A multitask deep learning approach for pulmonary embolism detection and identification.

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Pulmonary embolism (PE) is a life-threatening condition requiring prompt diagnosis and treatment.
• Chest computed tomographic pulmonary angiogram (CTPA) is the standard for PE detection.
• Deep learning offers potential to enhance radiologist workflow and diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate a two-phase multitask deep learning model for PE detection and characterization.
• To improve the accuracy and efficiency of diagnosing pulmonary embolism using CTPA.
• To reduce false-negative diagnoses by identifying PE presence and properties.

Main Methods:

• A two-phase multitask deep learning approach was implemented.
• The model was trained on the RSNA-STR Pulmonary Embolism CT Dataset.
• Interpretability was achieved using attention weight heatmaps and Grad-CAM.

Main Results:

• The model achieved an AUROC of 0.93, sensitivity of 0.86, and specificity of 0.85 on the test set.
• Performance is competitive with human radiologists.
• The model accurately identified PE presence, location, chronicity, and RV/LV ratio.

Conclusions:

• The proposed deep learning model shows significant promise for assisting in the diagnosis of pulmonary embolism.
• This AI tool can help prioritize critical cases and expedite diagnosis.
• The model's interpretability features support clinical application in PE diagnosis.