Liver steatosis classification using high-frequency ultrasound.

High-frequency B-mode images of 19 fresh human liver samples were obtained to evaluate their usefulness in determining the steatosis grade. The images were acquired by a mechanically controlled single-crystal probe at 25 MHz. Image features derived from gray-level concurrence and nonseparable wavelet transform were extracted to classify steatosis grade using a classifier known as the support vector machine. A subsequent histologic examination of each liver sample graded the steatosis from 0 to 3. The four grades were then combined into two, three and four classes. The classification results were correlated with histology. The best classification accuracies of the two, three and four classes were 90.5%, 85.8% and 82.6%, respectively, which were markedly better than those at 7 MHz. These results indicate that liver steatosis can be more accurately characterized using high-frequency B-mode ultrasound. Limitations and their potential solutions of applying high-frequency ultrasound to liver imaging are also discussed.