Multimodal "Radiomics" data analysis and visualization

Introduction:

The individualization of treatment regimens is increasingly finding its way into everyday clinical practice, especially in oncology. Multimodal databases are available for individual patients to identify outcome and treatment-relevant complex endotypes and multiparametric biomarker signatures. The radiological metadata-based "radiomics" signatures-which can be correlated with clinical, histological, and genetic data-are indicative of this. One of the biggest challenges is the processing and visualization of multimodal, large amounts of data.

Method:

We created a database of clinical, histological and radiographic metadata from 100 patients with head and neck squamous cell carcinoma and a follow-up of 2 years. For the evaluation of radiological metadata from pretherapeutic CT images, the primary tumor was manually and the lymph node metastases were semi-automatically segmented with software developed by the Fraunhofer Institute for this application. Radiomics features were created from the resulting metadata by appropriate algorithms.

Results:

From the available data individual radiomics characteristics could be created and put into clinical as well as histopathological context. The data allowed the identification of individual patient subgroups as well as a correlation and user-friendly visualization of the characteristics of the different data pools.

Conclusion:

An establishment of radiomics characteristics as well as their correlation could be presented and visualized in a novel way. The visualized analysis allows a more intuitive recognition of individual subgroups and data correlations, which facilitates the identification of a Radiomics signature.

Publication History

Publication Date:
18 April 2018 (online)