A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging - Brief Correspondence

Aug 5, 2026

Upload by Anda

Prostate magnetic resonance imaging has become the standard imaging approach for prostate cancer, with interpretation standardized using PI-RADS. Hundreds of studies are published annually on the diagnostic performance of PI-RADS, making systematic reviews and meta-analyses highly resource-intensive. A machine learning framework using a naïve Bayesian classifier in an active learning setting was assessed on a living systematic review dataset. In simulation, the approach enabled exclusion of studies after 95% of relevant studies were identified, with work saved over sampling (WSS@95%) of about 28%, implying reduced manual screening effort in practice.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages5

Reading Time13 min

AccessFree to download

Abstract

Background and objective

Methods

Results and implications

Patient summary

Keywords

EUR Research Information Portal A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging Published in: European Urology Open Science Publication status and date: Published: 01/10/2023 DOI (link to publisher): 10.1016/j.euros.2023.07.005 Document Version Publisher's PDF, also known as Version of record Document License/Available under: CC BY-NC-ND Citation for the published version (APA): Nedelcu, A. , Oerther, B. , Engel, H. , Sigle, A. , Schmucker, C. , Schoots, IG. , Eisenblätter, M. , & Benndorf, M. (2023) . A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging. European Urology Open Science, 56, 11-14. [https://doi.org/10.1016/j.euros.2023.07.005](https://doi.org/10.1016/j.euros.2023.07.005) Link to publication on the EUR Research Information Portal Terms and Conditions of Use Except as permitted by the applicable copyright law, you may not reproduce or make this material available to any third party without the prior written permission from the copyright holder(s) . Copyright law allows the following uses of this material without prior permission: • you may download, save and print a copy of this material for your personal use only; • you may share the EUR portal link to this material. In case the material is published with an open access license (e.g. a Creative Commons (CC) license), other uses may be allowed. Please check the terms and conditions of the specific license. Take-down policy If you believe that this material infringes your copyright and/or any other intellectual property rights, you may request its removal by contacting us at the following email address: [openaccess.library@eur.nl. Please](openaccess.library@eur.nl. Please) provide us with all the relevant information, including the reasons why you believe any of your rights have been infringed. In case of a legitimate complaint, we will make the material inaccessible and/or remove it from the website. EUROPEAN U ROLOGY OPEN SCIEN C E 5 6 ( 202 3 ) 1 1 – 14 available [at www.sciencedirect.com](at www.sciencedirect.com) journal [homepage: www.eu-openscience.europeanurology. com](homepage: www.eu-openscience.europeanurology. com) | Brief Correspondence
A Machine Learning Framework Reduces the Manual Workload for Systematic Reviews of the Diagnostic Performance of Prostate Magnetic Resonance Imaging
Andrea Nedelcu a, Benedict Oerthera, Hannes Engel a, August Sigleb,c, Christine Schmuckerd, Ivo G. Schootse, Michel Eisenbl¨atterf, Matthias Benndorfa,f, * | | | | --- | --- | --- | | Article info | Abstract
Prostate magnetic resonance imaging has become the imaging standard for prostate cancer in various clinical settings, with interpretation standardized according to the Prostate Imaging Reporting and Data System (PI-RADS). Each year, hundreds of scientiﬁc studies that report on the diagnostic performance of PI-RADS are published. To keep up with this ever-increasing evidence base, systematic reviews and meta-analyses are essential. As systematic reviews are highly resource-intensive, we investigated whether a machine learning framework can reduce the manual workload and speed up the screening process (title and abstract). We used search results from a living systematic review of the diagnostic performance of PI-RADS (1585 studies, of which 482 were potentially eligible after screening). A naïve Bayesian classiﬁer was implemented in an active learning environment for classiﬁcation of the titles and abstracts. Our outcome variable was the percentage of studies that can be excluded after 95% of relevant studies have been identiﬁed by the classiﬁer (work saved over sampling: WSS@95%). In simulation runs of the entire screening process (controlling for classiﬁer initiation and the frequency of classiﬁer updating), we obt

FAQs

Why are systematic reviews for PI-RADS diagnostic performance resource-intensive?

They require substantial manual screening of large numbers of titles and abstracts as new evidence accumulates each year.

What machine learning approach was used to reduce screening workload?

A naïve Bayesian classifier was implemented in an active learning environment to classify titles and abstracts.

How much work was saved by the proposed framework at 95% relevant-study identification?

Simulations yielded a WSS@95% of 28% (standard error of the mean ±0.1%), translating to a meaningful reduction in manual screening effort.

Aug 5, 2026

Upload by Anda