Automated Segmentation and Connectivity Analysis for Normal Pressure Hydrocephalus

Abstract

Objective and Impact Statement. We propose an automated method of predicting Normal Pressure Hydrocephalus (NPH) from CT scans. A deep convolutional network segments regions of interest from the scans. These regions are then combined with MRI information to predict NPH. To our knowledge, this is the first method which automatically predicts NPH from CT scans and incorporates diffusion tractography information for prediction. Introduction. Due to their low cost and high versatility, CT scans are often used in NPH diagnosis. No well-defined and effective protocol currently exists for analysis of CT scans for NPH. Evans’ index, an approximation of the ventricle to brain volume using one 2D image slice, has been proposed but is not robust. The proposed approach is an effective way to quantify regions of interest and offers a computational method for predicting NPH. Methods. We propose a novel method to predict NPH by combining regions of interest segmented from CT scans with connectome data to compute features which capture the impact of enlarged ventricles by excluding fiber tracts passing through these regions. The segmentation and network features are used to train a model for NPH prediction. Results. Our method outperforms the current state-of-the-art by 9 precision points and 29 recall points. Our segmentation model outperforms the current state-of-the-art in segmenting the ventricle, gray-white matter, and subarachnoid space in CT scans. Conclusion. Our experimental results demonstrate that fast and accurate volumetric segmentation of CT brain scans can help improve the NPH diagnosis process, and network properties can increase NPH prediction accuracy.

The Diagnostic Accuracy of Artificial Intelligence in Radiological Markers of Normal-Pressure Hydrocephalus (NPH) on Non-Contrast CT Scans of the Brain

“…Zhang et al propose an automated method of predicting NPH using the volumetric segmentation of CT brain scans, which is the first method that automatically predicts NPH from CT scans using AI. The connectome data to compute features, which capture the impact of enlarged ventricles and regions of interest segmented from CT scans using AI, provide the fast and accurate volumetric segmentation of CT brain scans, which can thus improve the NPH diagnosis accuracy [32]. However, their approach relies on a 3D U-Net model for segmentation, which is more computationally expensive compared to our study.…”

Structural neuroimaging markers of normal pressure hydrocephalus versus Alzheimer’s dementia and Parkinson’s disease, and hydrocephalus versus atrophy in chronic TBI—a narrative review

“…Considering an EI-x ≥ 0.32 to be an indicator of hydrocephalus, a transfer learning scheme was applied on a large dataset of CT scans to show a classification performance with AUC of 0.93, sensitivity of 93.6%, and specificity of 94.4% in distinguishing hydrocephalus from normal controls ( 113 ). Automated segmentation of CSF, subarachnoid, and cerebral spaces on non-contrast CT scans of 27 patients with possible NPH, integrated with indirectly inferred connectome data, was shown to be as effective as the EI-x for prediction with a specificity of 85% and sensitivity of 86% ( 114 ). Haber et al ( 115 ) recently developed a convolutional neural network (CNN) which was able to classify patients with definite improvement post-surgery as identified by the 2nd edition of the Japanese guidelines from healthy controls using CT scans setting a promising precedent in the application of deep learning to CT scans in NPH assessment.…”

Semisupervised 3D segmentation of pancreatic tumors in positron emission tomography/computed tomography images using a mutual information minimization and cross-fusion strategy

“…This delineation was then validated and confirmed by another radiologist with more than 10 years of experience in pancreatic disease diagnosis. Importantly, both radiologists were kept blind to the patients’ clinical outcomes ( 56 , 57 ). The pancreatic cancer datasets in this study were all confirmed by histopathological or cytological examination.…”