Selected WorkDILI-PREDICTION-APP

Drug-Induced Liver Injury (DILI) Rule Engine

Expert-system rules engine for DILI risk flagging using LFT-derived ratios and medication risk rules, built for deployment on Google Cloud and integration into clinical pipelines.

Healthcare Expert SystemsFastAPIGoogle CloudBigQuerySQLClinical Rules
Cover
Stack

Tech Stack

PythonFastAPIUvicornPydanticSQLAlchemyAlembicPostgreSQLBigQueryGoogle CloudSQLPandas
View on GitHub
SectionOverview

Overview

A clinical decision-support style package + API that evaluates liver function tests (LFTs) and medication history to flag potential drug-induced liver injury (DILI) risk using auditable clinical rules, with a data pipeline backed by Google Cloud services.

SectionThe Problem

The Problem

DILI screening is often delayed because clinicians must compute derived ratios across multiple labs, interpret injury patterns consistently, and cross-check medication risk,while lab + medication data lives across separate systems and formats.

SectionThe Solution

The Solution

Implemented a rules-based DILI evaluation framework using established heuristics (e.g., Hy’s Law criteria and R-ratio patterning) to classify hepatocellular/cholestatic/mixed injury patterns from LFT inputs. Built a FastAPI service (Uvicorn) with Pydantic validation and SQLAlchemy + Alembic for persistence, and integrated Google Cloud (BigQuery + GCP) for querying and analyzing longitudinal lab/medication data using SQL and Pandas-driven processing. Packaging is in progress to ship the engine as an installable module that can be embedded into existing clinical/EHR workflows.

SectionResults

Results & Impact

Generates consistent, explainable DILI risk flags and injury-pattern classifications from lab + medication inputs, with cloud-backed querying and reproducible rule execution suitable for integration into downstream alerting or clinical data pipelines.