Computed tomography-based machine learning for donor lung screening before transplantation

BACKGROUND

Assessment and selection of donor lungs remain 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 computed tomography (CT) images, we investigated the use of a CT-based machine learning algorithm for screening donor lungs before 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 before 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) before 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 intensive care unit and were at 19 times higher risk of developing chronic lung allograft dysfunction within 2 years posttransplant.

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 posttransplant complications.

The hemodynamic interplay between pulmonary ischemia-reperfusion injury and right ventricular function in lung transplantation: a translational porcine model

Lung transplantation (LTx) is a challenging procedure. Following the process of ischemia-reperfusion injury, the transplanted pulmonary graft might become severely damaged, resulting in primary graft dysfunction. In addition, during the intraoperative window, the right ventricle (RV) is at risk of acute failure. The interaction of right ventricular function with lung injury is, however, poorly understood. We aimed to address this interaction in a translational porcine model of pulmonary ischemia-reperfusion injury. Advanced pulmonary and hemodynamic assessment was used, including right ventricular pressure-volume loop analysis. The acute model was based on clamping and unclamping of the left lung hilus, respecting the different hemodynamic phases of a clinical lung transplantation. We found that forcing entire right ventricular cardiac output through a lung suffering from ischemia-reperfusion injury increased afterload (pulmonary vascular resistance from baseline to end experiment P < 0.0001) and induced right ventricular failure (RVF) in 5/9 animals. Notably, we identified different compensation patterns in failing versus nonfailing ventricles (arterial elastance P = 0.0008; stroke volume P < 0.0001). Furthermore, increased vascular pressure and flow produced by the right ventricle resulted in higher pulmonary injury, as measured by ex vivo CT density (correlation: pressure r = 0.8; flow r = 0.85). Finally, RV ischemia as measured by troponin-T was negatively correlated with pulmonary injury (r = -0.76); however, troponin-T values did not determine RVF in all animals. In conclusion, we demonstrate a delicate balance between development of pulmonary ischemia-reperfusion injury and right ventricular function during lung transplantation. Furthermore, we provide a physiological basis for potential benefit of extracorporeal life support technology.NEW & NOTEWORTHY In contrast to the abundant literature of mechanical pulmonary artery clamping to increase right ventricular afterload, we developed a model adding a biological factor of pulmonary ischemia-reperfusion injury. We did not only focus on the right ventricular behavior, but also on the interaction with the injured lung. We are the first to describe this interaction while addressing the hemodynamic intraoperative phases of clinical lung transplantation.

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

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.

Lung transplantation for acute respiratory distress syndrome: A multicenter experience

Acute respiratory distress syndrome (ARDS) is a rapidly progressive lung disease with a high mortality rate. Although lung transplantation (LTx) is a well-established treatment for a variety of chronic pulmonary diseases, LTx for acute lung failure (due to ARDS) remains controversial. We reviewed posttransplant outcome of ARDS patients from three high-volume European transplant centers. Demographics and clinical data were collected and analyzed. Viral infection was the main reason for ARDS (n = 7/13, 53.8%). All patients were admitted to ICU and required mechanical ventilation, 11/13 were supported with ECMO at the time of listing. They were granted a median LAS of 76 (IQR 50-85) and waited for a median of 3 days (IQR 1.5-14). Postoperatively, median length of mechanical ventilation was 33 days (IQR 17-52.5), median length of ICU and hospital stay were 39 days (IQR 19.5-58.5) and 54 days (IQR 43.5-127). Prolongation of peripheral postoperative ECMO was required in 7/13 (53.8%) patients with a median duration of 2 days (IQR 2-7). 30-day mortality was 7.7%, 1 and 5-year survival rates were calculated as 71.6% and 54.2%, respectively. Given the lack of alternative treatment options, the herein presented results support the concept of offering live-saving LTx to carefully selected ARDS patients.

Early protein expression profile in bronchoalveolar lavage fluid and clinical outcomes in primary graft dysfunction after lung transplantation

OBJECTIVES

Primary graft dysfunction (PGD) remains a major post-transplant complication and is associated with increased morbidity and mortality. Mechanisms evoking PGD are not completely clear, but inflammation plays a central role. We investigated the association between PGD and inflammatory proteins present in immediate postoperative bronchoalveolar lavage.

METHODS

All double-lung recipients transplanted at our institution from 2002 to 2018 were included in our study. We retrospectively selected 80 consecutive lung transplant recipients with different PGD grades (n = 20 for each PGD grades 0-1 to 2-3). In bronchoalveolar lavage performed within the first 24 h after donor aortic cross-clamping following lung transplantation, concentrations of 30 cytokines, chemokines and growth factors were assessed by enzyme-linked immunosorbent assay (ELISA) and correlated with donor and recipient demographics and outcomes. For analysis, 2 groups were defined: 'mild' PGD (grade 0-1) and 'severe' PGD (grades 2-3).

RESULTS

Significant differences between mild and severe PGD were found in 8 biomarkers [interleukin (IL)-6, IL-10, IL-13, eotaxin, granulocyte colony-stimulating factor, interferon γ, macrophage inflammatory protein 1α, surfactant protein D (SP-D); P < 0.05]. Increased IL-10 and IL-13, but none of the other proteins, were associated with short-term outcome (longer time to extubation; P = 0.005 and P < 0.0001; increased intensive care unit stay; P = 0.012 and P < 0.0001; and hospital stay; P = 0.041 and P = 0.002). There were no significant differences in donor and recipient characteristics between the groups.

CONCLUSIONS

Expression profiles of key inflammatory mediators in bronchoalveolar lavage fluid differed significantly between lung transplant recipients with severe versus mild PGD and correlated with clinical outcome variables. Further research should focus on the early mechanisms leading to PGD.

Histopathologic and radiologic assessment of nontransplanted donor lungs

Donor organ shortage results in significant waiting list mortality. Donor lung assessment is currently based on donors' history, gas exchange, chest X-ray, bronchoscopy findings, and ultimately in situ inspection but remains subjective. We correlated histopathology and radiology in nontransplanted donor lungs with the clinical indications to decline the offered organ. Sixty-two donor lungs, not used for transplantation (2010-2019), were procured, air-inflated, frozen, scanned with computed tomography, systematically sampled, and histologically and radiologically assessed. Thirty-nine (63%) lungs were declined for allograft-related reasons. In 13/39 (33%) lungs, histology could not confirm the reason for decline, in an additional 8/39 (21%) lungs, histologic abnormalities were only considered mild. In 16/39 (41%) lungs, radiology could not confirm the reason for decline. Twenty-three (37%) donor lungs were not transplanted due to extrapulmonary causes, of which three (13%) lungs displayed severe histologic abnormalities (pneumonia, n = 2; emphysema, n = 1), in addition to mild emphysema in 9 (39%) lungs and minor bronchopneumonia in 1 (4%). Radiology revealed ground-glass opacities in 8/23 (35%) and emphysema in 4/23 (17%) lungs. Histopathologic and radiologic assessment of nontransplanted donor lungs revealed substantial discrepancy with the clinical reason for decline. Optimization of donor lung assessment is necessary to improve current organ acceptance rates.

The resident's point of view in the learning curve of thymic MIS: why should I learn it?

Minimally invasive surgery (MIS) in thoracic surgery became quite popular during the last years. The aim of introducing and performing more MIS is to reduce surgical trauma, pain and complications in patients. Training in MIS increases operative time and thus cost in theatre but thus improves with experience. For a resident, the cases should be well selected with experienced supervision in a suitable setting with supporting staff and optimal instruments. Understanding the anatomy of the lung, using simulators, and attending workshops makes the learning curve shorter.