PERIPHERAL ARTERIAL DISEASE DETECTION USING MACHINE LEARNING

Abstract: Peripheral Arterial Disease (PAD) is a serious vascular condition that is often underdiagnosed due to reliance on manual interpretation of Doppler ultrasound reports. Early detection is critical to prevent complications such as critical limb ischemia and amputation. This study presents an automated PAD detection system that integrates Optical Character Recognition (OCR) with machine learning to analyze Color Doppler reports. The system extracts key clinical parameters, including the Ankle-Brachial Index (ABI), using text parsing techniques from reports in PDF and image formats. A hybrid decision framework combining a Random Forest classifier with rule-based clinical logic is employed to determine PAD presence and severity. A clinical override rule ensures that PAD is flagged when ABI ≤ 0.90, improving diagnostic reliability. To address class imbalance, Synthetic Minority Oversampling Technique (SMOTE) is applied, resulting in improved accuracy, precision, recall, and F1-score. The proposed system reduces manual effort, enhances diagnostic consistency, and enables real-time clinical decision support, making it suitable for practical healthcare applications.

Keywords: Peripheral Arterial Disease (PAD), Machine Learning, Optical Character Recognition (OCR), Random Forest, Ankle-Brachial Index (ABI), SMOTE, Healthcare Automation, Doppler Report Analysis.