CT-based Machine Learning for Donor Lung Screening Prior to Transplantation

Abstract

Background: Assessment and selection of donor lungs remains largely subjective and experience based. Criteria to accept or decline lungs are poorly standardized and are not compliant with the current donor pool. Using ex vivo CT images, we investigated the use of a CT-based machine learning algorithm for screening donor lungs prior to transplantation.

Methods: Clinical measures and ex-situ CT scans were collected from 100 cases as part of a prospective clinical trial. Following procurement, donor lungs were inflated, placed on ice according to routine clinical practice, and imaged using a clinical CT scanner prior to transplantation while stored in the icebox. We trained and tested a supervised machine learning method called dictionary learning , which uses CT scans and learns specific image patterns and features pertaining to each class for a classification task. The results were evaluated with donor and recipient clinical measures.

Results: Of the 100 lung pairs donated, 70 were considered acceptable for transplantation (based on standard clinical assessment) prior to CT screening and were consequently implanted. The remaining 30 pairs were screened but not transplanted. Our machine learning algorithm was able to detect pulmonary abnormalities on the CT scans. Among the patients who received donor lungs, our algorithm identified recipients who had extended stays in the ICU and were at 19 times higher risk of developing CLAD within 2 years post-transplant.

Conclusions: We have created a strategy to ex vivo screen donor lungs using a CT-based machine learning algorithm. As the use of suboptimal donor lungs rises, it is important to have in place objective techniques that will assist physicians in accurately screening donor lungs to identify recipients most at risk of post-transplant complications.

Figure 1:

Illustration of donor lung screening with computed tomography workflow. (A) Provided is an illustration of the inclusion of CT in routine donor lung screening process. Blue boxes represent standard-of-care, and red box represents CT-ML procedures. The approximate time for donor lung preparation and CT imaging is 5–10 minutes. (B) Corresponding axial, sagittal and coronal views of a CT scan from a declined donor lung (Figure 3C–D and Case 4 in Table 3).

Figure 2:

Representative CT scans with corresponding ML patch probability maps for (A and B) accepted and (C and D) declined donor lungs. The patch probabilities represent the likelihood that the lung tissue within the patch is “good” (red with probability of 1) or “bad” (blue with probability of 0). The accepted donor lung was obtained from a male, non-smoker, 45–50 years of age. The declined donor lung was obtained from a male, over 20 pack years, 65–70 years of age, found to have extensive emphysema.

Figure 3:

Kaplan-Meier plot showing potential of CT-ML strategy to predict ICU stay in lung transplant recipients (N=52). Green line and red dashed line represent agreement and disagreement, respectively, ML model to clinical decision. Lines correspond to color in confusion matrix (Supplement Figure 3). Statistical significance was determined using a logrank test.

Figure 4:

Representative CT scan with corresponding ML patch probability map from a declined donor lung identified by ML as acceptable for transplantation (False Negative). These images are from Case 10 in Table 3.