At the recent American Thoracic Society (ATS) annual meeting, Dr Michael Gregory Lester MD and Dr Fabien Maldonado MD, of Vanderbilt University Medical Centre, presented a poster ‘Utilizing a Deep Learning Radiomic Model for the Quantification of Emphysema. The poster showcases early findings in potential automated diagnosis and quantification of emphysema.
Emphysema is a type of Chronic Obstructive Pulmonary Disease (COPD), accounting for damaged alveoli, causing difficulty in breathing. If left untreated, emphysema can lead to further serious complications and previous literature suggests it may be linked to an increased risk of lung cancer 1 2. Emphysema is used in some current methods of lung cancer prediction, such as the Brock model.
Lung densitometry is the predominant method used for automatic quantification of emphysema but it comes with limitations. For example: it is susceptible to image reconstruction and the acquisition parameters of CT scans. While research studies suggest that visual assessment of emphysema in CT scans can improve lung cancer prediction, automated lung densitometry to identify emphysema has not shown an improvement in lung cancer prediction3.
An Optellum Deep Neural Network (DNN) was trained to carry out emphysema segmentation in CT scans in order to quantify the presence of emphysema and ultimately and account for emphysema in its Lung Cancer Prediction (LCP) score.
This investigation was done in collaboration with Dr Michael Gregory Lester MD, from Vanderbilt University Medical Centre (VUMC), an expert in pulmonary and critical care medicine, and Dr Fabien Maldonado MD, also from VUMC, whose expertise is in interventional pulmonology. Optellum has a long standing relationship with the Vanderbilt University Medical Centre, which stemmed from a cherished connection with the late Processor Pierre Massion, Cornelius Vanderbilt Chair of Medicine at the Vanderbilt University Medical Center and member of the Optellum medical advisory board until his untimely passing. His work continues at Vanderbilt.
In order to validate the results produced from the Optellum-trained DNN’s quantifications of emphysema, results were compared with those produced using standard lung densitometry methods.
The findings concluded that the DNN emphysema quantification showed less variability than densitometry on different reconstructions of the same CT acquisitions and, ultimately, that automated segmentations of emphysema using a trained DNN are equivalent to assessments of emphysema by the visual analysis of CT scans when they are used for nodule malignancy score in the Brock model.
Dr Michael Gregory LesterMD, of Vanderbilt University Medical Centre, commented:
“Artificial intelligence and deep learning are exceptionally powerful tools, and we have only just begun to scratch the surface of their potential contributions in COPD and emphysema. We are excited to see what new frontiers in this area the ongoing collaboration between Vanderbilt University Medical Center and Optellum will allow us to explore.”
Optellum, widely known for its world’s first AI-based early lung cancer decision support software, has a product roadmap that extends beyond cancer to other deadly diseases of the lung including Chronic Obstructive Pulmonary Disease (COPD). This research into the quantification of emphysema utilizing a Deep Learning radiomic model is a reflection of the early work the company is doing to accumulate a greater understanding of lung health and disease in order to improve patient care.
For more about new research and developments at Optellum, please subscribe to our newsletter.
 Turner, M. C., Chen, Y., Krewski, D., Calle, E. E., & Thun, M. J. (2007). Chronic obstructive pulmonary disease is associated with lung cancer mortality in a prospective study of never smokers. American journal of respiratory and critical care medicine, 176(3), 285–290. https://doi.org/10.1164/rccm.200612-1792OC
 Smith, B. M., Pinto, L., Ezer, N., Sverzellati, N., Muro, S., & Schwartzman, K. (2012). Emphysema detected on computed tomography and risk of lung cancer: a systematic review and meta-analysis. Lung cancer (Amsterdam, Netherlands), 77(1), 58–63. https://doi.org/10.1016/j.lungcan.2012.02.019