Authors :
Presenting Author: Callum Simpson, MSc – Newcastle University
Gerard Hall, PhD – Research Associate, School of Computing, Newcastle University; Peter Taylor, PhD – Reader in Computational Neurology, School of Computing, Newcastle University; Yujiang Wang, PhD – Reader in Computational Neurology, School of Computing, Newcastle University
Rationale:
Drug-resistant epilepsy often requires surgery that aims to remove the epileptogenic zone. Delineation of the resected tissue can be challenging due to post-operative brain shift. Therefore, most studies use manual approaches which are prohibitively time-consuming for large sample sizes. Furthermore, manual delineation can lead to variations between raters. This study aims to develop an automated pipeline to generate resection masks while preserving manual accuracy.
Methods:
We propose an automated pipeline to generate a 3D mask of the subsequently resected tissue. Our pipeline leverages existing software including FastSurfer (Henschel et al., 2020) and ANTs (Avants et al., 2009) to generate a mask in the same space as the patient’s pre-operative T1 weighted MRI. We compare our automated masks with manually drawn masks for validation. Finally, our pipeline produces an interactive report to allow for easy inspection of the resection mask and its anatomical makeup.
Results:
The automated and hand-drawn segmentations from 53 TLE subjects were compared using the Dice similarity coefficient (DSC) and achieved a median score of 0.78. An example patient with a subsequent anterior temporal lobe resection is shown in Figure 1. Manual and automated masks are also presented. These masks showed good agreement (DSC = 0.88).
Conclusions:
We present a user friendly, easy to use pipeline to automatically generate 3D resection masks and an interactive anatomical report (Figure 2). We will release our scripts openly for future use as a significant tool for epilepsy surgery research.
References
Avants, B.B., Tustison, N. and Song, G., 2009. Advanced normalization tools (ANTS). Insight j, 2(365), pp.1-35.
Henschel, L., Conjeti, S., Estrada, S., Diers, K., Fischl, B. and Reuter, M., 2020. Fastsurfer-a fast and accurate deep learning based neuroimaging pipeline. NeuroImage, 219, p.117012.
Funding:
Y.W. is supported by a UKRI Future Leaders Fellowship (MR/V026569/1). P.N.T. is supported by a UKRI Future Leaders Fellowship (MR/T04294X/1). C.M.S research is funded by Red Hat, and is supported by the Centre for Doctoral Training in Cloud Computing for Big Data (EP/L015358/1).