1194 A Practical Guide to Manual and Semi-Automated Neurosurgical Brain Lesion Segmentation

Aim To provide neurosurgeons with a practical guide for manual and semi-automated image segmentation which they can utilise for neurosurgical brain lesions, such as meningioma, glioblastoma and subarachnoid haemorrhage. Method Medical images of patients were downloaded from the International Brain Tumour Segmentation Challenges image database and from local patient data on Picture Archiving and Communication Systems with consent. Collected images were pre-processed by the MRIcro software. ITK-SNAP was used to manually segment the meningioma scan while classification via semi-automated segmentation was utilised for the glioblastoma and subarachnoid haemorrhage scans. Videos were recorded to illustrate the segmentation process with key teaching points. Quality control was achieved by utilising expert segmenters to review the segmentations. Results We devised an effective 4-stage pipeline consisting of image acquisition, pre-processing, segmentation, and post-segmentation quality control for novice neurosurgical segmenters to perform effective manual and semi-automated segmentation using ITK-SNAP. Videos have been generated to give detailed practical instructions with regard to segmenting subarachnoid haemorrhage, meningioma, and glioblastoma. Detailed segmentation figures have been created and included in the paper to illustrate the end result of following the video segmentation methods. We have also created a table which demonstrates key features of these three pathologies which would be of practical help for any novice segmenter. Conclusions Neurosurgical image segmentation has significant potential for education, research, and clinical practice. Our educational guideline provides a step-by-step pathway for neurosurgeons new to medical image segmentation to allow them to apply these methods within their research and clinical practice.