Currently accepted at: Journal of Medical Internet Research
Date Submitted: Dec 18, 2024
Date Accepted: Apr 17, 2025
This paper has been accepted and is currently in production.
It will appear shortly on 10.2196/70257
The final accepted version (not copyedited yet) is in this tab.
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
From text to treatment: a systematic review and meta-analysis of rule-based natural language processing algorithms in oncology clinical decision support
ABSTRACT
Background:
The rapid growth of big data in healthcare has led to increased interest in artificial intelligence (AI) and natural language processing (NLP) for clinical decision support.
Objective:
This systematic review assesses rule-based NLP algorithms for oncology.
Methods:
A PRISMA compliant systematic review was conducted across multiple databases, identifying studies that utilised rule-based NLP algorithms for clinical decision support in oncology. Data extraction, quality assessment, and meta-analysis were performed to evaluate the algorithms' sensitivity, specificity, and overall performance.
Results:
A total of 89 studies were included in the review, covering various cancer types and clinical tasks. Of these, 35 studies reported sufficient data to construct contingency tables, enabling calculation of test accuracy, with sensitivity ranging from 0.49 to 1.00% (mean 0.92, SD 0.10) and specificity ranging from 0.48 to 1.00 (mean 0.92, SD 0.13). Bivariate random-effects meta-analysis generated pooled sensitivity and specificity values of 0.96 (95% CI 0.93,0.97) and 0.98 (95% CI 0.95, 0.99). Notably, our study highlights the need for standardised reporting in NLP research and discusses the importance of addressing publication bias.
Conclusions:
Rule-based NLP algorithms demonstrate potential in oncology clinical decision support, but their performance varies based on the specific task and cancer type. The study emphasises the importance of standardising reporting practices and encourages further research to refine and evaluate these algorithms in real-world clinical settings. Clinical Trial: PROSPERO (CRD42020180676).
Citation
