Automated high-fidelity extraction of patient data from survival curves

StochIKM: Automated High-Fidelity Reconstruction of Individual Patient Data from Published Kaplan-Meier Curves

Artificial Intelligence

Summary

Medical studies often show patient survival using Kaplan-Meier curves, but the detailed patient data behind these charts is usually not shared. The authors offer a method called StochIKM that uses artificial intelligence to read these charts and recreate detailed patient data accurately. Their approach checks that the recreated data matches key details like survival rates and patient counts exactly. This method produces multiple possible datasets instead of just one, reflecting the uncertainties in the original chart. It was tested on real cancer study data and outperformed an existing method in matching published survival information.

What this means in practice

  • For clinical data analysts: Reconstruct detailed patient-level survival data from published Kaplan-Meier charts when original data is unavailable to support further study or meta-analysis.
  • For health technology assessment teams: Generate accurate input patient data for economic or treatment modeling based on published survival curves in oncology and other fields.

Authors

Liangze Ke, Daniel Zelterman, Wei Wei, Guannan Gong

Abstract

Individual patient data are the gold standard for secondary analysis of time-to-event evidence, but they are rarely released, and reconstruction from a published Kaplan-Meier figure could help. Existing methods all claim high fidelity, each judged by whichever agreements it happens to achieve, and each returns a single dataset although the map from figure to data is not bijective. We define fidelity explicitly: a dataset is admissible when it reproduces the risk table exactly, follows the digitized curve to within the image resolution, and respects the printed median, interval and event count. We propose StochIKM, a stochastic algorithm that digitizes the figure with a vision-capable language model and uses the high-fidelity definition as constraints, returning an admissible set of individual patient datasets. A solution is guaranteed to exist whenever the figure has been read correctly, so a failure indicts either the published figure or its digitization. In simulation and in 79 arms from published oncology trials, StochIKM reproduces the published risk tables and medians more closely than LLM-improved IPDfromKM, a high-impact alternative.
View PDFOpen arXiv