Abstract #3072

Comparing supervised and unsupervised machine learning frameworks based upon quantitative-MRI features in differentiation between non-enhancing tumor and vasogenic edema of glioma patients and validation using histopathological ground-truth

Neha Vats¹, Anirban Sengupta², Dinil Sasi³, Rakesh Kumar Gupta⁴, R.P. Chauhan¹, Virendra Kumar Yadav³, Sumeet Agarwal⁵, and Anup Singh³

¹NIT Kurukshetra, Kurukshetra, India, ²Centre for Biomedical Engineering, IIT Delhi, New Delhi, India, ³IIT Delhi, New Delhi, India, ⁴Fortis Memorial Research Institute, New Delhi, India, ⁵Electrical Engineering, IIT Delhi, Hauz Khas, New Delhi, India

The aim of this study was to compare the efficacy of unsupervised machine learning technique in differentiating non-enhancing tumor(NET) from surrounding vasogenic edema (VE) in high-grade glioma patients using T1-perfusion MRI parameters. Two unsupervised machine learning techniques, k-means clustering and Gaussian mixture model (GMM) were optimized with respect to their hyper-parameters for differentiating NET from VE and the results were compared with previously published results obtained using a supervised classifier Support Vector Machine (SVM). The results showed that SVM classifier was slightly superior to GMM and K-means clustering in differentiating NET from VE.

This abstract and the presentation materials are available to members only; a login is required.

Join Here