Efficient Detection of Hepatic Steatosis in Ultrasound Images Using Convolutional Neural Networks: A Comparative Study
Bibliographic record
Abstract
Introduction: Artificial Intelligence (AI) is widely used in medical studies to interpret imaging data and improve the efficiency of healthcare professionals.Nonalcoholic fatty liver disease (NAFLD) is a common liver abnormality associated with an increased risk of hepatic cirrhosis, hepatocellular carcinoma, and cardiovascular morbidity and mortality.This study explores the use of AI for automated detection of hepatic steatosis in ultrasound images.Background: Ultrasound is a non-invasive, cost-effective, and widely available method for hepatic steatosis screening.However, its accuracy depends on the operator's expertise, necessitating automated methods to enhance diagnostic accuracy.AI, particularly Convolutional Neural Network (CNN) models, can provide accurate and efficient analysis of ultrasound images, enabling automated detection, improving diagnostic accuracy, and facilitating real-time analysis.Problem Statement: This study aims to evaluate deep learning methods for binary classification of hepatic steatosis using ultrasound images.Methodology: Open-source data is used to prepare three groups (A, B, C) of ultrasound images in different sizes.Images are augmented using seven pre-processing approaches (resizing, flipping, rotating, zooming, contrasting, brightening, and wrapping) to increase image variations.Seven CNN classifiers (EfficientNet-B0, ResNet34, AlexNet, DenseNet121, ResNet18, ResNet50, and MobileNet_v2) are evaluated using stratified 10fold cross-validation.Six metrics (accuracy, sensitivity, specificity, precision, F1 score, and MCC) are employed, and the best-performing fold epochs are selected.Experiments and Results: The study evaluates seven models, finding EfficientNet-B0, ResNet34, DenseNet121, and AlexNet to perform well in groups A and B. EfficientNet-B0 shows the best overall performance.It achieves high scores for all six metrics, with accuracy rates of 98.9%, 98.4%, and 96.3% in groups A, B, and C, respectively.Discussion and Conclusion: EfficientNet-B0, ResNet34, and DenseNet121 exhibit potential for classifying fatty liver ultrasound images.EfficientNet-B0 demonstrates the best average accuracy, specificity, and sensitivity, although more training data is needed for generalization.Complete and mediumsized images are preferred for classification.Further evaluation of other classifiers is necessary to determine the best model.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".