Background
Diffusion tensor cardiac magnetic resonance (DT‐CMR) is an emerging technique that provides information on myocardial microstructure. When applied in-vivo, the technique requires significant amount of post-processing of the images before actual information can be retrieved. In a paper lead by Dr Pedro Ferreira, from Dr Sonia Nielles Vellespin’s team, they have proven the suitability of deep learning–based segmentation for automating the process, resulting in a faster and less prone to error analysis. Dr Ferreira developed a proof-of-concept MATLAB package that implements such deep learning approach, in addition to the manual post processing, and that had been used already to publish ~20 papers, most notably this from Dr Nielles-Vallespin.
Our Contribution
This MATLAB package worked as a proof-of-concept, however there were a number of limitations that the RSE Team was able to help with by re-implementing it in Python. The CADI tool was a Graphical User Interface (GUI) developed using the RSE team's guikit - a tool for building modular GUIs with wxPython. It allows users to process the imaging data through various modules, that conduct cropping, segmentation, rotation, registration and building a diffusion tensor. It was developed alongside an RSE embedded in the research group allowing us to sensibly divide work tasks so that the domain-specific code was conducted by those with the domain knowledge.
The Research team has continued to to develop the code so that it can be used for processing in-vivo as well as ex-vivo data.