Deep Learning Convolutional Neural Networks for the Estimation of Liver Fibrosis Severity from Ultrasound Texture - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2019 Feb:10950:109503E.
doi: 10.1117/12.2512592. Epub 2019 Mar 13.

Deep Learning Convolutional Neural Networks for the Estimation of Liver Fibrosis Severity from Ultrasound Texture

Alex Treacher  1 , Daniel Beauchamp  1 , Bilal Quadri  1 , David Fetzer  1 , Abhinav Vij  1 , Takeshi Yokoo  1 , Albert Montillo  1
Affiliations

Deep Learning Convolutional Neural Networks for the Estimation of Liver Fibrosis Severity from Ultrasound Texture

Alex Treacher et al. Proc SPIE Int Soc Opt Eng. 2019 Feb.

Abstract

Diagnosis and staging of liver fibrosis is a vital prognostic marker in chronic liver diseases. Due to the inaccuracies and risk of complications associated with liver core needle biopsy, the current standard for diagnosis, other less invasive methods are sought for diagnosis. One such method that has been shown to correlate well with liver fibrosis is shear wave velocity measured by ultrasound (US) shear wave elastography; however, this technique requires specific software, hardware, and training. A current perspective in the radiology community is that the texture pattern from an US image may be predictive of the stage of liver fibrosis. We propose the use of convolutional neural networks (CNNs), a framework shown to be well suited for real world image interpretation, to test whether the texture pattern in gray scale elastography images (B-mode US with fixed, subject-agnostic acquisition settings) is predictive of the shear wave velocity (SWV). In this study, gray scale elastography images from over 300 patients including 3,500 images with corresponding SWV measurements were preprocessed and used as input to 100 different CNN architectures that were trained to regress shear wave velocity. In this study, even the best performing CNN explained only negligible variation in the shear wave velocity measures. These extensive test results suggest that the gray scale elastography image texture provides little predictive information about shear wave velocity and liver fibrosis.

Keywords: Convolutional Neural Network; Deep Learning; Liver Fibrosis; Random Search; Shear Wave Velocity; Ultrasound.

PubMed Disclaimer

Figures

Figure 1:
Figure 1:
An gray scale elastography image (left) and the image region used for CNN prediction (red box on right). The white box in both panels shows the ROI positioned by the technologist as the target for SWE to measure the SWV. The image region for prediction (right) has dimensions that are 5x the width (w) and 1.66x the height (h) of the elastography targeting white box.
Figure 2:
Figure 2:
The minimum median MSE for the validation data on each network across epochs, ranked from lowest to highest. The first architecture (Table 1C) indicated with the red arrow was used later to evaluate performance on the on the test data.
Figure 3:
Figure 3:
Gray scale elastography ROIs for 10 high and 10 low subjects. Shown on the left are 10 ROIs from subjects with no or little fibrosis and who have the lowest SWV. Shown on the right are 10 ROIs with the highest SWV and high fibrosis. These images were used to quantify human expert accuracy.
Figure 4:
Figure 4:
A. Comparison of predicted SWV (m/s) versus actual elastography measured SWV (m/s). The ideal predicted=measured line is shown in black, while the actual linear fit line, shown in red, demonstrates only weak correlation between the measured and predicted values. B. The Bland-Altman plot showing average of the measured and predicted SWV values versus the difference: measured - predicted. The red lines show a 95% confidence interval which are +/− 0.848 m/s from the mean. C. The range of SWV for fibrosis stages: low fibrosis (green), significant fibrosis (yellow), and advanced fibrosis (red) as described in a pSWE study. The blue line shows the 95% confidence interval of the best performing model can span all three stages.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Centers for Disease Control and Prevention. Chronic Liver Disease and Cirrhosis. Available at https://www.cdc.gov/nchs/fastats/liver-disease.htm (2016).
    1. Barr RG et al. Elastography Assessment of Liver Fibrosis: Society of Radiologists in Ultrasound Consensus Conference Statement. Radiology 276, 845–861 (2015). - PubMed
    1. Seeff LB et al. Complication rate of percutaneous liver biopsies among persons with advanced chronic liver disease in the HALT-C trial. Clinical gastroenterology and hepatology 8, 877–883 (2010). - PMC - PubMed
    1. Regev A et al. Sampling error and intraobserver variation in liver biopsy in patients with chronic HCV infection. The American journal of gastroenterology 97, 2614–2618 (2002). - PubMed
    1. Vicas C, Lupsor M, Socaciu M, Badea R & Nedevschi Sergiu. Liver Fibrosis detection by the means of texture analysis. Limitations and further development directions. Automat. Comput. Appl. Math 19, 397–402 (2010).

LinkOut - more resources