Data Mining for Biomarker Discovery

Volume 65 of the series Springer Optimization and Its Applications pp 91-116


A Framework for Multimodal Imaging Biomarker Extraction with Application to Brain MRI

  • Kostas MariasAffiliated withInstitute of Computer Science, FORTH Email author 
  • , Vangelis SakkalisAffiliated withInstitute of Computer Science, FORTH
  • , Norbert GrafAffiliated withDepartment of Paediatric Oncology, USAAR

* Final gross prices may vary according to local VAT.

Get Access


The crucial role of imaging biomarkers is sparsely mentioned in the literature due to the complex nature of medical images, the interpretation variability and the multidisciplinary approach needed to extract, validate, and translate such biomarkers to the clinical setting. In the case of cancer, imaging biomarkers can play an important role in understanding the stage of the disease as well as the response (or not) to initial treatment as early as possible. In neurodegenerative diseases, imaging biomarkers can assist the early detection and diagnosis, before substantial symptoms appear. In this chapter, we describe the clinical importance of establishing robust imaging biomarkers as well as the limitations that need to be addressed. Then, we propose a clinically driven/ assisted image-analysis-based framework for extracting and assessing temporal image biomarkers comprising of geometrical normalization and image-information extraction. The proposed biomarker image discovery framework including a number of clinically useful tools developed by our group has been integrated in a platform called ‘DoctorEye’, a novel, open access and easy to use clinical multimodal image analysis environment. Based on this clinical platform, we describe three examples of imaging biomarker discovery involving our recent work for the case of brain MRI.