Abstract #3069

Grading of glioma using a machine learning framework based on optimized features obtained from T1 perfusion MRI and volumes of tumor components

Anirban Sengupta¹, Sumeet Agarwal², Rakesh Kumar Gupta³, Dinil Sasi⁴, Ayan Debnath^4,5, and Anup Singh⁴

¹Centre for Biomedical Engineering, IIT Delhi, New Delhi, India, ²Electrical Engineering, IIT Delhi, Hauz Khas, New Delhi, India, ³Fortis Memorial Research Institute, New Delhi, India, ⁴IIT Delhi, New Delhi, India, ⁵University of Pennysylvania, Philadelphia, PA, United States

Grading of glioma based on T₁ perfusion MRI parameters is well reported but it has certain challenges specially in differentiating intermediate glioma grades (Grade II vs. III and Grade III vs. IV). In this study, we have differentiated intermediate as well as multiclass glioma grades (Grade II vs. III vs. IV) using an optimized machine learning framework which uses quantitative T₁ perfusion MRI parameters in combination with volume of different components of tumor as a feature set. The results show that it is feasible to obtain low error in glioma grading using the proposed methodology. The results also emphasizes the utility of using volume of tumor subparts in conjunction with T₁ perfusion MRI parameters for glioma grading.

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

Join Here