Development and multicenter validation of FIB-6: A novel, machine learning, simple bedside score to rule out liver cirrhosis and compensated advanced chronic liver disease in patients with chronic hepatitis C

Use of artificial intelligence for liver diseases: A survey from the EASL congress 2024

Artificial intelligence (AI) methods enable humans to analyse large amounts of data, which would otherwise not be feasibly quantifiable. This is especially true for unstructured visual and textual data, which can contain invaluable insights into disease. The hepatology research landscape is complex and has generated large amounts of data to be mined. Many open questions can potentially be addressed with existing data through AI methods. However, the field of AI is sometimes obscured by hype cycles and imprecise terminologies. This can conceal the fact that numerous hepatology research groups already use AI methods in their scientific studies. In this review article, we aim to assess the contemporaneous use of AI methods in hepatology in Europe. To achieve this, we systematically surveyed all scientific contributions presented at the EASL Congress 2024. Out of 1,857 accepted abstracts (1,712 posters and 145 oral presentations), 6 presentations (∼4%) and 69 posters (∼4%) utilised AI methods. Of these, 55 posters were included in this review, while the others were excluded due to missing posters or incomplete methodologies. Finally, we summarise current academic trends in the use of AI methods and outline future directions, providing guidance for scientific stakeholders in the field of hepatology.

Liver fibrosis stage based on the four factors (FIB-4) score or Forns index in adults with chronic hepatitis C

BACKGROUND

The presence and severity of liver fibrosis are important prognostic variables when evaluating people with chronic hepatitis C (CHC). Although liver biopsy remains the reference standard, non-invasive serological markers, such as the four factors (FIB-4) score and the Forns index, can also be used to stage liver fibrosis.

OBJECTIVES

To determine the diagnostic accuracy of the FIB-4 score and Forns index in staging liver fibrosis in people with chronic hepatitis C (CHC) virus, using liver biopsy as the reference standard (primary objective). To compare the diagnostic accuracy of these tests for staging liver fibrosis in people with CHC and explore potential sources of heterogeneity (secondary objectives).

SEARCH METHODS

We used standard Cochrane search methods for diagnostic accuracy studies (search date: 13 April 2022).

SELECTION CRITERIA

We included diagnostic cross-sectional or case-control studies that evaluated the performance of the FIB-4 score, the Forns index, or both, against liver biopsy, in the assessment of liver fibrosis in participants with CHC. We imposed no language restrictions. We excluded studies in which: participants had causes of liver disease besides CHC; participants had successfully been treated for CHC; or the interval between the index test and liver biopsy exceeded six months.

DATA COLLECTION AND ANALYSIS

Two review authors independently extracted data. We performed meta-analyses using the bivariate model and calculated summary estimates. We evaluated the performance of both tests for three target conditions: significant fibrosis or worse (METAVIR stage ≥ F2); severe fibrosis or worse (METAVIR stage ≥ F3); and cirrhosis (METAVIR stage F4). We restricted the meta-analysis to studies reporting cut-offs in a specified range (+/-0.15 for FIB-4; +/-0.3 for Forns index) around the original validated cut-offs (1.45 and 3.25 for FIB-4; 4.2 and 6.9 for Forns index). We calculated the percentage of people who would receive an indeterminate result (i.e. above the rule-out threshold but below the rule-in threshold) for each index test/cut-off/target condition combination.

MAIN RESULTS

We included 84 studies (with a total of 107,583 participants) from 28 countries, published between 2002 and 2021, in the qualitative synthesis. Of the 84 studies, 82 (98%) were cross-sectional diagnostic accuracy studies with cohort-based sampling, and the remaining two (2%) were case-control studies. All studies were conducted in referral centres. Our main meta-analysis included 62 studies (100,605 participants). Overall, two studies (2%) had low risk of bias, 23 studies (27%) had unclear risk of bias, and 59 studies (73%) had high risk of bias. We judged 13 studies (15%) to have applicability concerns regarding participant selection. FIB-4 score The FIB-4 score's low cut-off (1.45) is designed to rule out people with at least severe fibrosis (≥ F3). Thirty-nine study cohorts (86,907 participants) yielded a summary sensitivity of 81.1% (95% confidence interval (CI) 75.6% to 85.6%), specificity of 62.3% (95% CI 57.4% to 66.9%), and negative likelihood ratio (LR-) of 0.30 (95% CI 0.24 to 0.38). The FIB-4 score's high cut-off (3.25) is designed to rule in people with at least severe fibrosis (≥ F3). Twenty-four study cohorts (81,350 participants) yielded a summary sensitivity of 41.4% (95% CI 33.0% to 50.4%), specificity of 92.6% (95% CI 89.5% to 94.9%), and positive likelihood ratio (LR+) of 5.6 (95% CI 4.4 to 7.1). Using the FIB-4 score to assess severe fibrosis and applying both cut-offs together, 30.9% of people would obtain an indeterminate result, requiring further investigations. We report the summary accuracy estimates for the FIB-4 score when used for assessing significant fibrosis (≥ F2) and cirrhosis (F4) in the main review text. Forns index The Forns index's low cut-off (4.2) is designed to rule out people with at least significant fibrosis (≥ F2). Seventeen study cohorts (4354 participants) yielded a summary sensitivity of 84.7% (95% CI 77.9% to 89.7%), specificity of 47.9% (95% CI 38.6% to 57.3%), and LR- of 0.32 (95% CI 0.25 to 0.41). The Forns index's high cut-off (6.9) is designed to rule in people with at least significant fibrosis (≥ F2). Twelve study cohorts (3245 participants) yielded a summary sensitivity of 34.1% (95% CI 26.4% to 42.8%), specificity of 97.3% (95% CI 92.9% to 99.0%), and LR+ of 12.5 (95% CI 5.7 to 27.2). Using the Forns index to assess significant fibrosis and applying both cut-offs together, 44.8% of people would obtain an indeterminate result, requiring further investigations. We report the summary accuracy estimates for the Forns index when used for assessing severe fibrosis (≥ F3) and cirrhosis (F4) in the main text. Comparing FIB-4 to Forns index There were insufficient studies to meta-analyse the performance of the Forns index for diagnosing severe fibrosis and cirrhosis. Therefore, comparisons of the two tests' performance were not possible for these target conditions. For diagnosing significant fibrosis and worse, there were no significant differences in their performance when using the high cut-off. The Forns index performed slightly better than FIB-4 when using the low/rule-out cut-off (relative sensitivity 1.12, 95% CI 1.00 to 1.25; P = 0.0573; relative specificity 0.69, 95% CI 0.57 to 0.84; P = 0.002).

AUTHORS' CONCLUSIONS

Both the FIB-4 score and the Forns index may be considered for the initial assessment of people with CHC. The FIB-4 score's low cut-off (1.45) can be used to rule out people with at least severe fibrosis (≥ F3) and cirrhosis (F4). The Forns index's high cut-off (6.9) can be used to diagnose people with at least significant fibrosis (≥ F2). We judged most of the included studies to be at unclear or high risk of bias. The overall quality of the body of evidence was low or very low, and more high-quality studies are needed. Our review only captured data from referral centres. Therefore, when generalising our results to a primary care population, the probability of false positives will likely be higher and false negatives will likely be lower. More research is needed in sub-Saharan Africa, since these tests may be of value in such resource-poor settings.

Validation of FIB-6 score in assessment of liver fibrosis in chronic hepatitis B

BACKGROUND

We recently developed a simple novel index called fibrosis 6 (FIB-6) using machine learning data analysis. We aimed to evaluate its performance in the diagnosis of liver fibrosis and cirrhosis in chronic hepatitis B (CHB).

METHODS

A retrospective observational analysis of data was obtained from seven countries (Egypt, Kingdom of Saudi Arabia (KSA), Turkey, Greece, Oman, Qatar, and Jordan) of CHB patients. The inclusion criteria were receiving an adequate liver biopsy and a complete biochemical and hematological data. The diagnostic performance analysis of the FIB-6 index was conducted and compared with other non-invasive scores.

RESULTS

A total of 603 patients were included for the analysis; the area under the receiver operating characteristic curve (AUROC) of FIB-6 for the discrimination of patients with cirrhosis (F4), compensated advanced chronic liver disease (cACLD) (F3 and F4), and significant fibrosis (F2-F4) was 0.854, 0.812, and 0.745, respectively. The analysis using the optimal cut-offs of FIB-6 showed a sensitivity of 70.9%, specificity of 84.1%, positive predictive value (PPV) of 40.3%, and negative predictive value (NPV) of 95.0% for the diagnosis of cirrhosis. For the diagnosis of cACLD, the results were 71.5%, 69.3%, 40.8%, and 89.2%, respectively, while for the diagnosis of significant fibrosis, the results were 68.3%, 67.5%, 59.9%, and 75.0%, respectively. When compared to those of fibrosis 4 (FIB-4) index, aspartate aminotransferase (AST)-to-platelet ratio index (APRI), and AST-to-alanine aminotransferase (ALT) ratio (AAR), the AUROC for the performance of FIB-6 was higher than that of FIB-4, APRI, and AAR in all fibrosis stages. FIB-6 gave the highest sensitivity and NPV (89.1% and 92.4%) in ruling out cACLD and cirrhosis, as compared to FIB-4 (63.8% and 83.0%), APRI (53.9% and 86.6%), and AAR (47.5% and 82.3%), respectively.

CONCLUSIONS

The FIB-6 index could be used in ruling out cACLD, fibrosis, and cirrhosis with good reliability.

Noninvasive Models to Assess Liver Inflammation and Fibrosis in Chronic HBV Infected Patients with Normal or Mildly Elevated Alanine Transaminase Levels: Which One Is Most Suitable?

The prevalence of substantial inflammation or fibrosis in treatment-naïve patients with chronic hepatitis B (CHB) and normal alanine transaminase (ALT) levels is high. A retrospective analysis was conducted on 559 consecutive patients with hepatitis B virus infection, who underwent liver biopsy, to investigate the value of noninvasive models based on routine serum markers for evaluating liver histology in CHB patients with normal or mildly elevated ALT levels and to provide treatment guidance. After comparing 55 models, we identified the top three models that exhibited excellent performance. The APGA model, based on the area under the receiver operating characteristic curve (AUROC), demonstrated a superior ability to evaluate significant (AUROC = 0.750) and advanced fibrosis (AUROC = 0.832) and demonstrated a good performance in assessing liver inflammation (AUROCs = 0.779 and 0.874 for stages G ≥ 2 and G ≥ 3, respectively). APGA also exhibited significant correlations with liver inflammation and fibrosis stage (correlation coefficients, 0.452 and 0.405, respectively (p < 0.001)). When the patients were stratified into groups based on HBeAg status and ALT level, APGA consistently outperformed the other 54 models. The other top two models, GAPI and XIE, also outperformed models based on other chronic hepatitis diseases. APGA may be the most suitable option for detecting liver fibrosis and inflammation in Chinese patients with CHB.

Noninvasive Tests to Assess Liver Stiffness in Patients with Chronic Hepatitis B: APRI, FIB‐4, and FIB‐5 Scores

Background. Invasive percutaneous liver parenchymal biopsy is the best test used to evaluate liver stiffness and fibrosis in the follow‐up and treatment of chronic hepatitis B (CHB) patients. In this study, we aimed to indirectly evaluate the severity of liver parenchymal fibrosis with tests used in the laboratory. Methods. This retrospective study was conducted with 201 patients diagnosed with CHB who underwent liver biopsy between 2021 and 2022. Preprocedural examination information, laboratory tests, and histopathological data of the patients were taken from the hospital database and examined. “Aspartate aminotransferase (AST)‐platelet ratio index” (APRI), “4 factor‐based fibrosis index” (FIB‐4) score, and “5 factor‐based fibrosis index” (FIB‐5) score were calculated and compared with liver histopathological features. Results. Of the 201 CHB patients, 76 were females and 125 were males. The average age of the patients was 38.05 ± 12.63 years. A weak, statistically significant correlation was observed between FIB‐4 and APRI scores. The patients’ significant fibrosis scores were 31.3% and 33.8%, respectively (r = 0.313; r = 0.338; p = 0.001; p < 0.01). The very weak negative correlation of 17.4% between the patients’ FIB‐5 score and fibrosis score was statistically significant (r = −0.174; p = 0.014; p < 0.05). Conclusions. According to the data we obtained in our study, while the APRI score and FIB‐4 score can be used safely, more comprehensive studies are needed for the reliability of the FIB‐5 score.

Multicenter validation of FIB-6 as a novel machine learning non-invasive score to rule out liver cirrhosis in biopsy-proven MAFLD

BACKGROUND AND AIMS

We previously developed and validated a non-invasive diagnostic index based on routine laboratory parameters for predicting the stage of hepatic fibrosis in patients with chronic hepatitis C (CHC) called FIB-6 through machine learning with random forests algorithm using retrospective data of 7238 biopsy-proven CHC patients. Our aim is to validate this novel score in patients with metabolic dysfunction-associated fatty liver disease (MAFLD).

METHOD

Performance of the new score was externally validated in cohorts from one site in Egypt (n = 674) and in 5 different countries (n = 1798) in Iran, KSA, Greece, Turkey and Oman. Experienced pathologists using METAVIR scoring system scored the biopsy samples. Results were compared with FIB-4, APRI, and AAR.

RESULTS

A total of 2472 and their liver biopsy results were included, using the optimal cutoffs of FIB-6 indicated a reliable performance in diagnosing cirrhosis, severe fibrosis, and significant fibrosis with sensitivity = 70.5%, specificity = 62.9%. PPV = 15.0% and NPV = 95.8% for diagnosis of cirrhosis. For diagnosis of severe fibrosis (F3 and F4), the results were 86.5%, 24.0%, 15.1% and 91.9% respectively, while for diagnosis of significant fibrosis (F2, F3 and F4), the results were 87.0%, 16.4%, 24.8% and 80.0%). Comparing the results of FIB-6 rule-out cutoffs with those of FIB-4, APRI, and AAR, FIB-6 had the highest sensitivity and NPV (97.0% and 94.7%), as compared to FIB-4 (71.6% and 94.7%), APRI (36.4% and 90.7%), and AAR (61.2% and 90.9%).

CONCLUSION

FIB-6 score is an accurate, simple, NIT for ruling out advanced fibrosis and liver cirrhosis in patients with MAFLD.

Hybrid model for precise hepatitis-C classification using improved random forest and SVM method

Hepatitis C Virus (HCV) is a viral infection that causes liver inflammation. Annually, approximately 3.4 million cases of HCV are reported worldwide. A diagnosis of HCV in earlier stages helps to save lives. In the HCV review, the authors used a single ML-based prediction model in the current research, which encounters several issues, i.e., poor accuracy, data imbalance, and overfitting. This research proposed a Hybrid Predictive Model (HPM) based on an improved random forest and support vector machine to overcome existing research limitations. The proposed model improves a random forest method by adding a bootstrapping approach. The existing RF method is enhanced by adding a bootstrapping process, which helps eliminate the tree's minor features iteratively to build a strong forest. It improves the performance of the HPM model. The proposed HPM model utilizes a 'Ranker method' to rank the dataset features and applies an IRF with SVM, selecting higher-ranked feature elements to build the prediction model. This research uses the online HCV dataset from UCI to measure the proposed model's performance. The dataset is highly imbalanced; to deal with this issue, we utilized the synthetic minority over-sampling technique (SMOTE). This research performs two experiments. The first experiment is based on data splitting methods, K-fold cross-validation, and training: testing-based splitting. The proposed method achieved an accuracy of 95.89% for k = 5 and 96.29% for k = 10; for the training and testing-based split, the proposed method achieved 91.24% for 80:20 and 92.39% for 70:30, which is the best compared to the existing SVM, MARS, RF, DT, and BGLM methods. In experiment 2, the analysis is performed using feature selection (with SMOTE and without SMOTE). The proposed method achieves an accuracy of 41.541% without SMOTE and 96.82% with SMOTE-based feature selection, which is better than existing ML methods. The experimental results prove the importance of feature selection to achieve higher accuracy in HCV research.