DILIGENT: An AI Clinical Assistant for Drug-Induced Liver Injury

Project Name: DILIGENT-Clinical-Copilot Core Positioning: A large language model-based AI clinical assistant specifically designed to assist doctors in detecting and managing Drug-Induced Liver Injury (DILI), providing case analysis, RAG retrieval, and conversation recording functions. Original Author/Maintainer: CTCycle Source Platform: GitHub Original Link: https://github.com/CTCycle/DILIGENT-Clinical-Copilot Release Time: June 3, 2026 Tech Stack: FastAPI (Backend), Angular+TypeScript (Frontend) License: Polyform Noncommercial License 1.0.0 (free for non-commercial use)

Drug-Induced Liver Injury (DILI) is one of the most common and severe types of adverse drug reactions, accounting for over 50% of all acute liver failure cases, with diverse clinical manifestations (from asymptomatic liver enzyme elevation to life-threatening liver failure).

DILI diagnosis faces four major challenges:

1. Complex Etiology: Need to exclude other causes such as viral hepatitis and alcoholic liver disease;
2. Time Correlation: Need to accurately trace the time relationship between medication history and liver injury;
3. Numerous Drugs: Over 1000 drugs are associated with liver injury, imposing a heavy memory burden;
4. Individual Differences: Genetic background, underlying diseases, and combined medications affect risk assessment.

DILIGENT is developed to address these clinical pain points, leveraging the knowledge integration and reasoning capabilities of LLM to provide systematic DILI assessment support.

System Architecture:

• Backend: Adopts FastAPI framework, which has advantages of high performance, asynchronous support, and automatic document generation, responsible for case storage and retrieval, LLM interaction, RAG pipeline implementation, and conversation management.
• Frontend: Built with Angular+TypeScript, providing type safety guarantees and adapting to complex clinical data models.
• Deployment Options: Supports Tauri-packaged Windows desktop version, which can run offline to meet medical data privacy compliance.

Environment Requirements:

• Python 3.14+
• Node.js 18+ & npm
• Optional: Ollama (local model operation)

Port Configuration:

• Frontend UI: http://127.0.0.1:9847
• Backend API: http://127.0.0.1:7690

Configuration Management: Manage operation modes (development/desktop version, etc.) through the settings/.env file.

Core Functions:

1. Structured Case Collection: Guide doctors to enter standardized information (medical history, medication records, laboratory data such as ALT/AST, etc.) to promote the standardization of clinical thinking.
2. AI-Assisted Analysis: Call LLM to complete multi-dimensional assessment, including RUCAM score calculation, drug-liver injury causal probability assessment, differential diagnosis suggestions, and supplementary examination plans.
3. RAG-Enhanced Retrieval: Combine medical literature, drug manuals, and clinical guidelines to reduce LLM hallucinations and ensure evidence-based suggestions.
4. Conversation Persistence: Save the analysis process, support reviewing historical results, updating assessments, tracking disease trends, and generating archiving/referral reports.

Clinical Value:

1. Improve Diagnostic Accuracy: Systematic assessment process + AI knowledge integration reduce missed diagnoses and misdiagnoses (especially in scenarios involving rare drugs or complex combined medications).
2. Enhance Work Efficiency: Automate causal assessment and report generation to save doctors' documentation time.
3. Promote Standardized Diagnosis and Treatment: Built-in internationally recognized RUCAM assessment framework to popularize standardized processes.
4. Support Medical Education: Serve as an interactive learning tool to help residents/medical students understand the diagnostic logic of DILI.

Current Limitations:

• Under active development, there may be incomplete functions or potential defects;
• Database schema may need reinitialization during version upgrades;
• Depends on the availability of external LLM API or local Ollama service.

Future Directions:

• Expand support for assessment of more types of adverse drug reactions;
• Integrate electronic medical record systems (EMR/EHR) to achieve automatic data synchronization;
• Multi-language support to serve the global medical market;
• Train machine learning models based on accumulated case data.

DILIGENT Clinical Copilot focuses on DILI, a specific clinical problem, and provides practical decision support for doctors through systematic information collection, knowledge-enhanced AI analysis, and traceable conversation management—rather than replacing doctors.

This "narrow and deep" vertical application model is more practically valuable than "broad and comprehensive" general solutions, and its design concept and implementation provide useful references for the development of AI-assisted diagnostic tools in other specialized fields.