Evaluating the Real World Performance of an AI Based Lung Nodule Detection Tool

Performance Estimation of Triaging Artificial Intelligence Based Computer-Aided Detection Algorithm in Routine Chest Radiography

chest x-rays will be analyzed by AI software for a secondary read of lung nodules. Chest x-rays will either be sent to the AI tool to be read or to radiologists to read. If the image is sent to the AI tool, the AI software will generate a report on if it detects a lung nodule or not. The image will then be sent to a radiologist to determine if there is agreement or disagreement with the AI tool.

The study is a prospective study for measuring the performance of an AI software in detecting lung nodules from chest X-rays. Data collected during the study will be analyzed for study purposes after end date of data collection.

There will be two study arms: the control arm and the interventional arm.

Control Arm:

There will be no interruption to the existing standard of care pathway.

Interventional Arm:

Use of AI will occur in parallel to the standard of care pathway.

Consistent with the control Arm, the radiologists or clinicians interpreting the chest x-ray images will proceed as usual based on the existing standard operating procedures of the study site. In addition, the AI software will function as a second reader; meaning images will be processed by the AI software which will generate a report.

In the event that the radiologist and the AI tool do not agree, cases will be reviewed by qualified study team members twice per week.

Inclusion Criteria:

• Chest X-ray images of patients aged 18 - 89 years.
• Modality: CR/DR/DX.
• PA/view
• Lung nodules measuring 6 mm -30 mm (for chest X-ray images where presence of nodules is required).

Exclusion Criteria:

• Incomplete view of the chest.
• Lateral view
• Known lung cancer at the time of Chest x-ray images.