IoMT-Enabled Computer-Aided Diagnosis of Pulmonary Embolism from Computed Tomography Scans Using Deep Learning

Area of Science:

• Medical Imaging and Artificial Intelligence
• Deep Learning for Healthcare
• Computer-Aided Diagnosis Systems

Background:

• The Internet of Medical Things (IoMT) generates vast medical data, necessitating advanced analytical models.
• Pulmonary Embolism (PE) is a leading cause of death, often diagnosed via Computed Tomography (CT) scans.
• Manual analysis of CT scans for PE is time-consuming and prone to diagnostic errors.

Purpose of the Study:

• To develop an automated Computer-Aided Diagnosis (CAD) system for Pulmonary Embolism (PE) detection.
• To leverage deep learning, specifically DenseNet201, for classifying PE in CT scans.
• To assist radiologists in making accurate and timely PE diagnoses.

Main Methods:

• A deep learning framework utilizing DenseNet201 as a feature extractor was proposed.
• Customized fully connected decision-making layers were integrated into the DenseNet201 architecture.
• The model was trained and evaluated on the RSNA-Pulmonary Embolism Detection Challenge (2020) Kaggle dataset.

Main Results:

• The proposed DenseNet201 model achieved promising diagnostic performance.
• Key performance metrics included 88% accuracy, 88% sensitivity, and 89% specificity.
• The model demonstrated a strong Area Under the Curve (AUC) of 90%.

Conclusions:

• The developed deep learning framework shows significant potential for automated PE detection in CT scans.
• This CAD system can assist medical professionals, potentially reducing misdiagnosis rates.
• Early and accurate PE diagnosis through AI can contribute to lowering mortality rates.