Quantitative ultrasound and machine learning for assessment of steatohepatitis in a rat model
- 185 Downloads
To develop a machine learning model based on quantitative ultrasound (QUS) parameters to improve classification of steatohepatitis with shear wave elastography in rats by using histopathology scoring as the reference standard.
This study received approval from the institutional animal care committee. Sixty male Sprague-Dawley rats were either fed a standard chow or a methionine- and choline-deficient diet. Ultrasound-based radiofrequency images were recorded in vivo to generate QUS and elastography maps. Random forests classification models and a bootstrap method were used to identify the QUS parameters that improved the classification accuracy of elastography. Receiver-operating characteristic analyses were performed.
For classification of not steatohepatitis vs borderline or steatohepatitis, the area under the receiver-operating characteristic curve (AUC) increased from 0.63 for elastography alone to 0.72 for a model that combined elastography and QUS techniques (p < 0.001). For detection of liver steatosis grades 0 vs ≥ 1, ≤ 1 vs ≥ 2, ≤ 2 vs 3, respectively, the AUCs increased from 0.70, 0.65, and 0.69 to 0.78, 0.78, and 0.75 (p < 0.001). For detection of liver inflammation grades 0 vs ≥ 1, ≤ 1 vs ≥ 2, ≤ 2 vs 3, respectively, the AUCs increased from 0.58, 0.77, and 0.78 to 0.66, 0.84, and 0.87 (p < 0.001). For staging of liver fibrosis grades 0 vs ≥ 1, ≤ 1 vs ≥ 2, and ≤ 2 vs ≥ 3, respectively, the AUCs increased from 0.79, 0.92, and 0.91 to 0.85, 0.98, and 0.97 (p < 0.001).
QUS parameters improved the classification accuracy of steatohepatitis, liver steatosis, inflammation, and fibrosis compared to shear wave elastography alone.
• Quantitative ultrasound and shear wave elastography improved classification accuracy of liver steatohepatitis and its histological features (liver steatosis, inflammation, and fibrosis) compared to elastography alone.
• A machine learning approach based on random forest models and incorporating local attenuation and homodyned-K tissue modeling shows promise for classification of nonalcoholic steatohepatitis.
• Further research should be performed to demonstrate the applicability of this multi-parametric QUS approach in a human cohort and to validate the combinations of parameters providing the highest classification accuracy.
KeywordsNonalcoholic steatohepatitis Non-alcoholic fatty liver disease Machine learning Ultrasonography Elasticity imaging techniques
Area under the receiver-operating characteristic curve
Controlled attenuation parameter
Hematoxylin phloxine saffron
Methionine and choline deficient
Nonalcoholic fatty liver disease
Receiver operating characteristic
Region of interest
We thank Mr. Jamal Ait Ichou for his assistance in the literature review.
This study has received funding by Fonds de Recherche du Québec—Nature et Technologies (FRQNT) (PR-174387), Canadian Institutes of Health Research, Institute of Nutrition, Metabolism, and Diabetes (grant nos. 273738 and 301520), Quebec Bio-imaging Network (QBIN/RBIQ #5886), New Researcher Startup Grant from the Centre de Recherche du Centre Hospitalier de l’Université de Montréal, and by a Chercheur-Boursier Junior 2 from the Fonds de Recherche du Québec en Santé and Fondation de l’association des radiologistes du Québec (FRQS-ARQ #34939) to An Tang.
Compliance with ethical standards
The scientific guarantor if this publication is An Tang.
Conflict of interest
The authors declare that they have no conflict of interest.
Statistics and biometry
One of the authors has significant statistical expertise.
Approval from the institutional animal care committee was obtained.
Institutional review board approval was obtained.
Study subjects or cohorts overlap
The study cohorts have been previously reported in . That prior article focused on a comparison between low-frequency versus high-frequency ultrasonographic shear wave elastography for the detection of steatohepatitis. In contrast, the present article investigates a machine learning model based on quantitative ultrasound parameters to improve classification of steatohepatitis compared to shear wave elastography alone.
• Diagnostic study
• Performed at one institution
- 23.Mamou J, Oelze ML (2013) Quantitative ultrasound in soft tissues. Springer, DordrechtGoogle Scholar
- 32.Destrempes F, Meunier J, Giroux MF, Soulez G, Cloutier G (2011) Segmentation of plaques in sequences of ultrasonic B-mode images of carotid arteries based on motion estimation and a Bayesian model. IEEE Trans Biomed Eng 58:2202–2211. https://doi.org/10.1109/TBME.2011.2127476
- 35.Bigelow TA, Labyed Y (2013) Attenuation compensation and estimation. In: Mamou J, Oelze ML (eds) Quantitative ultrasound in soft tissues. Springer, Dordrecht, pp 71–93Google Scholar
- 39.Efron B, Tibshirani R (1997) Improvements on cross-validation: the 632+ bootstrap method. J Am Stat Assoc 92:548–560Google Scholar
- 41.Davis J, Goadrich M (2006) The relationship between precision-recall and ROC curves. Proceedings of the 23rd international conference on machine learning. ACM, pp 233–240Google Scholar
- 46.Dong DR, Hao MN, Li C et al (2015) Acoustic radiation force impulse elastography, FibroScan(®), Forns’ index and their combination in the assessment of liver fibrosis in patients with chronic hepatitis B, and the impact of inflammatory activity and steatosis on these diagnostic methods. Mol Med Rep 11:4174–4182CrossRefGoogle Scholar
- 49.Mohareri O, Ruszkowski A, Lobo J et al (2014) Multi-parametric 3D quantitative ultrasound vibro-elastography imaging for detecting palpable prostate tumors. Med Image Comput Comput Assist Interv 17:561–568Google Scholar
- 50.Roy-Cardinal M, Destrempes F, Soulez G, Cloutier G (2016) Homodyned K-distribution parametric maps combined with elastograms for carotid artery plaque assessment. 2016 IEEE International Ultrasonics Symposium (IUS) pp 1–4Google Scholar