax, with the scans then analyzed with LungSIGHT, an AI-assisted software program for lung nodule detection. Dr. Li explained that this AI-assisted software is helping to address a significant staffing gap.

“In East Asia, we have a severe shortage of radiologists that has significantly hampered the implementation of CT screening,” he said.

In the LC-SHIELD study, the AI-assisted software acts as the first reader of the CT scans, reducing the need for manual first reads of the CT scans by radiologists. Lung nodules with a maximum diameter of at least 5 mm are classified as “positive” by the AI system.

“If the AI is positive for lung nodules, we will send it to lung radiologists for reporting and arrange follow-up according to the risks,” Dr. Li said.

Furthermore, as a “gold standard,” all scans are retrospectively reviewed by radiologists who are blinded to the LungSIGHT results. Dr. Li explained that in the first 9 months of the study, 400 participants were enrolled, with promising results so far.

“AI is able to screen out around 60% of cases without lung nodules, and we have identified a number of lung cancers as well,” he said.

In addition, the researchers conducted a retrospective validation of the AI program to evaluate its accuracy and sensitivity.

“We found a high sensitivity and negative predictive value of over 99% with the use of the AI program,” Dr. Li said. “We think that AI is a promising and also a very cost-effective tool.”

Dr. Li concluded by reflecting on the practical implications of the study and how it can help address barriers to lung cancer screening in this population.

“That will help us reduce significantly the demand for radiologists that will help us implement the use of CT screening in our locality where radiologist shortage and cost is a big issue in running such programs,” Dr. Li said.