A Radiomic-Based Machine Learning System to Diagnose Age-Related Macular Degeneration from Ultra-Widefield Fundus Retinography - Article

Aug 2, 2026

Radiomics is leveraged to build an explainable AI system for ultra-widefield fundus retinography, targeting early Age-related Macular Degeneration (AMD) detection and patient risk stratification. The approach integrates machine learning radiomics with a deep learning macular detector, extracting intensity and texture patterns in the macular region. A retrospective dataset of 226 annotated UWF-FRTs from two centers supports model training and testing. External evaluation reports 93% sensitivity and 74% specificity, with features linked to drusen and pigmentary abnormalities. Human-operator comparison yields a Cohen k of 0.79, indicating substantial concordance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Problem background and clinical significance

Methods

Radiomics with explainable feature extraction

Machine learning and deep learning components

Data and Evaluation

Dataset collection and annotation

External testing performance metrics

Results

Sensitivity, specificity, and concordance with human annotations

diagnostics Article A Radiomic-Based Machine Learning System to Diagnose Age-Related Macular Degeneration from Ultra-Wideﬁeld Fundus Retinography Matteo Interlenghi 1,†, Giancarlo Sborgia 2,†, Alessandro Venturi 1, Rodolfo Sardone 3,4, Valentina Pastore 2, Giacomo Boscia 2, Luca Landini 2, Giacomo Scotti 2, Alfredo Niro 5, Federico Moscara 2, Luca Bandi 1, Christian Salvatore 1,6, * and Isabella Castiglioni 7 Citation: Interlenghi, M.; Sborgia, G.; Venturi, A.; Sardone, R.; Pastore, V.; Boscia, G.; Landini, L.; Scotti, G.; Niro, A.; Moscara, F.; et al. A RadiomicBased Machine Learning System to Diagnose Age-Related Macular Degeneration from Ultra-Wideﬁeld Fundus Retinography. Diagnostics 2023, 13, 2965. [https://doi.org/](https://doi.org/)[ ](https://doi.org/)[10.3390/diagnostics13182965](10.3390/diagnostics13182965) Academic Editor: Jae-Ho Han Received: 28 July 2023 Revised: 4 September 2023 Accepted: 13 September 2023 Published: 15 September 2023 Copyright: © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ([https://](https://)[ ](https://)[creativecommons.org/licenses/by/](creativecommons.org/licenses/by/)[ ](creativecommons.org/licenses/by/)[4.0/](4.0/)) . 1 DeepTrace Technologies S.R.L., 20122 Milan, Italy; [interlenghi@deeptracetech.com](interlenghi@deeptracetech.com) (M.I.); [venturi@deeptracetech.com](venturi@deeptracetech.com) (A.V.); [bandi@deeptracetech.com](bandi@deeptracetech.com) (L.B.) 2 Department of Medical Science, Neuroscience and Sense Organs, Eye Clinic, University of Bari Aldo Moro, 70121 Bari, Italy; [gcsborgia@hotmail.it](gcsborgia@hotmail.it) (G.S.); [valentinapastore@hotmail.it](valentinapastore@hotmail.it) (V.P.); bosciagiacomo@gmail.com (G.B.); [lucalandi4@gmail.com](lucalandi4@gmail.com) (L.L.); [giacomo.scotti@tiscali.it](giacomo.scotti@tiscali.it) (G.S.); [federico.moscara@gmail.com](federico.moscara@gmail.com) (F.M.) 3 National Institute of Gastroenterology—IRCCS “Saverio de Bellis”, 70013 Castellana Grotte, Italy; [rodolfo.sardone@irccsdebellis.it](rodolfo.sardone@irccsdebellis.it) 4 Unit of Statistics and Epidemiology, Local Healthcare Authority of Taranto, 74121 Taranto, Italy 5 Eye Clinic, Hospital “SS. Annunziata”, ASL Taranto, 74121 Taranto, Italy; [alfred.nir@tiscali.it](alfred.nir@tiscali.it) 6 Department of Science, Technology and Society, University School for Advanced Studies IUSS Pavia, 27100 Pavia, Italy 7 Department of Physics “Giuseppe Occhialini”, University of Milan-Bicocca, 20126 Milan, Italy; [isabella.castiglioni@unimib.it](isabella.castiglioni@unimib.it) * Correspondence: salvatore@deeptracetech.com or [christian.salvatore@iusspavia.it](christian.salvatore@iusspavia.it)[ ](christian.salvatore@iusspavia.it)† These authors contributed equally to this work. Abstract: The present study was conducted to investigate the potential of radiomics to develop an explainable AI-based system to be applied to ultra-wideﬁeld fundus retinographies (UWF-FRTs) with the objective of predicting the presence of the early signs of Age-related Macular Degeneration (AMD) and stratifying subjects with low-versus high-risk of AMD. The ultimate aim was to provide clinicians with an automatic classiﬁer and a signature of objective quantitative image biomarkers of AMD. The use of Machine Learning (ML) and radiomics was based on intensity and texture analysis in the macular region, detected by a Deep Learning (DL)-based macular detector. Two-hundred and twenty six UWF-FRTs were retrospectively collected from two centres and manually annotated to train and test the algorithms. Notably, the combination of the ML-based radiomics model and the DL-based macular detector reported 93% sensitivity and 74% speciﬁcity when applied to the data of the centre used for external testing, capturing explainable features associated with drusen

FAQs

What does the radiomic-based system aim to achieve for AMD patients?

The system predicts early signs of Age-related Macular Degeneration and stratifies subjects by low- versus high-risk. It also provides clinicians with an automatic classifier and quantitative image biomarker signature.

How are ultra-widefield fundus retinographies processed in the proposed method?

The method combines intensity and texture analysis in the macular region with a deep learning-based macular detector. Radiomics features are extracted to support an explainable machine learning pipeline.

What performance did the system show during external testing and agreement with human annotations?

On external testing, the combined model achieved 93% sensitivity and 74% specificity. Compared with human operator annotations, the system obtained a Cohen k of 0.79, indicating substantial concordance.

Aug 2, 2026

Upload by Lucas Martin