A Clinical Guideline Driven Automated Linear Feature Extraction for Vestibular Schwannoma

TL;DR

The paper addresses the need for standardized vestibular schwannoma measurements by automating the extraction of intrameatal/extrameatal and maximum tumour diameters using a deep learning segmentation pipeline. It introduces a two-stage nnU-Net segmentation framework for pre- and post-operative scans and a rule-based feature extraction module to select the most clinically relevant diameter, $\\mathcal{D}$, as either $\\mathcal{D}^{WT}$ or $\\mathcal{D}^{EM}$. Validation on 50 patients (187 sessions) shows strong agreement with manual measurements (e.g., $r$ approaching 0.97, favorable Bland–Altman biases), supporting its potential as a clinical decision aid in MDT meetings. The approach standardizes VS management, offering visual and quantitative guidance while highlighting areas for improvement, such as segmentation overestimation and modality biases, with avenues to extend to other benign brain tumours.

Abstract

Vestibular Schwannoma is a benign brain tumour that grows from one of the balance nerves. Patients may be treated by surgery, radiosurgery or with a conservative "wait-and-scan" strategy. Clinicians typically use manually extracted linear measurements to aid clinical decision making. This work aims to automate and improve this process by using deep learning based segmentation to extract relevant clinical features through computational algorithms. To the best of our knowledge, our study is the first to propose an automated approach to replicate local clinical guidelines. Our deep learning based segmentation provided Dice-scores of 0.8124 +- 0.2343 and 0.8969 +- 0.0521 for extrameatal and whole tumour regions respectively for T2 weighted MRI, whereas 0.8222 +- 0.2108 and 0.9049 +- 0.0646 were obtained for T1 weighted MRI. We propose a novel algorithm to choose and extract the most appropriate maximum linear measurement from the segmented regions based on the size of the extrameatal portion of the tumour. Using this tool, clinicians will be provided with a visual guide and related metrics relating to tumour progression that will function as a clinical decision aid. In this study, we utilize 187 scans obtained from 50 patients referred to a tertiary specialist neurosurgical service in the United Kingdom. The measurements extracted manually by an expert neuroradiologist indicated a significant correlation with the automated measurements (p < 0.0001).

Figures (4)

• Figure 1: A: intra-/extra-meatal regions are shown in green and yellow labels respectively. B: Petrous pyramid region is shown in red dashed margin and the boundary separating intra-/extra-meatal regions is shown in blue. C: Maximum extrameatal diameter. D: Maximum whole tumour diameter.
• Figure 2: The proposed framework: Segmentation module consists of a two-stage approach for pre-operative scans, whereas the finetuned stage 1 model is used for the post-operative scans. This is integrated with the feature extraction module to provide clinical features for 50 patients.
• Figure 3: Rows 1 & 2 are different cases from the KCLH MC-RC dataset. Columns A, B, C show the additional features measured when deciding the presented feature in the column D. The green line in column A is the $\textbf{d}_{(intra,\parallel)}$. The blue line in column B is the $\textbf{d}_{(extra,\parallel)}$. The yellow line in column C is the $\textbf{d}_{(extra,\bot)}$. The posterior petrous pyramid, or the boundary between the intra-/extra-meatal region is shown in white dotted line. In the column D, the maximum extrameatal diameter is shown in red and the maximum whole tumour diameter is shown in orange.
• Figure 4: Plots with the assessment of agreement between two methods of clinical measurement; automated vs manual. A & D: represent the relationship between maximum extrameatal diameter feature. B & E: represent the relationship between maximum whole tumour diameter feature for pre-operative sessions. C & F: represent the relationship between maximum whole tumour diameter feature for post-operative sessions. In Bland-Altman plots, red dashed lines indicate the bias, (mean of the difference) and black dashed lines indicate upper and lower 95% limits of agreement. AM: Automated Measurement, EMM: Expert Manual Measurement