Automated Ventricular System Segmentation in Paediatric Patients Treated for Hydrocephalus Using Deep Learning Methods

Abstract

Hydrocephalus is a common neurological condition that can have traumatic ramifications and can be lethal without treatment. Nowadays, during therapy radiologists have to spend a vast amount of time assessing the volume of cerebrospinal fluid (CSF) by manual segmentation on Computed Tomography (CT) images. Further, some of the segmentations are prone to radiologist bias and high intraobserver variability. To improve this, researchers are exploring methods to automate the process, which would enable faster and more unbiased results. In this study, we propose the application of U-Net convolutional neural network in order to automatically segment CT brain scans for location of CSF. U-Net is a neural network that has proven to be successful for various interdisciplinary segmentation tasks. We optimised training using state of the art methods, including “1cycle” learning rate policy, transfer learning, generalized dice loss function, mixed float precision, self-attention, and data augmentation. Even though the study was performed using a limited amount of data (80 CT images), our experiment has shown near human-level performance. We managed to achieve a 0.917 mean dice score with 0.0352 standard deviation on cross validation across the training data and a 0.9506 mean dice score on a separate test set. To our knowledge, these results are better than any known method for CSF segmentation in hydrocephalic patients, and thus, it is promising for potential practical applications.

Hydrocephalus: Ventricular Volume Quantification Using Three-Dimensional Brain CT Data and Semiautomatic Three-Dimensional Threshold-Based Segmentation Approach

“…However, the inclusion of cerebral cisterns adjacent to the ventricles, such as the basal cistern, was unavoidable, and this resulted in an overestimation of ventricular volume ( 3 ). A few recent studies demonstrated that automated intracranial ventricle segmentation software using deep learning methods could reduce processing time (e.g., to less than 1 minute) with high accuracy ( 18 19 ). However, these automatic tools produced segmentation errors and remained at a pre-clinical stage of research ( 18 19 ).…”

“…A few recent studies demonstrated that automated intracranial ventricle segmentation software using deep learning methods could reduce processing time (e.g., to less than 1 minute) with high accuracy ( 18 19 ). However, these automatic tools produced segmentation errors and remained at a pre-clinical stage of research ( 18 19 ). The accuracy of deep learning methods depends on the quality and the number of ground truth data used for training.…”

Semantic segmentation of cerebrospinal fluid and brain volume with a convolutional neural network in pediatric hydrocephalus—transfer learning from existing algorithms

“…Here, CSF without brain volume was considered. In this study, a Dice coefficient CSF of 95% could be achieved [18]. Overall, our segmentation results of the CNN on the BrainWeb data set were very accurate; however, the Dice coefficient cut off was marginally inferior due to slightly higher false negative values in the tissue area, particularly in the lower layers.…”

“…Here, CSF without brain volume was considered. In this study, a Dice coefficient CSF of 95% could be achieved [ 18 ].…”

Hydrocephalus after aneurysmal subarachnoid hemorrhage: Epidemiology, Pathogenesis, Diagnosis, and Management

“…The assessment of these segments is often highly subjected to intraobserver variability [123]. To improve this, Klimont et al proposed the use of a U-Net convolutional neural network for automatic segmentation of brain CT scans [124].…”