A fast and fully-automated deep-learning approach for accurate hemorrhage segmentation and volume quantification in non-contrast whole-head CT

Abstract

This project aimed to develop and evaluate a fast and fully-automated deep-learning method applying convolutional neural networks with deep supervision (CNN-DS) for accurate hematoma segmentation and volume quantification in computed tomography (CT) scans. Non-contrast whole-head CT scans of 55 patients with hemorrhagic stroke were used. Individual scans were standardized to 64 axial slices of 128 × 128 voxels. Each voxel was annotated independently by experienced raters, generating a binary label of hematoma versus normal brain tissue based on majority voting. The dataset was split randomly into training (n = 45) and testing (n = 10) subsets. A CNN-DS model was built applying the training data and examined using the testing data. Performance of the CNN-DS solution was compared with three previously established methods. The CNN-DS achieved a Dice coefficient score of 0.84 ± 0.06 and recall of 0.83 ± 0.07, higher than patch-wise U-Net (< 0.76). CNN-DS average running time of 0.74 ± 0.07 s was faster than PItcHPERFeCT (> 1412 s) and slice-based U-Net (> 12 s). Comparable interrater agreement rates were observed between “method-human” vs. “human–human” (Cohen’s kappa coefficients > 0.82). The fully automated CNN-DS approach demonstrated expert-level accuracy in fast segmentation and quantification of hematoma, substantially improving over previous methods. Further research is warranted to test the CNN-DS solution as a software tool in clinical settings for effective stroke management.

Editorial: Machine Learning in Neuroimaging

“…The above-mentioned results published in this Research Topic confirm the great potential of machine learning methods to improve various aspects of neuroimaging (1). Applications of artificial methods to diagnostic neuroradiology are really only emerging and have a bright future (2), in particular to assist the neuroradiologists in tedious tasks such as the segmentation of multiple sclerosis (3) or stroke (4) lesions, or to help in the emergency room setting, for example in detecting hemorrhage on non-contrast CT (5).…”

A deep-learning method using computed tomography scout images for estimating patient body weight

“…Recently, deep learning has shown many encouraging results in the field of medical image processing and analysis. Particularly, a convolutional neural network (CNN) has been shown to be well suited for object detection 11 , 12 , semantic segmentation 13 , 14 , image classification 15 , 16 , and prediction 17 , 18 tasks in radiological research. This study aimed to develop a CNN-based method using chest and abdominal CT scout images for predicting body weight and to evaluate the correlation between actual and predicted body weights.…”

Automated Detection and Screening of Traumatic Brain Injury (TBI) Using Computed Tomography Images: A Comprehensive Review and Future Perspectives

“…Chang et al [ 97 ] reported a hybrid 3D/2D mask ROI-based CNN framework with efficient hematoma detection, classification, and segmentation capabilities in parallel. Arab et al [ 98 ] presented a deep learning model with deep supervision for quick and automated segmentation of whole-head CT. Desai et al [ 99 ] presented a deep learning model using pre-trained Google Net to detect the presence of basal ganglia hematoma in a dataset consisting of 170 CT images. Hssayeni et al [ 100 ] proposed a fully automated U-Net model for segmentation of hematoma regions from 82 CT scans.…”

“…Chang et al [ 97 ] developed a novel hybrid ROI-based CNN to estimate the 3D volumes of IPH, SDH, and EDH, respectively. Arab et al [ 98 ] presented a CNN model with deep supervision (CNN–DS) to perform hematoma quantification on whole-head CT rapidly and more efficiently. Jain et al [ 114 ] developed a U-Net based CNN model to compute the volume of acute hematoma lesions, and validated their technique using a multi-centre dataset.…”