NIMG-13. SEGMENTATION AND VOLUMETRIC ANALYSIS IMPROVES DETECTION OF PROGRESSION IN LOW GRADE GLIOMAS

• Published in: Neuro-oncology (Charlottesville, Va.) Vol. 20; no. suppl_6; p. vi178
• Main Authors: Fathallah-Shaykh, Hassan, Nabors, L Burt, DeAtkine, Andrew, Bag, Asim, Bredel, Markus, Warren, Paula, Han, Xiaosi, Bouaynaya, Nidhal
• Format: Journal Article
• Language: English
• Published: US Oxford University Press 05-11-2018
• Subjects: Abstracts
• ISSN: 1522-8517; 1523-5866
• DOI: 10.1093/neuonc/noy148.740

Abstract
Tumor surveillance is a common practice in oncology; its goal is timely detection of cancer progression, which is typically diagnosed by Visual Comparison (VC) or bidirectional measurement of longitudinal radiological images. Low-grade gliomas (LGG) can cause significant neurological morbidity by progressive brain invasion over time. We analyzed the longitudinal magnetic resonance imaging (MRI) studies of patients, diagnosed with grade 2 gliomas without any radiation therapy at the initial diagnosis, who had at least 4 MRI studies. Forty-eight patients met the inclusion criteria, including 13 oligodendrogliomas, 24 astrocytomas and 11 mixed gliomas. Thirty-four patients had clinical progression (CP) and 14 patients were considered clinically stable (CS) by VC at the time of the study. The computer assisted diagnosis (CAD) method consisted of segmentation of the fluid-attenuated inversion recovery (FLAIR) images, computing tumor volumes, and determining progression by the online abrupt change of point method, which considered only past MRIs. In the CP group, CAD detected progression at earlier times than the clinical diagnosis in 26/34 patients. In the CS group, CAD detected progression in 10/14 patients. Six physicians from the departments of neurology (neuro-oncology), radiology, and radiation oncology reviewed and agreed with the progression determined by the CAD method including the study subjects in the CS group. CAD was able to diagnose tumor progression when the tumor size grew at an average of 74% (range 14% to 364%) since the base line compared to 320% (range 25% to 2019%) in the VC method. The average time to progression was only 17 months when CAD was used compared to 50 months when VC method was used. Early detection of tumor progression is critical for LGG patients because it can optimize therapeutic options with significantly lower volume of brain resection or irradiation