noamko128 / image-processing-project

Image Segmentation Project- using CT scans in DICOM format, as part of a 'Biomedical Image Processing' course

Image segmentation project

Data backround-

The data set we worked on is based on data from patients with pulmonary embolism. It is taken from the KAGGLE website (https://www.kaggle.com/datasets/andrewmvd/pulmonary-embolism-in-ct-images) and contains CT scans in DICOM format which is a standard for saving medical images. The data set consists of angiography scans of 35 different patients, where each patient has ~200 scans.

Method-

The final goal of the project was to present a clean three-dimensional scan of the rib cage. We tried to work with several different methods to reach the desired result and among them: performing a simple segmentation and using the Canny algorithm. After realizing that segmentation on a single image is not the way to reach the desired result, we switched to working with a segmentation method on volume scans. We chose the Connected Components algorithm that divides the largest areas that are connected to each other into separate components. After reaching a better result, we also made corrections to the scans using morphology.

Project conclusions-

• An experiment and questioning must be carried out separately for each patient in order to find the optimal number of Connected Components for him.
• Our model will not work optimally on every patient and every scan due to the wide variety of scans.

*This project is my final project in the Biomedical Image Processing course as part of my Bachelor's degree in Digital Medical Technologies.