Deep-learning algorithm diagnostic support for usual interstitial pneumonia pattern recognition in fibrotic interstitial lung disease

Caitlin C Fermoyle; John A Mackintosh; Vidya Navaratnam; Samantha J Ellis; Wendy A Cooper; Nicole S L Goh; Yuben Moodley; Paul N Reynolds; Christopher J Zappala; Peter Hopkins; Ian N Glaspole; Tamera J Corte; Simon L F Walsh; SOFIA Project Consortium

doi:10.1002/resp.70246

Deep-learning algorithm diagnostic support for usual interstitial pneumonia pattern recognition in fibrotic interstitial lung disease

Caitlin C Fermoyle
, John A Mackintosh
, Vidya Navaratnam
, Samantha J Ellis
, Wendy A Cooper
, Nicole S L Goh
, Yuben Moodley
, Paul N Reynolds
, Christopher J Zappala
, Peter Hopkins
, Ian N Glaspole
, Tamera J Corte
, Simon L F Walsh^*
, SOFIA Project Consortium

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

Abstract

Background and objective: High resolution computed tomography (HRCT) scan diagnostic classification for usual interstitial pneumonia (UIP) plays a critical role in therapeutic decision-making and clinical trial eligibility for interstitial lung disease (ILD) patients, but is subject to variability. A deep learning algorithm, the Systematic Objective Fibrotic Imaging Analysis Algorithm (SOFIA), has been validated to assist classification of HRCTs based on current guidelines. In this study, we evaluate the impact of SOFIA on inter-observer agreement for UIP classification and prognostic accuracy of clinicians' assessment of ILD HRCTs.

Methods: Radiologists and pulmonologists (reviewers) were invited to evaluate 203 HRCTs from a national fibrotic ILD registry, scoring each of four UIP categories (definite UIP, probable UIP, indeterminate, or alternative diagnosis). SOFIA outputs were then provided, and reviewers were able to reevaluate their scores. Changes in interobserver agreement for UIP classification and prognostic accuracy were calculated.

Results: Three hundred twelve reviewers (120 radiologists, 192 pulmonologists) from 49 countries evaluated 203 HRCT scans. Following SOFIA, inter-observer diagnostic agreement improved for definite UIP from moderate to good (ICC_pre= 0.54[0.50-0.60]; ICC_post = 0.70[0.66-0.74]), and for probable UIP from fair to moderate (ICC_pre = 0.30[0.27-0.35]; ICC_post = 0.53[0.49-0.58]). Following SOFIA, there was improved prognostic accuracy for reviewers' definite UIP, probable UIP, and indeterminate scores (significant change in c-index), and the proportion of reviewers whose probable UIP scores were significantly predictive of transplant-free survival increased by 42%.

Conclusion: Providing SOFIA algorithm output to clinicians reviewing HRCT scans improved diagnostic agreement and prognostic accuracy for fibrotic ILD. SOFIA may be a useful automated assistive tool to support improved diagnostic consistency.

Original language	English
Pages (from-to)	1-10
Number of pages	10
Journal	Respirology
Volume	Early View
Early online date	1 Apr 2026
DOIs	https://doi.org/10.1002/resp.70246
Publication status	E-pub ahead of print - 1 Apr 2026

Keywords

Deep learning
Disease progression
Idiopathic pulmonary fibrosis
Radiology
Usual interstitial pneumonia

Access to Document

10.1002/resp.70246Licence: CC BY

Fermoyle-et-al-2026- Deep‐learning-algorithm-diagnostic-support-Respirology-70246-CCBY
Copyright © 2026 The Author(s). Respirology published by John Wiley & Sons Australia, Ltd on behalf of Asian Pacific Society of Respirology. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Final published version, 3.25 MBLicence: CC BY

Handle.net

Persistent link to portal record

Cite this

Fermoyle, C. C., Mackintosh, J. A., Navaratnam, V., Ellis, S. J., Cooper, W. A., Goh, N. S. L., Moodley, Y., Reynolds, P. N., Zappala, C. J., Hopkins, P., Glaspole, I. N., Corte, T. J., Walsh, S. L. F., & SOFIA Project Consortium (2026). Deep-learning algorithm diagnostic support for usual interstitial pneumonia pattern recognition in fibrotic interstitial lung disease. Respirology, Early View, 1-10. Advance online publication. https://doi.org/10.1002/resp.70246

@article{68b9681f04ab4848a25c262b6d46a671,

title = "Deep-learning algorithm diagnostic support for usual interstitial pneumonia pattern recognition in fibrotic interstitial lung disease",

abstract = "Background and objective: High resolution computed tomography (HRCT) scan diagnostic classification for usual interstitial pneumonia (UIP) plays a critical role in therapeutic decision-making and clinical trial eligibility for interstitial lung disease (ILD) patients, but is subject to variability. A deep learning algorithm, the Systematic Objective Fibrotic Imaging Analysis Algorithm (SOFIA), has been validated to assist classification of HRCTs based on current guidelines. In this study, we evaluate the impact of SOFIA on inter-observer agreement for UIP classification and prognostic accuracy of clinicians' assessment of ILD HRCTs.Methods: Radiologists and pulmonologists (reviewers) were invited to evaluate 203 HRCTs from a national fibrotic ILD registry, scoring each of four UIP categories (definite UIP, probable UIP, indeterminate, or alternative diagnosis). SOFIA outputs were then provided, and reviewers were able to reevaluate their scores. Changes in interobserver agreement for UIP classification and prognostic accuracy were calculated.Results: Three hundred twelve reviewers (120 radiologists, 192 pulmonologists) from 49 countries evaluated 203 HRCT scans. Following SOFIA, inter-observer diagnostic agreement improved for definite UIP from moderate to good (ICCpre = 0.54[0.50-0.60]; ICCpost = 0.70[0.66-0.74]), and for probable UIP from fair to moderate (ICCpre = 0.30[0.27-0.35]; ICCpost = 0.53[0.49-0.58]). Following SOFIA, there was improved prognostic accuracy for reviewers' definite UIP, probable UIP, and indeterminate scores (significant change in c-index), and the proportion of reviewers whose probable UIP scores were significantly predictive of transplant-free survival increased by 42\%.Conclusion: Providing SOFIA algorithm output to clinicians reviewing HRCT scans improved diagnostic agreement and prognostic accuracy for fibrotic ILD. SOFIA may be a useful automated assistive tool to support improved diagnostic consistency.",

keywords = "Deep learning, Disease progression, Idiopathic pulmonary fibrosis, Radiology, Usual interstitial pneumonia",

author = "Fermoyle, \{Caitlin C\} and Mackintosh, \{John A\} and Vidya Navaratnam and Ellis, \{Samantha J\} and Cooper, \{Wendy A\} and Goh, \{Nicole S L\} and Yuben Moodley and Reynolds, \{Paul N\} and Zappala, \{Christopher J\} and Peter Hopkins and Glaspole, \{Ian N\} and Corte, \{Tamera J\} and Walsh, \{Simon L F\} and \{SOFIA Project Consortium\} and Dhasmana, \{Devesh J\}",

note = "Funding: National Health and Medical Research Council. Grant Number: 2017479; Boehringer Ingelheim, Roche Products Pty. Limited; Lung Foundation Australia \& the Centre of Research Excellence in Pulmonary Fibrosis (CRE-PF; CREATE HOPE Fellowship).",

year = "2026",

month = apr,

day = "1",

doi = "10.1002/resp.70246",

language = "English",

volume = "Early View",

pages = "1--10",

journal = "Respirology",

issn = "1323-7799",

publisher = "John Wiley \& Sons, Ltd",

}

Fermoyle, CC, Mackintosh, JA, Navaratnam, V, Ellis, SJ, Cooper, WA, Goh, NSL, Moodley, Y, Reynolds, PN, Zappala, CJ, Hopkins, P, Glaspole, IN, Corte, TJ, Walsh, SLF & SOFIA Project Consortium 2026, 'Deep-learning algorithm diagnostic support for usual interstitial pneumonia pattern recognition in fibrotic interstitial lung disease', Respirology, vol. Early View, pp. 1-10. https://doi.org/10.1002/resp.70246

TY - JOUR

T1 - Deep-learning algorithm diagnostic support for usual interstitial pneumonia pattern recognition in fibrotic interstitial lung disease

AU - Fermoyle, Caitlin C

AU - Mackintosh, John A

AU - Navaratnam, Vidya

AU - Ellis, Samantha J

AU - Cooper, Wendy A

AU - Goh, Nicole S L

AU - Moodley, Yuben

AU - Reynolds, Paul N

AU - Zappala, Christopher J

AU - Hopkins, Peter

AU - Glaspole, Ian N

AU - Corte, Tamera J

AU - Walsh, Simon L F

AU - SOFIA Project Consortium

AU - Dhasmana, Devesh J

N1 - Funding: National Health and Medical Research Council. Grant Number: 2017479; Boehringer Ingelheim, Roche Products Pty. Limited; Lung Foundation Australia & the Centre of Research Excellence in Pulmonary Fibrosis (CRE-PF; CREATE HOPE Fellowship).

PY - 2026/4/1

Y1 - 2026/4/1

N2 - Background and objective: High resolution computed tomography (HRCT) scan diagnostic classification for usual interstitial pneumonia (UIP) plays a critical role in therapeutic decision-making and clinical trial eligibility for interstitial lung disease (ILD) patients, but is subject to variability. A deep learning algorithm, the Systematic Objective Fibrotic Imaging Analysis Algorithm (SOFIA), has been validated to assist classification of HRCTs based on current guidelines. In this study, we evaluate the impact of SOFIA on inter-observer agreement for UIP classification and prognostic accuracy of clinicians' assessment of ILD HRCTs.Methods: Radiologists and pulmonologists (reviewers) were invited to evaluate 203 HRCTs from a national fibrotic ILD registry, scoring each of four UIP categories (definite UIP, probable UIP, indeterminate, or alternative diagnosis). SOFIA outputs were then provided, and reviewers were able to reevaluate their scores. Changes in interobserver agreement for UIP classification and prognostic accuracy were calculated.Results: Three hundred twelve reviewers (120 radiologists, 192 pulmonologists) from 49 countries evaluated 203 HRCT scans. Following SOFIA, inter-observer diagnostic agreement improved for definite UIP from moderate to good (ICCpre = 0.54[0.50-0.60]; ICCpost = 0.70[0.66-0.74]), and for probable UIP from fair to moderate (ICCpre = 0.30[0.27-0.35]; ICCpost = 0.53[0.49-0.58]). Following SOFIA, there was improved prognostic accuracy for reviewers' definite UIP, probable UIP, and indeterminate scores (significant change in c-index), and the proportion of reviewers whose probable UIP scores were significantly predictive of transplant-free survival increased by 42%.Conclusion: Providing SOFIA algorithm output to clinicians reviewing HRCT scans improved diagnostic agreement and prognostic accuracy for fibrotic ILD. SOFIA may be a useful automated assistive tool to support improved diagnostic consistency.

AB - Background and objective: High resolution computed tomography (HRCT) scan diagnostic classification for usual interstitial pneumonia (UIP) plays a critical role in therapeutic decision-making and clinical trial eligibility for interstitial lung disease (ILD) patients, but is subject to variability. A deep learning algorithm, the Systematic Objective Fibrotic Imaging Analysis Algorithm (SOFIA), has been validated to assist classification of HRCTs based on current guidelines. In this study, we evaluate the impact of SOFIA on inter-observer agreement for UIP classification and prognostic accuracy of clinicians' assessment of ILD HRCTs.Methods: Radiologists and pulmonologists (reviewers) were invited to evaluate 203 HRCTs from a national fibrotic ILD registry, scoring each of four UIP categories (definite UIP, probable UIP, indeterminate, or alternative diagnosis). SOFIA outputs were then provided, and reviewers were able to reevaluate their scores. Changes in interobserver agreement for UIP classification and prognostic accuracy were calculated.Results: Three hundred twelve reviewers (120 radiologists, 192 pulmonologists) from 49 countries evaluated 203 HRCT scans. Following SOFIA, inter-observer diagnostic agreement improved for definite UIP from moderate to good (ICCpre = 0.54[0.50-0.60]; ICCpost = 0.70[0.66-0.74]), and for probable UIP from fair to moderate (ICCpre = 0.30[0.27-0.35]; ICCpost = 0.53[0.49-0.58]). Following SOFIA, there was improved prognostic accuracy for reviewers' definite UIP, probable UIP, and indeterminate scores (significant change in c-index), and the proportion of reviewers whose probable UIP scores were significantly predictive of transplant-free survival increased by 42%.Conclusion: Providing SOFIA algorithm output to clinicians reviewing HRCT scans improved diagnostic agreement and prognostic accuracy for fibrotic ILD. SOFIA may be a useful automated assistive tool to support improved diagnostic consistency.

KW - Deep learning

KW - Disease progression

KW - Idiopathic pulmonary fibrosis

KW - Radiology

KW - Usual interstitial pneumonia

U2 - 10.1002/resp.70246

DO - 10.1002/resp.70246

M3 - Article

C2 - 41923409

SN - 1323-7799

VL - Early View

SP - 1

EP - 10

JO - Respirology

JF - Respirology

ER -

Deep-learning algorithm diagnostic support for usual interstitial pneumonia pattern recognition in fibrotic interstitial lung disease

Abstract

Keywords

Access to Document

Handle.net

Fingerprint

Cite this