Role of Complex Networks for Integrating Medical Images and Radiomic Features of Intracranial Ependymoma Patients in Response to Proton Radiotherapy

Abstract

Human cancers exhibit phenotypic diversity that medical imaging can precisely and non-invasively detect. Multiple factors underlying innovations and progresses in the medical imaging field exert diagnostic and therapeutic impacts. The emerging field of radiomics has shown unprecedented ability to use imaging information in guiding clinical decisions. To achieve clinical assessment that exploits radiomic knowledge sources, integration between diverse data types is required. A current gap is the accuracy with which radiomics aligns with clinical endpoints. We propose a novel methodological approach that synergizes data volumes (images), tissue-contextualized information breadth, and network-driven resolution depth. Following the Precision Medicine paradigm, disease monitoring and prognostic assessment are tackled at the individual level by examining medical images acquired from two patients affected by intracranial ependymoma (with and without relapse). The challenge of spatially characterizing intratumor heterogeneity is tackled by a network approach that presents two main advantages: (a) Increased detection in the image domain power from high spatial resolution, (b) Superior accuracy in generating hypotheses underlying clinical decisions.

Reproducibility of radiomic features using network analysis and its application in Wasserstein k-means clustering

“…Recently, a study was performed in a network analysis framework, using imaging features extracted from magnetic resonance images in intracranial ependymoma, and showed that subnetworks clearly separated tumor and healthy tissues, and even reflected tissue heterogeneity inside the tumor. 29 However, to our knowledge, the present study is the first to exploit network analysis for identifying reproducible radiomic features.…”

Radiomics at a Glance: A Few Lessons Learned from Learning Approaches

“…At another end, new methods will emerge to provide better representations for the encoded inputs via concepts like networks deconvolution, inversion, and dissection, among others (see [ 95 ]). Fifth and last, in order to face the challenge of intratumor heterogeneity, the quantification of tumor abundance at the voxel level is becoming an important direction in response assessment and recurrence risk studies [ 4 , 95 , 96 , 97 ]. This might help the identification of subregions, for instance, those metabolically active and defined as high risk [ 92 ], and may also inspire strategy to mitigate the effects of unbalanced data (for instance, when an outcome is over-represented) and thus decisional bias [ 12 ].…”

Translating Data Science Results into Precision Oncology Decisions: A Mini Review

“…Fourth, the integration of high dimensional data, including imaging, genetic sequence, clinical data, and health records, introduces errors, approximations, and inaccuracies. Some solutions have been proposed, such as a complex network approach [ 70 ]. Finally, it is difficult to compare model performance among different studies for the lack of standards in algorithm implementation and data collection [ 71 ].…”