Automatic heart segmentation using machine learning techniques

Principal Investigator

Name: Julian Bernard
Degrees: M.Sc
Institution: LARALAB UG
Position Title: Research manager
Email: julian.bernard@laralab.de

Project Information

Study: NLST
Project ID: NLST-477
Title: Automatic heart segmentation using machine learning techniques

Summary

Radiotherapy for lymphoma, breast and lung cancer induces cardiac side effects, which often manifest years after treatment and compromise therapy outcomes. It has been shown that cardiac exposure is a negative prognostic factor for patients with lung cancer after concurrent chemotherapy and radiotherapy. A precise segmentation of the heart from CT images allows to calculate cardiac radiation exposure and create comprehensive therapy plans. In this project we will develop machine learning algorithms to automatically segment the heart including cardiac substructures and peripheral vascular structures. The segmentation models shall be applied to cardiac computed tomography images as well as non-contrast CT images. A subset of data will be used to train the neural network, remaining datasets will be used to establish ground truth data with clinical experts.

Aims

1. Development of an automated heart segmentation algorithm for cardiac computed tomography
2. Development of an automated heart segmentation algorithm for non-contrast CT

Collaborators

All Laralab R&D members.