Texture Analysis of Standard Magnetic Resonance Images to Predict Response to Gamma Knife Radiosurgery in Vestibular Schwannomas

• Published in: World neurosurgery 2019-12, Vol.132, p.e228-e234
• Main Authors: Speckter, Herwin, Santana, Jairo, Bido, José, Hernandez, Giancarlo, Rivera, Diones, Suazo, Luis, Valenzuela, Santiago, Oviedo, Jairo, Gonzalez, Cesar F., Stoeter, Peter
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Aged; Child; Disease Progression; Female; Humans; Image Interpretation, Computer-Assisted - methods; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Male; Middle Aged; Neuroimaging - methods; Neuroma, Acoustic - diagnostic imaging; Neuroma, Acoustic - pathology; Neuroma, Acoustic - surgery; Radiosurgery; Texture analysis; Vestibular schwannoma; Young Adult
• ISSN: 1878-8750; 1878-8769
• DOI: 10.1016/j.wneu.2019.08.193

To search for texture features of routine magnetic resonance imaging to predict tumor volume reduction and transient versus permanent tumor progression of vestibular schwannomas treated by Gamma Knife stereotactic radiosurgery. Included were 23 patients with vestibular schwannomas treated in our center and followed over a period of 23.7–80.3 months (mean 42.7). Magnetic resonance imaging was performed on a 3-Tesla scanner and included T1-weighted images with and without contrast enhancement, T2-weighted, and fluid-attenuated inversion recovery images. Volumetric results were followed longitudinally over time and correlated to texture features as mean, minimum, maximum, standard deviation, skewness, and kurtosis of normalized signals taken from regions of interest covering the total tumor volume. In total, 14 tumors showed early progression during the first 5–18 months (2 cases permanent, 12 cases transient), whereas 9 tumors regressed immediately after SRS. Kurtosis of T2-weighted image intensity values turned out to predict progression best with a sensitivity and specificity of 71% and 78%. From all texture feature parameters, only the minimum of the normalized T2-weighted image intensity values correlated significantly to the final reduction of tumor volume per month (correlation coefficient = −0.634, P < 0.05, corrected for false discovery rate). Texture feature analysis helps to predict permanent versus transient enlargement and final volume reduction of schwannomas after SRS. Thus, alternative treatment strategies might be considered, mainly in large tumors, where further clinical deterioration cannot be excluded. To confirm these results, a prospective study including more cases and a longer follow-up period is necessary.