Artificial Intelligence System Reduces Missed Lung Cancer Diagnoses by 60%

Artificial Intelligence System Reduces Missed Lung Cancer Diagnoses by 60%

Artificial intelligence solutions have been shown to improve the accuracy of cancer diagnoses, with a new study suggesting artificial intelligence in combination with the expert eyes of trained radiologists can reduce the rate of missed lung cancer diagnoses by 60%.

CT scans can greatly improve cancer detection rates in the early stages when treatments are most likely to be effective, but while CT screening is increasing, in the early stages of the disease lung cancer is most commonly detected by radiologists from chest X-rays. Even trained radiologists can sometimes miss the early stages of lung cancer, as the interpretation of x-ray images is often a complex and subjective task.

According to researchers, around 90% of lung cancer misdiagnoses occur with chest x-rays. This is an area where artificial intelligence systems can be really beneficial. To test the effectiveness of AI for assisting in cancer diagnosis, researchers at the Department of Radiology at Southend University Hospital in Essex used a commercially available artificial intelligence algorithm to read chest x-rays.

The AI algorithm was found to have a similar detection rate to trained radiologists when used on its own; however, when combined with human analyses of x-rays, AI significantly enhanced detection rates.

For the study, a dataset of 400 chest x-ray images was obtained, with the sample including 200 lung cancer cases which were difficult to diagnose. The researchers used an AI algorithm developed by Behold.ai and the same images were provided to three expert radiologists. The AI and radiologist labels were then combined to simulate the triage workflow.

When used on its own, the AI algorithm had a tumor detection rate of 87% which is equivalent to the detection rate by an average radiologist. When rads and AI were combined, the performance of the radiologists was much better, with the detection rate increasing to 90.67%, with 91.33% sensitivity and 90% specificity.

“The overall reduction in missed cancers of 60% has great promise in improving patient survival rates through the early identification of lung cancers,” wrote the researchers.

You can read more about the study in the paper – Augmenting lung cancer diagnosis on chest radiographs: positioning artificial intelligence to improve radiologist performance – which was recently published in the Clinical Radiology. DOI: 10.1016/j.crad.2021.03.021