November 18, 2026
Health news, explained with evidence
LIVE

New AI Framework Enhances Non-Alcoholic Fatty Liver Disease Detection via Ultrasound

🔬 Research 2 min read 👁 4

Researchers have developed a new deep learning framework called EPFH-Net to detect non-alcoholic fatty liver disease using ultrasound images with high accuracy.

🔬Research

Key points

  • EPFH-Net combines EfficientNet, PyramidNet, and harmonic analysis to detect NAFLD.
  • The model analyzes B-mode ultrasound images using advanced feature extraction techniques.
  • Experiments demonstrated an accuracy of 92.12% and a true positive rate of 92.52%.

A team of researchers has introduced a novel diagnostic framework known as the Efficient Pyramid Forward Harmonic Network, or EPFH-Net, aimed at improving the detection of non-alcoholic fatty liver disease (NAFLD). Published in Scientific Reports, the study addresses historical limitations associated with traditional liver biopsies, which have often been time-consuming and prone to inaccuracies in identifying fat accumulation within liver cells not linked to excessive alcohol consumption.

The newly proposed computational method works by processing B-mode fatty liver ultrasound images. It utilizes a region of interest extraction process alongside advanced feature extraction techniques, including Convolutional Neural Networks, Gray Level Co-occurrence Matrix, Pyramid Histogram of Oriented Gradients, and Center Symmetric-Local Binary Patterns. By combining an Efficient Network, a Pyramid Network, and harmonic analysis, EPFH-Net classifies the resulting outputs as either normal or fatty.

According to the experimental outcomes detailed in the study, the framework achieved an accuracy rate of 92.12%, a true positive rate of 92.52%, and a true negative rate of 91.88% at a k-value of 9. While these computational results suggest a promising avenue for non-invasive screening, the findings are based on an early-stage technological model rather than large-scale clinical trials.

What to know: EPFH-Net is an experimental diagnostic tool designed to analyze ultrasound data for signs of NAFLD. Readers should consult a qualified healthcare provider for any personal medical evaluations or diagnostic decisions regarding liver health.

Source Nature Health Sciences — nature.com/articles/s41598-026-77862-2 ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.