Diagnosis of Liver Disease Using Machine Learning

Abstract

Liver disease continues to be a notable Worldwide health concern, with numerous patients receiving diagnoses at a delayed stage, with far reaching adverse consequences including higher rates of reaction requiring intensive medical treatments or death. Traditionally, diagnosis depended out on manual examination of medical parameters and imaging which is all time taking process and thus may cause errors as they are of human. This is compounded further by limited availability of experienced radiologists and expert diagnostic tools in areas. The current study meets the requirement of a reliable, clinically translatable liver disease prediction algorithm incorporating various diagnostic tools within limited time and resource constraints. LiverCare AI: Our livercareAI, developed as a Flask web application, offers two ways of prediction; using your body parameters (on test values) and by looking at the image for representations from the scan format(ultrasound/CT/MRI). Aided by machine learning algorithms trained on a dataset of Indian liver patient records, the system also incorporates image classification models that deliver high-confidence predictions. It features a modern, responsive user interface with educational resources on liver health and personalized recommendations according to the prediction results.First efficacy testing shows a highly predictive ability and amazing performance in both parameter-based and image-based settings. This approach combines the processing of clinical data with AI-based imaging detection, providing earlier diagnosis and reducing diagnostic delays. This provides an adjustable and broad resource-limited telemedicine integration solution.

Introduction

1. Introduction & Problem Statement

Liver diseases like cirrhosis, hepatitis, and fatty liver disease are major global health issues. Current diagnostic methods—blood tests, clinical exams, and imaging—are effective but often:

• Unavailable or delayed in rural or resource-poor areas.

• Dependent on specialist interpretation, introducing subjectivity and inconsistent outcomes.

2. Proposed Solution: LiverCare AI

LiverCare AI is a dual-mode AI system designed to improve early, accurate diagnosis of liver diseases through:

• Parameter-based prediction using health records (via Random Forest model).

• Image-based classification using liver scans (via Convolutional Neural Networks or CNNs).

Built as a Flask-based web application, it features:

• User-friendly and mobile-compatible interface.

• Confidence scores, educational content, and personalized lifestyle recommendations.

• Accessibility for both urban and rural healthcare settings.

3. Literature Review Highlights

• Traditional methods are often unavailable or inefficient in low-resource areas.

• AI and ML offer automation, accuracy, and early detection capabilities.

• CNNs can identify subtle patterns in images, outperforming human diagnosis in some contexts.

4. Methodology

A. Dataset Used

• Indian Liver Patient Dataset (ILPD): Includes biochemical markers like ALT, AST, Bilirubin, etc., along with demographic data.

• Liver scan images (ultrasound, CT, MRI): Collected from open-source repositories for image-based predictions.

B. Data Preprocessing

• Removal of incomplete records.

• StandardScaler used to normalize numerical data.

• Image resizing and normalization for CNN compatibility.

C. Model Development

• Random Forest: Used for predicting based on patient parameters.

• CNN: Used for analyzing liver images.

• Model Training & Testing: Focused on balancing accuracy and generalizability.

D. Flask Web Integration

• Combines both models into a single interface.

• Accepts either patient data or liver scans as input.

• Offers instant results, confidence levels, and health tips.

5. Evaluation & Results

Random Forest (Parameter-based model)

• Metrics Used:

  • Accuracy: Overall correctness.

  • Precision: Reduces false positives.

  • Recall: Emphasizes detecting actual cases.

  • F1-Score: Balanced measure (harmonic mean of precision and recall).

CNN (Image-based model)

• AUC-ROC: Evaluates the ability to distinguish between diseased and healthy cases.

• Confusion Matrix: Analyzes error types (false positives/negatives).

Additional Evaluations

• Inference Time: Fast response (within seconds), suitable for real-time use.

• Resource Efficiency: Lightweight, deployable even in low-resource settings.

6. Key Advantages

• Automated & Early Diagnosis: Reduces reliance on specialists.

• Scalable: Suitable for both rural clinics and urban hospitals.

• Educational: Enhances patient understanding with resources.

• Mobile-Compatible: Broadens accessibility.

Conclusion

LiverCare AI is a two-prone artificial intelligence framework proposed in this study to counter the pressing issues related to on-time, precise, and low-cost liver disorders detection. It comprises a Flask-based web application in which two different classifiers are integrated: One for predicting based on parameters using Random Forest, and the other for analysing images using Convolutional Neural Network (CNN).Overall Flow: User input either patient medical parameters or liver scan images -pre processing and model inference result generation suggestions for educational material. To discriminate between normal and defective objects, the implementation phase AME model achieved strong results in all metrics (high accuracy with precision, recall, and AUC-ROC), showing that the proposed solution is reliable for detection even low false-positives and false-negatives. The integration of tabular and image analysis in a single framework directly fits the abstract problem statementpaving the way for effective earlier diagnostics in resource poor settings.In the future, this might involve increasing and broadening the dataset demographics, incorporating more imaging such as elastography into treatment response prediction, and development of XAI methods to enhance clinical interpretability for model outputs.

References

[1] A. K. Das, S. Mishra, and S. Mukherjee, \"Liver disease detection utilizing machine learning algorithms,\" International Journal of Computer Applications, vol. 192, no. 2, pp. 1–6, 2021. [2] Deshmukh, P. B., and Kshirsagar, V. P. \"Development of a predictive model for liver disease utilizing decision tree and random forest algorithms.\" International Journal of Engineering Research & Technology, vol. 10, no. 6, pp. 512–517, 2021. [3] S. V. Mohan, B. S. Naidu, and K. V. Reddy, “Diagnosis of liver diseases from medical parameters using deep learning,” Biomedical Signal Processing and Control, vol. 70, pp. 103060, 2021. [4] In the publication by F. Chollet titled \"Xception: Deep learning leveraging depthwise separable convolutions,\" presented at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) in 2017, pages 1251-1258. [5] S. M. Kamal, R. K. Das, and T. K. Das, “Liver fibrosis detection through automated analysis of ultrasound images using convolutional neural networks,” Journal of Digital Imaging, vol. 34, no. 4, pp. 854–865, 2021. [6] Y. Lecun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, 2015. [7] Kumar, R., & Gopal, P. (2021). Hybrid ensemble machine learning model for liver disease prediction. Expert Systems with Applications, 185, 115596. [8] J. Brownlee, Python Machine Learning, Machine Learning Mastery (2020). [9] T. Esteva et al., “A guide to deep learning in healthcare,” Nature Medicine, vol. 25, pp. 24–29, 2019. [10] M. Abdar et al., “A review of uncertainty quantification in deep learning: Techniques, applications, and challenges,” Information Fusion, vol. 76, pp. 243–297, 2021.

Copyright

Copyright © 2025 Mrs. Subhashree D C, Dr. Girish Kumar D, Ms. Pooja, Mrs. Sharvani V, Ms. M M Harshitha. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.