Behavioral Analysis of Turbulent Exhale Flows
Shane Transue (PhD student), Sayed Mohsin Reza (PhD student), Ann C. Halbower (MD Pulmonary Medicine) and Min Choi (CMTC Co-Director)
This research introduces a new set of non-contact vision-based methods for respiratory analysis, including direct CO2 imaging using thermal cameras and monitoring chest deformations. These methods allow patients to breath normally without interference from contact-based methods that use respiratory tubes or other contact devices.
Problem that Inspired Research:
Current contact-based methods for respiratory analysis are not well suited for extended monitoring periods and can be uncomfortable to use, typically requiring the patient to breathe through one or more tubes. While improvements in contact methods can be addressed for comfort, they still remain inaccurate because they modify natural breathing behaviors. Typically, patients do not breathe through a tube during normal breathing, therefore introducing this form of contact monitoring fundamentally alters their respiratory behavior.
Objective / Proposed Solution:
Our objective is to eliminate the artificial interference of normal breathing behaviors introduced through the use of tubes to monitory respiratory functions while obtaining the same metrics required by existing medical diagnosis techniques. This includes using vision-based methods for obtaining respiratory rate, tidal volume, and exhale behavior.
Greatest Challenge to Overcome:
The development of non-contact methods is inherently difficult; no direct measurements of area, volume, flow, or pressure can be obtained. Therefore, every method introduced within the non-contact domain must find alternative techniques for obtaining these standard metrics that are normally easy to obtain in contact methods. The primary methods of using CO2 imaging include how to handle background heat sources, how to precisely extract exhale CO2 signatures from thermal images, and how to reconstruct volumetric models from 2D image sequences.
Benefits of Research:
This research has a significant impact on any patient that requires routine or extended respiratory monitoring sessions for clinical diagnosis or ongoing treatment evaluations. The core contribution of this method is to allow a patient to breathe naturally (without using tubes or other devices connected to them) during the monitoring process. This makes the environment less stressful and much less strenuous on the patient. Additionally, without requiring the patient to breathe through a tube, this method is much more comfortable and much more applicable to sleep studies.
The application of this research is within the medical domain. Since we obtain high-resolution CO2 images of a patient’s respiratory process, this method is less adaptable for at-home care, but provides professional clinics with a new method for ensuring patient care during monitoring procedures. The application of these techniques will be employed within a hospital environment for patient testing.
Innovations to Media and Technology:
This research contributes to medical imaging technology through the development of new algorithms and computer vision techniques that can be used to efficiently and accurately track respiratory health. Using these methods, we can gain further insight into new medical diagnosis channels that did not previously exist including exhale behavioral analysis and lung functionality testing through non-contact methods.
Cutting-edge Technology Being Used:
This research utilizes cutting edge technology to generate high-resolution thermal images tuned to the specific wavelength of CO2. The visualization of this gas at room temperature requires a continuously cooled thermal camera (sensor temperature ~ -200[C]) with extremely tight machine tolerances. To obtain this device, we worked with FLIR, the global leader in thermal camera development, to build a custom CO2 imaging camera specifically designed for respiratory analysis.
This research combines computer vision, graphics, and simulation with medical domain experts to develop new and meaningful contributions to non-contact respiratory analysis.
Awarded a core science grant by the National Science Foundation (NSF) grant no. 1602428; 2018 Biomedical and Health Informatics Conference Best Paper Award (~14% oral acceptance rate).
This research is interdisciplinary project between Computer Science (Dr. Min-Hyung Choi) and the University of Colorado Medical School and Children’s Hospital Colorado (Dr. Ann Halbower, MD of Pulmonary Medicine).