Project description

Research and development project.

This project focuses on classifying coronary artery calcium (CAC) in low-dose chest CT images using deep learning. It integrates both CT images and Electronic Medical Records (EMR) data to improve classification accuracy. The study aims to automate CAC scoring, which is crucial for early detection of coronary artery disease, by developing a deep learning-based classification model. .

Role

Designed and implemented a deep learning-based classification model using multimodal data (CT + EMR).

Extracted patches from CT images and applied pre-trained encoders (VGG16, ResNet-50, DenseNet-201).

Compared results with existing CAC classification models and reported accuracy and F1-score improvements.

Challenge:

Merging CT image features with structured EMR data required advanced fusion techniques also traditional augmentation techniques impacted accuracy; thus, preprocessing methods like histogram equalization and thresholding were used.

Skills

Design Development Python TensorFlow 2.4.2, Keras VGG16, ResNet-50, DenseNet-201 OpenCV, DICOM image handling NumPy, Pandas, Scikit-learn Quadro RTX 8000 Accuracy, F1-score Adam optimizer with a learning rate of 0.0001