top of page

Optimizing Cardiovascular Health: Leveraging Deep Learning for Enhanced Real-Time Detection of Coronary Artery Stenosis

Overview:
Developed and implemented advanced deep learning models to detect and classify coronary artery stenosis in angiographic images, providing a robust solution for automating cardiovascular diagnostics. This project combined data analytics, machine learning, and image processing to address a critical challenge in medical imaging.

​

Key Responsibilities:

  1. Data Analysis and Preparation:

    • Collected and analyzed a dataset of 8,325 angiographic images sourced from multiple imaging systems.

    • Preprocessed data using techniques like normalization and augmentation to enhance model performance.

    • Annotated medical images with bounding boxes using LabelBox and ensured accurate classification of stenosis severity.

  2. Machine Learning Model Development:

    • Trained, fine-tuned, and evaluated object detection models, including Faster R-CNN, DETR, and YOLOv5, using Python and PyTorch.

    • Optimized model hyperparameters for improved precision, recall, and real-time processing speed.

    • Conducted a comparative analysis of model performance based on metrics such as mean Average Precision (mAP), F1 Score, and ROC AUC.

  3. Data Workflow and Visualization:

    • Designed data workflows and created insightful visualizations using Python, Excel, and Power BI, aligning with data accuracy and reporting requirements.

    • Performed exploratory data analysis (EDA) to identify patterns, trends, and correlations within the imaging dataset.

    • Presented findings through compelling visualizations to aid decision-making and stakeholder communication.

  4. Documentation and Reporting:

    • Documented methodologies, workflows, and findings in an 18,000-word dissertation, showcasing advanced data communication, technical writing, and project management skills.

  5. Innovation in Real-Time Applications:

    • Demonstrated YOLOv5 as the best-fit model for real-time clinical environments, balancing speed and accuracy.

    • Developed a framework for future integration of AI models into clinical workflows to aid decision-making.

 

Results and Impact:

  • Achieved high diagnostic accuracy and reduced false positives and negatives, paving the way for early intervention and better patient outcomes.

  • Provided a scalable, real-time solution to support healthcare professionals in diagnosing coronary artery stenosis.

  • Contributed to advancements in AI-driven healthcare innovations by exploring generalizability across diverse patient populations.

 

Skills and Technologies:

  • Programming & Tools: Python, PyTorch, TensorFlow, Excel, Power BI, LabelBox, Roboflow

  • Data Analysis: Exploratory Data Analysis (EDA), data preprocessing, and statistical analysis

  • Machine Learning & Deep Learning: Faster R-CNN, DETR (Detection Transformer), YOLOv5

  • Evaluation Metrics: mAP, F1 Score, ROC AUC, Precision-Recall Curve

  • Visualization: Python libraries, Power BI, Excel for data-driven insights

 

Relevance for Data Roles:
This project showcases expertise in designing data workflows, visualizing complex datasets, applying machine learning and deep learning techniques, and documenting findings—core skills for data analyst, data scientist, and machine learning roles.

bottom of page