Caring Better for Our CommunitieModelling the math of bile secretion assists clinical decision making in obstructive jaundices Using Informatics Driven Population-Based Health in Primary Care

Authors: Adam Strier MD/Bnai Zion Surgery department, Haifa Israel Ronen Tal-Bozer PhD/ Sense/ Medical Bioinformatics-Bar-Ilan University, Israel Dean Keren MD/ Bnai Zion Gastroenterology dept., Haifa Israel Ibrahim Mattar MD/ Head of Bnai Zion Surgery department, Haifa Israel Ron Unger PhD/ Associate Professor, Head of the Biomedical Informatics Program at Bar-Ilan: adam.strier@gmail.com

Biliary disease affects approximately 10-15% of the western population. While biliary colic is largely self-limiting, acute cholecystitis and ascending cholangitis may often be life-threatening and generally require endoscopic and/or surgical intervention, exposing patients to possible peri-procedural complications.

One of the most common procedures performed for gallstone disease is Endoscopic Retrograde Cholangio-Pancreatography (ERCP), used to both diagnose and treat impacted bile duct stones in numerous ways. This intervention largely replaced surgical exploration of bile ducts with all its complications, but is nevertheless considered as an invasive intervention with a complication rate of 7%, of which 1.67% graded severe, including post-ERCP mortality rate of 0.33%. The decision to perform an ERCP is therefore a difficult one, and is usually made by a multi-disciplinary consultation.

Current guidelines published by the American Society of Gastroenterologists stratify the patients to High-, Intermediate-, and Low- probability of Bile Duct Stones, using serum Bilirubin level as the main index. While rises in serum bilirubin levels are definitely considered a hallmark of obstructive jaundice, many cases of choledocholithiasis present with mild or no serum bilirubin elevation. Four more liver enzymes are important markers of hepatic injury and cholestasis, and accurate understanding of their individual and orchestrated dynamics in time can lead us to distinctive clusters of patients with different disease dynamics. This will enable us to more accurately infer probability of success by ERCP. No comprehensive model of liver enzyme elevation in cholestasis has been so far proposed. We present a novel, computer-assisted pre-ERCP risk-benefit patient stratification method, using graph-analysis algorithms, thus laying the groundwork for a systems-biology approach to a more comprehensive understanding of cholestasis as a pathological process.

We used retrospective data from patients' EMR in the last 10 years, from a single medical center in Haifa, Israel, with ERCP and general surgical and laparoscopic capabilities. For every patient several blood samples are recorded, including complete liver enzyme panels, as well as textual imaging results from abdominal ultrasound (US) and computerized tomography (CT) scans, and their textual ERCP reports.
We used a natural language processing (NLP) algorithm to identify patients radiographically diagnosed with choledocholithiasis, as well as to differentiate between 2 clusters of patients according to their ERCP results - "Success" vs "Failure".

We implemented an advanced multi-objective genetic algorithm already published under the name "Profilase" to cluster patients according to their serums liver enzyme levels time-series to propose a model for prediction of success of ERCP for future patients presenting with obstructive jaundice, thus assisting the clinical decision-making.