Automated Detection of Vestibular Schwannoma Growth Using a Two‐Dimensional U‐Net Convolutional Neural Network

Abstract

Objectives/Hypothesis To determine if an automated vestibular schwannoma (VS) segmentation model has comparable performance to using the greatest linear dimension to detect growth. Study Design Case‐control Study. Methods Patients were selected from an internal database who had an initial gadolinium‐enhanced T1‐weighted magnetic resonance imaging scan and a follow‐up scan captured at least 5 months later. Two observers manually segmented the VS to compute volumes, and one observer's segmentations were used to train a convolutional neural network model to automatically segment the VS and determine the volume. The results of automatic segmentation were compared to the observer whose measurements were not used in model development to measure agreement. We then examined the sensitivity, specificity, and area under the receiver‐operating characteristic curve (AUC) to compare automated volumetric growth detection versus using the greatest linear dimension. Growth detection determined by the external observer's measurements served as the gold standard. Results A total of 65 patients and 130 scans were studied. The automated method of segmentation demonstrated excellent agreement with the observer whose measurements were not used for model development for the initial scan (interclass correlational coefficient [ICC] = 0.995; 95% confidence interval [CI]: 0.991‐0.997) and follow‐up scan (ICC = 0.960; 95% CI: 0.935–0.975). The automated method of segmentation demonstrated increased sensitivity (72.2% vs. 63.9%), specificity (79.3% vs. 69.0%), and AUC (0.822 vs. 0.701) compared to using the greatest linear dimension for growth detection. Conclusions In detecting VS growth, a convolutional neural network model outperformed using the greatest linear dimension, demonstrating a potential application of artificial intelligence methods to VS surveillance. Level of Evidence 4 Laryngoscope, 131:E619–E624, 2021

Machine Learning for the Detection and Segmentation of Benign Tumors of the Central Nervous System: A Systematic Review

“…Similarly, only four manuscripts reported interrater variability. The most frequently used metric for interrater variability, the dice coefficient, was between 0.89 and 0.94 [ 5 , 112 , 114 , 115 ].…”

“…George-Jones et al analyzed a cohort of 65 patients with a median tumor volume of only 0.28 mL [ 114 ]. Unlike other publications, the authors did not report dice coefficients, but instead tried to analyze how well the model was able to detect growth compared to the manual segmentations which were used as the ground truth.…”

Segmentation of Vestibular Schwannomas on Postoperative Gadolinium-Enhanced T1-Weighted and Noncontrast T2-Weighted Magnetic Resonance Imaging Using Deep Learning

“…George-Jones et al (13) demonstrated the use of an automated segmentation model to detect growth in vestibular schwannomas on GdT1WI. A direct comparison in segmentation performance with our model cannot be made because George-Jones et al did not report performance metrics such as the Dice score for the model itself, instead focusing on the ability of the model to detect tumor growth in comparison to manual segmentation using the greatest linear dimension.…”

Convolutional Neural Networks for Classifying Laterality of Vestibular Schwannomas on Single MRI Slices—A Feasibility Study

“…One publication used an NN to predict vs. recurrence following surgery from clinical parameters in tabular format [23]. The remaining four publications used NNs to segment VS for either radiotherapy planning or response assessment [24][25][26][27].…”