Artificial Intelligence for Detecting Pulmonary Embolisms via CT: A Workflow-oriented Implementation.

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Pulmonary embolism (PE) detection is critical for patient care.
• Artificial intelligence (AI) shows promise in assisting radiologists with PE detection.
• Evaluating AI diagnostic accuracy and clinical staff acceptance is crucial for workflow integration.

Purpose of the Study:

• To evaluate the performance of an AI algorithm in detecting PE on CTPA.
• To assess the acceptance of the AI algorithm among radiology department staff.

Main Methods:

• Retrospective analysis of anonymized CTPA data from a university clinic.
• Surveys conducted at 3 and 9 months post-AI implementation.
• Reference standard provided by thoracic and cardiac radiologists.
• AI analyzed 59 cases initially and 46 in follow-up.

Main Results:

• AI sensitivity improved from 84.6% to 90.9%; specificity from 94.3% to 96.7%.
• Radiologists' acceptance of the AI tool increased over time.
• Despite increased acceptance, some preferred hiring additional physicians over AI if costs were comparable.

Conclusions:

• AI algorithm demonstrated high, improving sensitivity and specificity for PE detection on CTPA.
• Growing user confidence and trust in the AI tool were observed.
• AI shows promise as a supportive 'second-look' tool in emergency radiology, complementing physician expertise.