Cancer accounted for 12 million deaths worldwide in 2020. This statistic implies that 1 in every 6 deaths happened due to cancer. Research shows that early identification and effective treatment can reduce fatalities resulting from certain cancers. Early identification is usually achieved by screening where the body is checked for cancer before symptoms appear. Screening for breast, colorectal, cervical and lung cancer is U.S. Preventive Services Task Force external icon (USPSTF). Screening is done via different tests out of which radiological imaging is quite popular.

In this study, we evaluate how low-dose computed tomography scans, the USPSTF’s current recommended screening method for lung cancer, can be utilized to detect lung cancer using deep learning. False positives from cancer screening occur when a person without cancer is determined to have cancer. Additional tests and treatments are typically administered afterward, which results in inflated expenditures and treatment fatigue. There is frequently a difference in opinion among doctors. Deep Learning has occasionally outperformed clinicians in its ability to predict lung cancer.

Since lung cancer screening is compulsory in high-risk persons in the US, insurance payers are responsible for covering treatments that are neither necessary nor desirable in the event of a false positive. This research attempts to forecast the financial impact of a decrease in false positives per patient, which can ultimately result in cost savings for insurance payers if deep learning treatments were implemented in lung cancer screening.