Artificial Intelligence Applications in Hepatocellular Carcinoma Diagnosis

Advances in artificial intelligence have revolutionised the detection, classification and prognostication of hepatocellular carcinoma (HCC). By harnessing vast clinical and imaging datasets, machine-learning and deep-learning frameworks extract subtle patterns beyond human perception. Convolutional neural networks applied to contrast-enhanced magnetic resonance imaging and multiphase computed tomography achieve high accuracy in distinguishing HCC from benign lesions. Simultaneously, predictive models based on biochemical markers and clinical variables employ ensemble learning and survival forests to flag malignancy and estimate recurrence risk. Radiomic analysis further quantifies tumour heterogeneity, while class activation mapping highlights diagnostic regions within images. Integration of these AI tools in multidisciplinary workflows has demonstrated improvements in sensitivity, specificity and workflow efficiency, paving the way for non-invasive, standardised decision support in HCC diagnosis.

Research from Nature Portfolio

A fully automated diagnostic system evaluated in over 12 000 patients across multiple centres applied deep learning to computed tomography scans. This system accurately triaged patients into benign and malignant categories, yielding high negative predictive values and enhancing radiologist performance, particularly for hepatocellular carcinoma. Additionally, convolutional neural networks fine-tuned on hepatobiliary phase magnetic resonance images demonstrated robust detection of HCC lesions, achieving sensitivity above 85 % and specificity exceeding 90 % on external validation cohorts. Moreover, a machine-learning framework integrating gradient boosting and tumour biomarkers surpassed single-marker approaches by halving the misclassification rate of HCC presence, delivering an area under the curve approaching 0.94 and demonstrating the translational potential of real-world clinical data.

Artificial Intelligence Applications in Hepatocellular Carcinoma Diagnosis publication trend

The graph below shows the total number of articles in artificial intelligence applications in hepatocellular carcinoma diagnosis across all publications each year (not limited to Nature Index journals).

Technical terms

Convolutional neural network (CNN): A deep-learning architecture that applies convolutional filters to extract hierarchical features from imaging data.

Radiomics: The high-throughput extraction of quantitative features from medical images to characterise tumour phenotype.

Random survival forest: An ensemble machine-learning method for analysing time-to-event data by aggregating decision trees to predict survival outcomes.

F1-score: The harmonic mean of precision and recall, reflecting a balance between false positives and false negatives.

Class activation map (CAM): A visualisation technique that highlights image regions most influential to a neural network’s classification decision.

References

1. A multicenter clinical AI system study for detection and diagnosis of focal liver lesions. Nature Communications (2024).
2. Detection of Hepatocellular Carcinoma in Contrast-Enhanced Magnetic Resonance Imaging Using Deep Learning Classifier: A Multi-Center Retrospective Study. Scientific Reports (2020).
3. Machine-learning Approach for the Development of a Novel Predictive Model for the Diagnosis of Hepatocellular Carcinoma. Scientific Reports (2019).
4. Feature reduction for hepatocellular carcinoma prediction using machine learning algorithms. Journal of Big Data (2024).
5. Deep Learning for Accurate Diagnosis of Liver Tumor Based on Magnetic Resonance Imaging and Clinical Data. Frontiers in Oncology (2020).
6. Artificial intelligence in liver diseases: Improving diagnostics, prognostics and response prediction. JHEP Reports (2022).
7. Deep learning for differential diagnosis of malignant hepatic tumors based on multi-phase contrast-enhanced CT and clinical data. Journal of Hematology & Oncology (2021).