Texture analysis of B-mode ultrasound images to stage hepatic lipidosis in the dairy cow: A methodological study

- A methodology to estimate the degree of fatty infiltration of the liver in the dairy cow is proposed.
- Texture analysis of ultrasonographic images of the liver was performed.
- Hepatic biopsies were performed to quantitatively measure liver fat content.
- A regression equation to predict hepatic fat content from texture parameters was generated.
- The methodology proposed has high diagnostic performance in the detection of hepatic lipidosis.

Hepatic lipidosis is the most diffused hepatic disease in the lactating cow. A new methodology to estimate the degree of fatty infiltration of the liver in lactating cows by means of texture analysis of B-mode ultrasound images is proposed. B-mode ultrasonography of the liver was performed in 48 Holstein Friesian cows using standardized ultrasound parameters. Liver biopsies to determine the triacylglycerol content of the liver (TAGqa) were obtained from each animal. A large number of texture parameters were calculated on the ultrasound images by means of a free software. Based on the TAGqa content of the liver, 29 samples were classified as mild (TAGqa < 50 mg/g), 6 as moderate (50 mg/g < TAGqa > 100 mg/g) and 13 as severe (TAG > 100 mg/g) in steatosis. Stepwise linear regression analysis was performed to predict the TAGqa content of the liver (TAGpred) from the texture parameters calculated on the ultrasound images. A five-variable model was used to predict the TAG content from the ultrasound images. The regression model explained 83.4% of the variance. An area under the curve (AUC) of 0.949 was calculated for < 50 mg/g vs > 50 mg/g of TAGqa; using an optimal cut-off value of 72 mg/g TAGpred had a sensitivity of 86.2% and a specificity of 84.2%. An AUC of 0.978 for < 100 mg/g vs > 100 mg/g of TAGqa was calculated; using an optimal cut-off value of 89 mg/g, TAGpred sensitivity was 92.3% and specificity was 88.6%. Texture analysis of B-mode ultrasound images may therefore be used to accurately predict the TAG content of the liver in lactating cows.