AI Detection Of Incidental PE: More Common Than Expected

Purpose: An incidental pulmonary embolism (IPE) can be detected on a contrast enhanced chest computerized tomography (CT). This study aimed to evaluate the feasibility of using AI to detect PE on a contrast enhanced chest CT and its clinical impact.

Methods and Materials: This retrospective analysis included all contrast enhanced chest CT’s (performed as chest only CT or CT including the chest) at a single tertiary center over eight months (9/2020-4/2021), a period that included the COVID-19 pandemic. IPE was considered positive in a chest CT’s performed for indications other than detection of PE and following an expert radiologist conformation. All cases were evaluated by an AI algorithm (AIDOC). Cases were further in-depth analyzed for IPE features (central vs. subsegmental, multifocal vs. unifocal) and patient management.

Results: A total of 10408 CT examinations were included with 300 (2.9%) PE’s identified. The CTPA group comprised a total of 1775 cases with 193 (10.8%) positive cases while the non CTPA group comprised a total of 8633 cases with 107 (1.2%) positive cases identified. IPE comprised 35% out of total PE cases. Most of the IPE were unilocular (57%) with lobar and segmental being the most common location (52%). All IPE patients were treated following diagnosis anti-coagulation.

Conclusions: A significant percentage (35%) of all detected PE cases are found on contrast enhanced CT of the chest. AI provides a reliable technique for detection of these incidental PE’s. Following AI based diagnosis the management of these clinically underdiagnosed patients included anticoagulant drugs, indicating that a higher index of suspicion is required.

Clinical Relevance/Application: Usage of an AI product to detect IPE could potentially be implemented to increase patient safety.