Improving prognostic accuracy in lung transplantation using unique features of isolated human lung radiographs

Bonnie T Chao^{1

2}, Andrew T Sage^{1

3

4}, Micheal C McInnis^{5

6}, Jun Ma^{7

8

9}, Micah Grubert Van Iderstine¹, Xuanzi Zhou^{1

3}, Jerome Valero¹, Marcelo Cypel^{1

3

4}, Mingyao Liu^{1

3

4}, Bo Wang^{7

8

9

10

11}, Shaf Keshavjee^{12

13

14

15

16}

Affiliations

¹ Latner Thoracic Research Laboratories, Toronto General Hospital Research Institute, University Health Network, Toronto, ON, Canada.
² Institute of Biomedical Engineering, University of Toronto, Toronto, ON, Canada.
³ Institute of Medical Science, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada.
⁴ Department of Surgery, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada.
⁵ University Medical Imaging Toronto, Toronto General Hospital, University Health Network, Toronto, ON, Canada.
⁶ Department of Medical Imaging, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada.
⁷ Peter Munk Cardiac Centre, University Health Network, Toronto, ON, Canada.
⁸ Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada.
⁹ Vector Institute, University of Toronto, Toronto, ON, Canada.
¹⁰ AI Hub, University Health Network, Toronto, ON, Canada.
¹¹ Department of Computer Science, University of Toronto, Toronto, ON, Canada.
¹² Latner Thoracic Research Laboratories, Toronto General Hospital Research Institute, University Health Network, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹³ Institute of Biomedical Engineering, University of Toronto, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹⁴ Institute of Medical Science, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹⁵ Department of Surgery, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹⁶ AI Hub, University Health Network, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.

PMID: 39363013
PMCID: PMC11452202
DOI: 10.1038/s41746-024-01260-z

Improving prognostic accuracy in lung transplantation using unique features of isolated human lung radiographs

Bonnie T Chao et al. NPJ Digit Med. 2024.

. 2024 Oct 3;7(1):272.

doi: 10.1038/s41746-024-01260-z.

Authors

Affiliations

¹ Latner Thoracic Research Laboratories, Toronto General Hospital Research Institute, University Health Network, Toronto, ON, Canada.
² Institute of Biomedical Engineering, University of Toronto, Toronto, ON, Canada.
³ Institute of Medical Science, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada.
⁴ Department of Surgery, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada.
⁵ University Medical Imaging Toronto, Toronto General Hospital, University Health Network, Toronto, ON, Canada.
⁶ Department of Medical Imaging, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada.
⁷ Peter Munk Cardiac Centre, University Health Network, Toronto, ON, Canada.
⁸ Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada.
⁹ Vector Institute, University of Toronto, Toronto, ON, Canada.
¹⁰ AI Hub, University Health Network, Toronto, ON, Canada.
¹¹ Department of Computer Science, University of Toronto, Toronto, ON, Canada.
¹² Latner Thoracic Research Laboratories, Toronto General Hospital Research Institute, University Health Network, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹³ Institute of Biomedical Engineering, University of Toronto, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹⁴ Institute of Medical Science, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹⁵ Department of Surgery, Temerty Faculty of Medicine, University of Toronto, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.
¹⁶ AI Hub, University Health Network, Toronto, ON, Canada. shaf.keshavjee@uhn.ca.

PMID: 39363013
PMCID: PMC11452202
DOI: 10.1038/s41746-024-01260-z

Abstract

Ex vivo lung perfusion (EVLP) enables advanced assessment of human lungs for transplant suitability. We developed a convolutional neural network (CNN)-based approach to analyze the largest cohort of isolated lung radiographs to date. CNNs were trained to process 1300 longitudinal radiographs from n = 650 clinical EVLP cases. Latent features were transformed into principal components (PC) and correlated with known radiographic findings. PCs were combined with physiological data to classify clinical outcomes: (1) recipient time to extubation of <72 h, (2) ≥ 72 h, and (3) lungs unsuitable for transplantation. The top PC was significantly correlated with infiltration (Spearman R: 0·72, p < 0·0001), and adding radiographic PCs significantly improved the discrimination for clinical outcomes (Accuracy: 73 vs 78%, p = 0·014). CNN-derived radiographic lung features therefore add substantial value to the current assessments. This approach can be adopted by EVLP centers worldwide to harness radiographic information without requiring real-time radiological expertise.

PubMed Disclaimer

Conflict of interest statement

Keshavjee serves as Chief Medical Officer of Traferox Technologies and reports personal fees from Lung Bioengineering, outside the submitted work. A.T.S., M.C.M., M.C., B.W., and S.K. declare ongoing patent applications with University Health Network (No.US63/314,930 & No. US63/315,042) related to machine learning models for ex vivo perfusion used in this study. The investigators fully adhere to policies at University Health Network that ensure academic integrity and management of potential conflicts of interest. B.T.C., J.M., M.G.V.I., X.Z., J.V., M.L. declare no competing interests.

Figures

**Fig. 1. Class activation mapping for isolated lung images.**
Important regions for different predictions are visualized using Grad-CAM. Darker blue colors indicate regions significant for a particular class prediction.

**Fig. 2. Correlation heat map depicting the associations between PCs from image features and radiographic scores.**
Features analyzed by computer vision are consistent with manually labeled radiographic abnormalities. The shaded bar on the right represents the Spearman correlation coefficients from +1.0 to −1.0. The shade and size of each circle in the grid indicate the extent to which each PC correlated with each radiographic abnormality. (PC principal component, *p < 0.05, **p < 0.01, ***p < 0.001).

**Fig. 3. CNN-based image processing pipeline.**
1 h and 3 h X-ray images are simultaneously processed by convolutional layers, and one single classification is performed for both images of the same lungs. (CNN convolutional neural network).

See this image and copyright information in PMC

References

1. Ali, A. & Cypel, M. Ex-vivo lung perfusion and ventilation: where to from here? Curr. Opin. Organ Transplant.24, 297–304 (2019). - PubMed
1. Tian, D. et al. Outcomes of marginal donors for lung transplantation after ex vivo lung perfusion: a systematic review and meta-analysis. J. Thorac. Cardiovasc. Surg.159, 720–730.e6 (2020). - PubMed
1. Keshavjee, S. Human organ repair centers: Fact or fiction? JTCVS Open3, 164–168 (2020). - PMC - PubMed
1. Cypel, M. et al. Normothermic ex vivo lung perfusion in clinical lung transplantation. N. Engl. J. Med. 10, 1431–1440 (2011). - PubMed
1. Yeung, J. C. et al. Physiologic assessment of the ex vivo donor lung for transplantation. J. Heart Lung Transplant.31, 1120–1126 (2012). - PubMed

LinkOut - more resources

Full Text Sources
- Nature Publishing Group
- PubMed Central

Save citation to file

Email citation

Add to Collections

Add to My Bibliography

Your saved search

Create a file for external citation management software

Your RSS Feed

Improving prognostic accuracy in lung transplantation using unique features of isolated human lung radiographs

Affiliations

Improving prognostic accuracy in lung transplantation using unique features of isolated human lung radiographs

Authors

Affiliations

Abstract

Conflict of interest statement

Figures

References

LinkOut - more resources

Full Text Sources