Prediction of Vestibular Schwannoma Enlargement After Radiosurgery Using Tumor Shape and MRI Texture Features

Abstract

Objective: Determine if vestibular schwannoma (VS) shape and MRI texture features predict significant enlargement after stereotactic radiosurgery (SRS). Study Design: Retrospective case review. Setting: Tertiary referral center. Patients: Fifty-three patients were selected who underwent SRS and had a contrast-enhanced T1 sequence planning MRI scan and a follow-up contrast enhanced T1 MRI available for review. Median follow-up of 6.5 months (interquartile range/IQR, 5.9–7.4). Median pretreatment tumor volume was 1,006 mm3 (IQR, 465–1,794). Intervention(s): Stereotactic radiosurgery. Main Outcome Measure(s): Texture and shape features from the SRS planning scans were extracted and used to train a linear support vector machine binary classifier to predict post-SRS enlargement >20% of the pretreatment volume. Sensitivity, specificity, area under the receiver-operating-characteristic curve (AUC), and positive likelihood ratio were computed. A stratified analysis based on pretreatment tumor volume greater or less than the median volume was also performed. Results: The model had a sensitivity of 92%, specificity of 65%, AUC of 0.75, and a positive likelihood ratio of 2.6 (95% CI 1.4–5.0) for predicting post-SRS enlargement of >20%. In the larger tumor subgroup, the model had a sensitivity of 87%, specificity of 73%, AUC of 0.76, and a positive likelihood ratio of 3.2 (95% CI 1.2–8.5). In the smaller tumor subgroup, the model had a sensitivity of 95%, specificity of 50%, AUC of 0.65, and a positive likelihood ratio of 1.9 (95% CI 0.8–4.3). Conclusions: VS shape and texture features may be useful inputs for machine learning models that predict VS enlargement after SRS.

Joint Vestibular Schwannoma Enlargement Prediction and Segmentation Using a Deep Multi‐task Model

“…Recently, radiomics‐based methods combined with traditional machine learning techniques have been applied to different prediction tasks for VS 14,15,31–33 . Although promising prediction performance was achieved for post‐radiosurgery response/enlargement prediction and active treatment requirement prediction, none of these methods provides end‐to‐end prediction, because tumor delineation is a prerequisite for MRI radiomics feature extraction.…”

“…Different from tumor enlargement detection which is based on tumor size measurement on follow‐up images, tumor enlargement prediction (TEP) methods predict the probability of significant tumor volume growth in the early stage of VS diagnosis. Recent works have shown that VS MRI radiomic feature analysis with machine learning methods is predictive for TEP, which has the potential to be used as a basis for individualized treatment planning 14–16 . However, these studies focus on TEP for post‐radiosurgery VS only, and 3D delineation of VS is needed as a prerequisite for radiomic feature extraction.…”

Emerging artificial intelligence applications in otological imaging

“…The recent literature consists of AI applied to otological imaging in a range of modalities and clinical contexts (Table 1). Groups have combined AI with computed tomography (CT) [4–14], magnetic resonance imaging (MRI) [15–17,18 ▪ ,19,20] and light otoscopy/otoendoscopy [21 ▪ ,22,23,24 ▪▪ ,25,26,27 ▪ ,28–31,32 ▪▪ ]. Most of the recent literature focuses on two themes: automated, image-based otoscopic diagnosis and automated segmentation of temporal bone CT and MRI scans for use in virtual reality, surgical simulation and surgical planning.…”

“…George-Jones trained a support vector machine (SVM) algorithm to use tumour shape and texture information from MRI scans of vestibular schwannomas (VS) to predict a greater than 20% volumetric growth 6 months post radiosurgery with a sensitivity of 92% and specificity of 65% [15]. Langhuizen's SVM model examined a longer follow up period of 129 months post VS radiosurgery and demonstrated 83% sensitivity and 82% specificity [17].…”

MRI Texture Features are Associated with Vestibular Schwannoma Histology

“…In the context of VS, previous work has shown that a cystic or heterogenous appearance on MRI was associated with an increase in Antoni type B/mixed tissue 7 . In addition, recent studies have established relationships between MRI texture parameters and the prediction of VS enlargement after radiosurgery 8–10 . Although these results are promising and exciting, the aforementioned texture analysis studies in VS all take place in the context of tumors that did not undergo surgical resection, and therefore, there has been no correlation between these texture parameters and underling tumor histology.…”