Planar single plane area determination is a viable substitute for total volumetry of CSF and brain in childhood hydrocephalus

Exploiting lamina terminalis appearance and motion in prediction of hydrocephalus using convolutional LSTM network

BACKGROUND

Evaluation of the lamina terminalis (LT) is crucial for non-invasive evaluation of the CSF diversion for the treatment of hydrocephalus. Together with deep learning algorithms, morphological and physiological analyses of the LT may play an important role in the management of hydrocephalus.

AIM

We aim to show that exploiting the motion of LT can contribute to the evaluation of hydrocephalus using deep learning algorithms.

METHODS

The dataset contains 61 True-fisp data with routine sequences 37 of which are labeled as 'hydrocephalus' and the others as 'normal condition'. A fifteen-year experienced neuroradiologist divided data into two groups. The first group, 'hydrocephalus', consists of patients with typical MRI findings (ventriculomegaly, enlargement of the third ventricular recesses and lateral ventricular horns, decreased mamillo-pontine distance, reduced frontal horn angle, thinning/elevation of the corpus callosum, and non-dilated convexity sulci), and the second group contains samples that did not show any symptoms or neurologic abnormality and labeled as 'normal condition'. The region of interest was determined by the radiologist supervisor to cover the LT. To achieve our purpose, we used both spatial and spatio-temporal analysis with two different deep learning architectures. We utilized Convolutional Neural Networks (CNN) for spatial and Convolutional Long Short-Term Memory (ConvLSTM) models for spatio-temporal analysis using an ROI around LT on sagittal True-fisp images.

RESULTS

Our results show that 80.7% classification accuracy was achieved with the ConvLSTM model exploiting LT motion, whereas 76.5% and 71.6% accuracies were obtained by the 2D CNN model using all frames, and only the first frame from only spatial information, respectively.

CONCLUSION

We suggest that the motion of the LT can be used as an additional attribute to the spatial information to evaluate the hydrocephalus.

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

Background

For the segmentation of medical imaging data, a multitude of precise but very specific algorithms exist. In previous studies, we investigated the possibility of segmenting MRI data to determine cerebrospinal fluid and brain volume using a classical machine learning algorithm. It demonstrated good clinical usability and a very accurate correlation of the volumes to the single area determination in a reproducible axial layer. This study aims to investigate whether these established segmentation algorithms can be transferred to new, more generalizable deep learning algorithms employing an extended transfer learning procedure and whether medically meaningful segmentation is possible.

Methods

Ninety-five routinely performed true FISP MRI sequences were retrospectively analyzed in 43 patients with pediatric hydrocephalus. Using a freely available and clinically established segmentation algorithm based on a hidden Markov random field model, four classes of segmentation (brain, cerebrospinal fluid (CSF), background, and tissue) were generated. Fifty-nine randomly selected data sets (10,432 slices) were used as a training data set. Images were augmented for contrast, brightness, and random left/right and X/Y translation. A convolutional neural network (CNN) for semantic image segmentation composed of an encoder and corresponding decoder subnetwork was set up. The network was pre-initialized with layers and weights from a pre-trained VGG 16 model. Following the network was trained with the labeled image data set. A validation data set of 18 scans (3289 slices) was used to monitor the performance as the deep CNN trained. The classification results were tested on 18 randomly allocated labeled data sets (3319 slices) and on a T2-weighted BrainWeb data set with known ground truth.

Results

The segmentation of clinical test data provided reliable results (global accuracy 0.90, Dice coefficient 0.86), while the CNN segmentation of data from the BrainWeb data set showed comparable results (global accuracy 0.89, Dice coefficient 0.84). The segmentation of the BrainWeb data set with the classical FAST algorithm produced consistent findings (global accuracy 0.90, Dice coefficient 0.87). Likewise, the area development of brain and CSF in the long-term clinical course of three patients was presented.

Conclusion

Using the presented methods, we showed that conventional segmentation algorithms can be transferred to new advances in deep learning with comparable accuracy, generating a large number of training data sets with relatively little effort. A clinically meaningful segmentation possibility was demonstrated.