Title: AutoForest: End-to-End Evidence Extraction and Synthesis for Automated Forest Plot Generation from Biomedical Research

Abstract:

While systematic reviews depend heavily on forest plots to aggregate quantitative evidence from various biomedical studies, the current workflow for creating these visuals is disjointed and resource-heavy. Scientists are required to decipher intricate clinical narratives, manually pull outcome statistics from trial reports, specify suitable interventions and control groups, reconcile divergent study methodologies, and execute meta-analytic calculations. This process typically involves specialized software that necessitates structured data entry and significant domain knowledge. Although recent advancements show that large language models can successfully pull study-level information from unstructured text, no prior tool has automated the entire journey from raw manuscripts to synthesized forest plots.

To bridge this void, we present AutoForest, a pioneering end-to-end platform capable of producing publication-quality forest plots directly from biomedical literature. By inputting one or more research papers, AutoForest automatically proposes ICO (Intervention, Comparator, Outcome) components, retrieves outcome data, conducts statistical synthesis, and generates the final graphical representation. This paper outlines the system’s architecture and interface, while a user study with clinicians validates its efficacy through real-world applications. The findings illustrate how AutoForest streamlines evidence synthesis and significantly reduces the obstacles associated with performing meta-analyses.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC