Abstract

This study aims to develop and validate an artificial intelligence model based on deep learning to predict early hematoma enlargement (HE) in patients with intracerebral hemorrhage. A total of 1,899 noncontrast computed tomography (NCCT) images of cerebral hemorrhage patients were retrospectively analyzed to establish a predicting model and 1,117 to validate the model. And a total of 118 patients with intracerebral hemorrhage were selected based on inclusion and exclusion criteria so as to validate the value of the model for clinical prediction. The baseline noncontrast computed tomography images within 6 h of intracerebral hemorrhage onset and the second noncontrast computed tomography performed at 24 ± 3 h from the onset were used to evaluate the prediction of intracerebral hemorrhage growth. In validation dataset 1, the AUC was 0.778 (95% CI, 0.768–0.786), the sensitivity was 0.818 (95% CI, 0.790–0.843), and the specificity was 0.601 (95% CI, 0.565–0.632). In validation dataset 2, the AUC was 0.780 (95% CI, 0.761–0.798), the sensitivity was 0.732 (95% CI, 0.682–0.788), and the specificity was 0.709 (95% CI, 0.658–0.759). The sensitivity of intracerebral hemorrhage hematoma expansion as predicted by an artificial intelligence imaging system was 89.3%, with a specificity of 77.8%, a positive predictive value of 55.6%, a negative predictive value of 95.9%, and a Yoden index of 0.671, which were much higher than those based on the manually labeled noncontrast computed tomography signs. Compared with the existing prediction methods through computed tomographic angiography (CTA) image features and noncontrast computed tomography image features analysis, the artificial intelligence model has higher specificity and sensitivity in the prediction of early hematoma enlargement in patients with intracerebral hemorrhage.

Highlights

  • MATERIALS AND METHODSIntracerebral hemorrhage refers to hemorrhage caused by the rupture of the blood vessels in the brain parenchyma, with the mortality rate as high as 40% and 54% after 1 month and 1 year of the rupture, respectively (Zia et al, 2009; Hansen et al, 2013; Poon et al, 2014)

  • The DLS model produces a segmentation mask indicating the location of hematoma and a confidence score representing the risk of hematoma enlargement (HE)

  • In validation dataset 1 (n = 615, 15,980 images), the slice-level pixel-wise IoU was 0.863, and the patient-level IoU was 0.831

Read more

Summary

MATERIALS AND METHODS

Intracerebral hemorrhage refers to hemorrhage caused by the rupture of the blood vessels in the brain parenchyma, with the mortality rate as high as 40% and 54% after 1 month and 1 year of the rupture, respectively (Zia et al, 2009; Hansen et al, 2013; Poon et al, 2014). The recent years have witnessed a surge in research on the prediction of hematoma enlargement with NCCT imaging markers, with the concepts of black hole sign, blend sign, CT hypodensities, island sign, and hematoma enlargement border proposed and validated for their clinical value in the prediction of hematoma enlargement(Lu et al, 2007; Ji et al, 2009; Boulouis et al, 2016; Li et al, 2017; Sporns et al, 2017; Yu et al, 2017) These markers have to be interpreted by trained doctors only because it is easy to be influenced by the readers’ experience and subjective judgment, and the related sensitivity is not high. We attempted to develop a prediction model of hematoma enlargement risk in patients with intracerebral hemorrhage based on the NCCT images so as to rapidly screen out the high-risk population of hematoma enlargement for preparing individualized clinical treatment and guidelines. The NCCT images of patients with spontaneous intracerebral hemorrhage from 84 public hospitals were collected for retrospective analysis between November 2011 and May 2018. P < 0.05 was considered to be statistically significant

RESULTS
DISCUSSION
DATA AVAILABILITY STATEMENT
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call