A multi-label dataset and its evaluation for automated scoring system for cleanliness assessment in video capsule endoscopy.

An automated scoring system for cleanliness assessment during video capsule endoscopy (VCE) is presently lacking. The present study focused on developing an approach to automatically assess the cleanliness in VCE frames as per the latest scoring i.e., Korea-Canada (KODA). Initially, an easy-to-use mobile application called artificial intelligence-KODA (AI-KODA) score was developed to collect a multi-label image dataset of twenty-eight patient capsule videos. Three readers (gastroenterology fellows), who had been trained in reading VCE, rated this dataset in a duplicate manner. The labels were saved automatically in real-time. Inter-rater and intra-rater reliability were checked. The developed dataset was then randomly split into train:validate:test ratio of 70:20:10 and 60:20:20. It was followed by a comprehensive benchmarking and evaluation of three multi-label classification tasks using ten machine learning and two deep learning algorithms. Reliability estimation was found to be overall good among the three readers. Overall, random forest classifier achieved the best evaluation metrics, followed by Adaboost, KNeighbours, and Gaussian naive bayes in the machine learning-based classification tasks. Deep learning algorithms outperformed the machine learning-based classification tasks for only VM labels. Thorough analysis indicates that the proposed approach has the potential to save time in cleanliness assessment and is user-friendly for research and clinical use. Further research is required for the improvement of intra-rater reliability of KODA, and the development of automated multi-task classification in this field.

Liver stiffness assessment as an alternative to hepatic venous pressure gradient for predicting rebleed after acute variceal bleed: A proof-of-concept study.

BACKGROUND AND AIM: Hepatic venous pressure gradient (HVPG), although an important determinant in predicting rebleeding after an episode of acute variceal bleed (AVB), is seldom utilized in clinical practice. We aimed to study the role of liver stiffness measurement (LSM) after variceal bleeding as a potential noninvasive predictor of rebleed. METHODS: This was a post hoc analysis of clinical trial of patients undergoing HVPG (postbleed HVPG) and LSM (postbleed LSM) assessment within 3-5 days of index AVB. HVPG response was assessed after 4 weeks of pharmacotherapy. Comparative assessment of long-term rebleeding rates stratified using postbleed LSM, postbleed HVPG, and HVPG response was performed. Decision curve analysis (DCA) was conducted to identify the most appropriate tool for routine use. RESULTS: Long-term clinical and HVPG response data were available for 48 patients post-AVB, of whom 45 patients had valid postbleed LSM. Rebleeding occurred in 13 (28%) patients over a median follow-up of 4 years with no early rebleeds. Postbleed LSM >30 kPa and baseline HVPG >15 mm Hg were optimal cutoffs for identifying patients at high risk of rebleeding. Time-dependent receiver operating characteristic curves and competing risk analysis accounting for death showed similar discriminative values for all three stratification tools. At usual risk thresholds, HVPG response had maximum benefit on DCA followed by postbleed LSM. On DCA, 50-60 additional HVPGs were required to detect one additional patient at high risk of rebleed. CONCLUSION: Liver stiffness measurement during AVB can potentially be used as an alternative to portal pressure indices in decompensated cirrhosis to identify those at high risk of late-onset rebleed.

Body mass index-based controlled attenuation parameter cut-offs for assessment of hepatic steatosis in non-alcoholic fatty liver disease.

BACKGROUND AND AIM: In patients with liver disease, etiology and body mass index (BMI) affects controlled attenuation parameter (CAP) assessment using FibroScan. We aimed to assess the performance characteristics of CAP for hepatic steatosis in patients with non-alcoholic fatty liver disease (NAFLD) stratified into obese (BMI ≥ 30 kg/m2) and non-obese (BMI < 30 kg/m2) subgroups. METHODS: In this prospective study, 219 consecutive adult NAFLD patients, with an available FibroScan value (liver stiffness measurement-[LSM] and CAP) and liver biopsy, were included. Receiver operating characteristic curves were used for assessment of the CAP cut-off values predicting different stages of hepatic steatosis. RESULTS: The mean ± standard deviation age of patients was 39.7 ± 10.5 years, 116 (53%) were males, and median (interquartile range) BMI was 31.8 (25.7-43.8) kg/m2. One hundred (45.7%) and 119 (54.3%) patients were non-obese and obese, respectively. The median values of CAP and LSM were significantly higher among obese patients as compared with the non-obese ones: 333 (304-368) vs. 320 (296-345) dB/m, p = 0.002 and 8.3 (6.1-11.4) vs. 6.6 (5.7-10.3) kPa, p = 0.012, respectively. Among non-obese NAFLD, optimal CAP cut-off values for steatosis (S) ≥ S1, ≥ S2, and ≥ S3 were 275 dB/m, 319 dB/m, and 337 dB/m, respectively. The corresponding CAP values among obese patients were higher as 285 dB/m, 340 dB/m, and 355 dB/m, respectively. BMI independently predicted CAP on multivariate analysis. The discordance of 2-grades between CAP and biopsy measured steatosis was seen in 13% in non-obese and 19.3% in obese NAFLD. CAP overestimated steatosis more often than underestimating it, with a higher proportion in obese NAFLD. CONCLUSION: In patients with NAFLD, interpretation of CAP requires consideration of BMI.

Controlled Attenuation Parameter for Assessment of Hepatic Steatosis in Indian Patients.

BACKGROUND/AIMS: The gold standard method for measurement of hepatic steatosis is liver histology. Controlled Attenuation Parameter (CAP) can measure hepatic steatosis non-invasively. We aimed to assess the accuracy of CAP for detection of hepatic steatosis. METHODS: A total of 462 patients (May 2012-January 2017)-89 non-alcoholic fatty liver disease, 182 chronic hepatitis B, 88 chronic hepatitis C and 103 patients with other etiologies who underwent simultaneous liver biopsy and CAP estimation using Transient Elastography (TE) were included. Steatosis was graded as S0: steatosis in 0-5% of hepatocytes, S1: 6-33%, S2: 34-66% and S3: 67-100%. Receiver Operating Characteristic (ROC) curves were plotted to evaluate the accuracy of CAP in detecting hepatic steatosis. Predictors of CAP were assessed by multivariate linear regression model. RESULTS: The mean age ± SD was 33.8 ± 11.6 years; 296 (64.1%) were males. On liver histology, steatosis grades S0, S1, S2 and S3 were seen in 331 (71.6%), 74 (16.0%), 39 (8.4%) and 18 (3.9%), respectively. The median CAP (IQR) values for S0, S1, S2, and S3 steatosis were 206 (176-252) dB/m, 295 (257-331) dB/m, 320 (296-356) dB/m, and 349 (306-363) dB/m, respectively. For estimation of ≥S1, ≥S2, and ≥S3 using CAP, AUROC were 0.879, 0.893, and 0.883, respectively. In multivariate analysis, only BMI (OR 1.18; CI, 1.11-1.26, P < 0.001) and grade of hepatic steatosis (grade 1, OR, 3.94; 95% CI, 1.58-9.84, P = 0.003; grade 2, OR 42.04; 95% CI, 4.97-355.31, P = 0.001 and grade 3, OR 35.83; 95% CI 4.31-297.61, P = 0.001) independently predicted CAP. CONCLUSIONS: CAP detects hepatic steatosis with good accuracy in Indian patients with various etiologies.