Author: John Kildea, Odette Rios-Ibacache, Amal Zouaq 👨🔬
Affiliation: McGill University, Polytechnique Montréal 🌍
Purpose:
Even though Electronic medical records (EHRs) are now in widespread use in healthcare, and Artificial Intelligence tools incorporating radiomics are used to identify tumors in medical images and automate the assessment of outcomes in radiotherapy (RT) treatments, current lack of standardization in the interoperability of health data limits the potential of real-world research. This study proposes an ontological extension for the Minimal Common Oncology Data Elements (mCODE) STU4, a set of structured data elements for cancer data interoperability, linking patients' medical image information, radiomics and dosiomics with their EHR data in a standardized way, to facilitate multicenter research studies.
Methods:
We analyzed the structured set mCODE as an ontology and designed an extension for medical images, focusing on images related to cancer diagnosis and RT treatment. A review of the existing literature on the harmonization and standardization of radiomics and dosiomics extraction methods was conducted to include the minimum parameters that would impact the extraction of radiomics and dosiomics. We included data elements recommended by the Image Biomarker Standardisation Initiative (IBSI) guidelines.
Results:
Currently, our proposed extension has 20 key elements, divided into classes and subclasses, and 99 attributes. The radiomics portion of the extension (for diagnostic and radiotherapy images) has 14 classes, 54 attributes and 17 relationships between the classes. The dosiomics portion has 6 classes, 16 attributes, and 8 relationships. Our next step involves developing a radiomics feature-extractor module to extract features from images and information used in our ontology to enrich Knowledge Bases (KBs) and facilitate multicenter and multi-omics studies.
Conclusion:
This study is one of the first steps in our development of an mCODE-compliant KB for cancer patients, aiming to standardize medical information used in medical physics research for data storage, increase reproducibility in radiomics and dosiomics studies, and set the path for multicentre studies.