Project: Exploration of the complexity levels of radiomic characteristics

Status

In progress (2021-today)

Type

Master’s

Team

• Mahdi Ait Lhaj Loutfi¹ (2021-today)
• Martin Lepage² (2021-today)
• Martin Vallières¹ (2021-today)

¹ Computer science department, Université de Sherbrooke, Sherbrooke (QC), Canada

² Nuclear medicine and radiobiology department, Université de Sherbrooke, Sherbrooke (QC), Canada

Objective

The project consists of a systematic evaluation of the potential of prediction for medical image-based analysis of increasing complexity in the precision oncology spectrum.

Issue and objective

Recent advances in “omics” technology have created opportunities to characterize biological processes correlated with tumor phenotypes. We anticipate that integrative modeling of available oncology data will provide the ability to make accurate tumor phenotype predictions, which in turn would allow for better monitoring and tailoring of patient-specific treatments (i.e. “precision oncology”). To potentially improve radiomic analysis (i.e.: Quantitative extraction of useful data from all types of medical images), radiomic features of the tumor region can be extracted from medical images filtered with different types of filters (e.g.: texture filtering or texture smoothing). These features can be extracted using the MEDimage package, but they expose different levels of complexity, which leads us to investigate in this project, the spectrum of complexity of medical images for the resolving of several problems in precision oncology and study the changes in these different levels of complexity from an image modality to another.

Database

We already have a large database on which the analysis of the levels of complexity will be carried out. It contains multiple combinations of image modalities and prediction problems (13 combinations in all), as shown in the table below:

Levels of complexity to be investigated