TY - JOUR
T1 - Patient Reidentification from Chest Radiographs
T2 - AInterpretable Deep Metric Learning Approach and Its Applications
AU - MacPherson, Matthew
AU - Hutchinson, Charles
AU - Horst, Carolyn
AU - Goh, Vicky
AU - Montana, Giovanni
N1 - Funding Information:
From the Mathematics Institute (M.S.M.), Warwick Medical School (C.E.H.), Department of Statistics (G.M.), and Warwick Manufacturing Group (G.M.), University of Warwick, Coventry CV4 7AL, United Kingdom; Department of Radiology, University Hospitals Coventry and Warwickshire NHS Trust, Coventry, United Kingdom (C.E.H.); School of Biomedical Engineering & Imaging Sciences, King’s College London, London, United Kingdom (C.H., V.G.); Department of Radiology, Guy’s and St Thomas’ NHS Foundation Trust, London, United Kingdom (V.G.); and Alan Turing Institute, London, United Kingdom (G.M.). Received January 25, 2023; revision requested March 14; revision received August 24; accepted September 7. Address correspondence to G.M. (email: [email protected]). Supported by the Wellcome Trust (research grant) and by the Engineering and Physical Sciences Research Council (student funding). Conflicts of interest are listed at the end of this article.
Publisher Copyright:
© RSNA, 2023.
PY - 2023/11
Y1 - 2023/11
N2 - Purpose: To train an explainable deep learning model for patient reidentification in chest radiograph datasets and assess changes in model-perceived patient identity as a marker for emerging radiologic abnormalities in longitudinal image sets. Materials and Methods: This retrospective study used a set of 1 094 537 frontal chest radiographs and free-text reports from 259 152 patients obtained from six hospitals between 2006 and 2019, with validation on the public ChestX-ray14, CheXpert, and MIMIC-CXR datasets. A deep learning model was trained for patient reidentification and assessed on patient identity confirmation, retrieval of patient images from a database based on a query image, and radiologic abnormality prediction in longitudinal image sets. The representa-tion learned was incorporated into a generative adversarial network, allowing visual explanations of the relevant features. Performance was evaluated with sensitivity, specificity, F1 score, Precision at 1, R-Precision, and area under the receiver operating characteristic curve (AUC) for normal and abnormal prediction. Results: Patient reidentification was achieved with a mean F1 score of 0.996 ± 0.001 (2 SD) on the internal test set (26 152 patients) and F1 scores of 0.947–0.993 on the external test data. Database retrieval yielded a mean Precision at 1 score of 0.976 ± 0.005 at 299 × 299 resolution on the internal test set and Precision at 1 scores between 0.868 and 0.950 on the external datasets. Patient sex, age, weight, and other factors were identified as key model features. The model achieved an AUC of 0.73 ± 0.01 for abnormality prediction versus an AUC of 0.58 ± 0.01 for age prediction error. Conclusion: The image features used by a deep learning patient reidentification model for chest radiographs corresponded to intui-tive human-interpretable characteristics, and changes in these identifying features over time may act as markers for an emerging abnormality.
AB - Purpose: To train an explainable deep learning model for patient reidentification in chest radiograph datasets and assess changes in model-perceived patient identity as a marker for emerging radiologic abnormalities in longitudinal image sets. Materials and Methods: This retrospective study used a set of 1 094 537 frontal chest radiographs and free-text reports from 259 152 patients obtained from six hospitals between 2006 and 2019, with validation on the public ChestX-ray14, CheXpert, and MIMIC-CXR datasets. A deep learning model was trained for patient reidentification and assessed on patient identity confirmation, retrieval of patient images from a database based on a query image, and radiologic abnormality prediction in longitudinal image sets. The representa-tion learned was incorporated into a generative adversarial network, allowing visual explanations of the relevant features. Performance was evaluated with sensitivity, specificity, F1 score, Precision at 1, R-Precision, and area under the receiver operating characteristic curve (AUC) for normal and abnormal prediction. Results: Patient reidentification was achieved with a mean F1 score of 0.996 ± 0.001 (2 SD) on the internal test set (26 152 patients) and F1 scores of 0.947–0.993 on the external test data. Database retrieval yielded a mean Precision at 1 score of 0.976 ± 0.005 at 299 × 299 resolution on the internal test set and Precision at 1 scores between 0.868 and 0.950 on the external datasets. Patient sex, age, weight, and other factors were identified as key model features. The model achieved an AUC of 0.73 ± 0.01 for abnormality prediction versus an AUC of 0.58 ± 0.01 for age prediction error. Conclusion: The image features used by a deep learning patient reidentification model for chest radiographs corresponded to intui-tive human-interpretable characteristics, and changes in these identifying features over time may act as markers for an emerging abnormality.
UR - http://www.scopus.com/inward/record.url?scp=85177859186&partnerID=8YFLogxK
U2 - 10.1148/ryai.230019
DO - 10.1148/ryai.230019
M3 - Article
SN - 2638-6100
VL - 5
JO - Radiology: Artificial intelligence
JF - Radiology: Artificial intelligence
IS - 6
M1 - e230019
ER -