Segmentation of glioma structures is vital for therapy planning.Although state of the art algorithms achieve impressive resultswhen compared to ground-truth manual delineations, one could arguethat the binary nature of these labels does not properly reflect the underlyingbiology, nor does it account for uncertainties in the predictedsegmentations. Moreover, the tumor infiltration beyond the contrastenhancedlesion – visually imperceptible on imaging – is often ignoreddespite its potential role in tumor recurrence. We propose an intensitybasedprobabilistic model for brain tissue mapping based on conventionalMRI sequences. We evaluated its value in the binary segmentation of thetumor and its subregions, and in the visualisation of possible infiltration.The model achieves a median Dice of 0.82 in the detection of thewhole tumor, but suffers from confusion between different subregions.Preliminary results for the tumor probability maps encourage furtherinvestigation of the model regarding infiltration detection.
De Sutter, S, Geens, W, Bossa Bossa, MN, Vanbinst, A-M, Duerinck, J & Vandemeulebroucke, J 2022, Probabilistic tissue mapping for tumor segmentation and infiltration detection of glioma. in Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries. Springer.
De Sutter, S., Geens, W., Bossa Bossa, M. N., Vanbinst, A.-M., Duerinck, J., & Vandemeulebroucke, J. (Accepted/In press). Probabilistic tissue mapping for tumor segmentation and infiltration detection of glioma. In Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries Springer.
@inproceedings{6ea62581bdaf4ff08d615e7c8bffceef,
title = "Probabilistic tissue mapping for tumor segmentation and infiltration detection of glioma",
abstract = "Segmentation of glioma structures is vital for therapy planning.Although state of the art algorithms achieve impressive resultswhen compared to ground-truth manual delineations, one could arguethat the binary nature of these labels does not properly reflect the underlyingbiology, nor does it account for uncertainties in the predictedsegmentations. Moreover, the tumor infiltration beyond the contrastenhancedlesion – visually imperceptible on imaging – is often ignoreddespite its potential role in tumor recurrence. We propose an intensitybasedprobabilistic model for brain tissue mapping based on conventionalMRI sequences. We evaluated its value in the binary segmentation of thetumor and its subregions, and in the visualisation of possible infiltration.The model achieves a median Dice of 0.82 in the detection of thewhole tumor, but suffers from confusion between different subregions.Preliminary results for the tumor probability maps encourage furtherinvestigation of the model regarding infiltration detection.",
author = "{De Sutter}, Selene and Wietse Geens and {Bossa Bossa}, {Mat{\'i}as Nicol{\'a}s} and Anne-Marie Vanbinst and Johnny Duerinck and Jef Vandemeulebroucke",
year = "2022",
language = "English",
booktitle = "Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries",
publisher = "Springer",
}