Chronic obstructive pulmonary disease (COPD) is one of the most common respiratory conditions that affects more than 22 million Americans; it has climbed to be the third leading cause of death and accounts for $32 billion in associated healthcare cost.

https://www.youtube.com/watch?v=lLkXqGk8bIM

COPD remains underdiagnosed, and its associated comorbidities include cardiovascular disease, musculoskeletal disorders, and lung cancer. Proper population management of COPD patients rests on the ability to define who is at a higher risk based on their disease prognosis and underlying comorbidities.

Chest Computed Tomography (CT) and X-ray (CXR) imaging provide a unique snapshot of patients’ health status that includes three major systems: heart, lung and skeletal muscle. I will present the latest advancements that my group has developed in image-based artificial intelligence approaches to providing risk assessment of patients with COPD. Our risk models are based on deep learning approaches to quantify emphysema, air trapping, heart size, body composition and bone mineral density from both chest CT and CXR images. We have also developed prognostication models based on canonical landmark views of a CT scan that can stage and prognosticate acute respiratory event and death in COPD.

With annual utilization rates of CT and CXR of 400 and 900 exams respectively per 1,000 beneficiaries just in the Medicare population alone, the deployment of these tools to mine standardized Picture Archiving and Communications Systems (PACS) can enable healthcare analytic companies with population management solutions based solely on imaging information.

For more information about Dr. San José Estépar’s research, please contact Partners HealthCare Innovation by clicking here.

Figure 1: Response of the mortality prediction convolutional neural network for a test subject.

Figure 2: Quantification of Emphysema from X-Ray images and corresponding network activations.