How AI is Revolutionizing Ultrasound Diagnoses of the Liver

By Campion Quinn, MD

Introduction

Ultrasound is a widely used imaging modality for liver disease because it is fast, cost-effective, and non-invasive. Unlike MRI, which is expensive and often requires long wait times, ultrasound can be performed at the point of care, enabling immediate decision-making. Additionally, ultrasound is particularly valuable in low-resource settings where advanced imaging modalities may not be available. However, traditional ultrasound interpretation depends heavily on operator expertise, leading to variability in results. Artificial intelligence (AI) is transforming liver ultrasound by automating image analysis, improving diagnostic accuracy, and standardizing assessments. This article explores how AI enhances liver ultrasound studies, streamlines clinical workflows, and improves patient outcomes.

Current Challenges in Liver Ultrasound

Liver ultrasound is widely used due to its accessibility, affordability, and non-invasive nature. However, it has several limitations:

1. Operator Dependence: Interpretation varies based on the skill and experience of the sonographer.

2. Limited Sensitivity for Early Disease: Mild fibrosis and fatty liver disease may go undetected until they progress [7].

3. Difficulty in Differentiating Lesions: Distinguishing benign from malignant tumors can be challenging [3].

4. Measurement Inconsistencies: Liver stiffness and steatosis assessments rely on manual techniques, introducing variability [5].

AI-enhanced ultrasound overcomes these issues by standardizing image interpretation, detecting subtle abnormalities, and providing objective measurements.

AI in Liver Fibrosis Staging

Liver fibrosis is a key indicator of chronic liver disease. AI improves fibrosis detection through:

· Deep Learning on B-mode Ultrasound: AI models analyze ultrasound images for texture changes associated with fibrosis. Studies show AI can distinguish significant fibrosis (≥F2) with an area under the curve (AUC) of 0.92, surpassing traditional methods [3].

· AI-Assisted Elastography: Shear wave elastography (SWE) measures liver stiffness, but manual selection of the region of interest (ROI) can introduce errors. AI optimizes ROI selection and filters suboptimal frames, improving diagnostic reliability [7].

· Integration with Biomarkers: AI combining ultrasound with blood test markers (e.g., FIB-4) reduces unnecessary referrals by 42% [5].

By increasing fibrosis detection accuracy, AI enables earlier interventions, slowing disease progression and reducing liver-related mortality.

AI in Fatty Liver Disease (MAFLD)

Fatty liver disease, a growing concern due to rising obesity rates, can progress to cirrhosis if undiagnosed. AI enhances MAFLD detection by:

· Automating Fat Quantification: AI detects liver steatosis with 98% accuracy compared to expert radiologists [1].

· Standardizing Hepato-Renal Index (HRI) Measurements: AI improves segmentation of the liver and kidney, reducing inter-observer variability [4].

· Predicting Fat Fraction: AI models using raw ultrasound data achieve MRI-equivalent accuracy in estimating liver fat percentage, providing a cost-effective alternative [2].

Earlier and more precise detection of MAFLD allows for timely lifestyle modifications and pharmacologic interventions.

AI in Liver Tumor Detection and Classification

Accurately distinguishing benign from malignant liver lesions is one of the biggest challenges in ultrasound diagnostics. AI aids in:

· Automated Lesion Detection: Deep learning algorithms analyze shape, echotexture, and vascular patterns, achieving an AUC of 0.92, comparable to MRI findings [6].

· Classification of Liver Tumors: Convolutional neural network (CNN)-based AI models classify tumors with 96.3% accuracy and an AUC of 0.994, outperforming traditional methods [3].

· Radiomics and Contrast-Enhanced Ultrasound (CEUS): AI detects enhancement patterns over time, improving hepatocellular carcinoma (HCC) detection rates [8].

· Predicting Tumor Characteristics: AI models can assess microvascular invasion in liver cancer, helping guide treatment decisions [7].

These advancements allow for more precise diagnoses, reducing the need for unnecessary biopsies and enabling earlier cancer treatment.

AI for Automated Liver Measurements

AI also enhances routine liver ultrasound evaluations by automating key measurements:

· Liver Size Estimation: AI accurately segments liver contours for precise volume measurement [4].

· Standardized Echogenicity Analysis: AI quantifies liver brightness to objectively assess fat content and fibrosis [1].

· Quality Control in Ultrasound Imaging: AI assists sonographers by optimizing probe positioning, reducing operator-dependent variability [7].

These automated features improve reproducibility, saving time and enhancing diagnostic reliability.

References

1. Barre, A., et al. (2023). AI-based ultrasound for fatty liver disease diagnosis. Journal of Personalized Medicine. https://doi.org/10.3390/jpm13121704
2. Ensemble Model Study. (2024). Precision NAFLD diagnosis. MDPI Bioengineering. https://doi.org/10.3390/bioengineering11060600
3. Huang, Y., et al. (2022). The feasibility of using artificial intelligence to aid in detecting focal liver lesions. Scientific Reports. https://doi.org/10.1038/s41598-022-11506-z
4. Intelligent Ultrasound Group. (2023). AI-based diagnostic tools for liver disease. Intelligent Ultrasound. https://www.intelligentultrasound.com
5. Kwan, A., et al. (2025). Artificial intelligence helps find undiagnosed liver disease. New England Journal of Medicine AI.
6. Lee, J., et al. (2024). AI for ultrasonographic detection of liver lesions. Scientific Reports. https://doi.org/10.1038/s41598-024-71657-z
7. Lin, X., et al. (2021). Application of artificial intelligence in non-alcoholic fatty liver disease. Diagnostics. https://doi.org/10.3390/diagnostics11122357
8. Yasaka, K., et al. (2022). Artificial intelligence models for the diagnosis and management of liver diseases. PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC9816706/