Structural and Funcional Vascular Phenotyping in Smokers with Pulmonary Hypertension using Artificial Intelligence - Project Summary Pulmonary hypertension (PH) is a heterogeneous condition that occurs most frequently as complicating comorbidity of very common disease in the western world, including COPD, heart failure, and blood clots in the lung. In patients with PH secondary to COPD (Group III) mortality risk is about twice as high than in patients with COPD and normal pulmonary artery pressure. Medications developed successfully for Group I disease have not shown much utility in Group III PH. Additionally, non-invasive screening and early detection remain challenging. For this reason, etiology and evolution of PH are still poorly understood and the response to potential interventions is very difficult to measure. In recent years, radiologists have been making observation of the loss of distal vasculature (pruning), increased vascular “tortuosity”, and proximal vascular dilation in patients with PH. In this career development proposal, we hypothesize that the great power and recent advances in artificial intelligence can be leveraged to identify new functional and structural markers that define the presence of PH in smokers from non-contrast CT images. These CT-based features will significantly improve our ability to define the main structural and functional effects of PH in COPD patients and assess pathological conditions, leading to a better understanding of the genetic and biological underpinnings of the disease. Aim 1 of this application involves refining our algorithm for vascular morphology assessment and applying it to investigate how the pulmonary vasculature is affected in presence of PH Group III. Five markers of presence of disease will be explored. The goal of Aim 2 is to develop an innovative approach to assess the level of pulmonary perfusion from non-contrast non cardiac gated CT images and to apply it to explore new functional markers of disease. Both in Aims 1 and 2 association with biological outcomes will be performed. Finally, in Aim 3 we will validate the biological relevance of the endpoints determined in Aims 1 and 2 by performing common and rare variant association studies. Preliminary data obtained with our published methods show promising results. We show that using artificial intelligence we can effectively segment pulmonary vessels, separate arteries and veins, and measure small vessels with an accuracy which outperforms state-of-the-art methods. Additionally, we developed an innovative technique that synthesized perfusion maps from non-contrast CT to assess perfusion defects in PH patients. Together, the research proposed in the aims of this award will take full advantage of the comprehensive dataset available through the COPDGene study. The execution of the aims in this proposal will be possible through active collaboration with Dr. Raul San Jose Estepar, Ph.D. as the mentor and an outstanding Advisory Committee including renowned leaders in the fields of medical image analysis, translational research, quantitative imaging in pulmonary vascular disease, and the genetic epidemiology of COPD.
