A recent study published in Nature Genetics reports a breakthrough in the early detection and prediction of chronic obstructive pulmonary disease (COPD) using machine learning (ML) technology. The study was conducted by a team of medical researchers, computer scientists, and engineers from various institutions across the United States.
COPD is a progressive lung disease that obstructs airflow and causes breathing difficulties, often resulting in chronic bronchitis and emphysema. It is the third leading cause of death worldwide and is responsible for a significant economic burden on healthcare systems. Early detection and treatment of COPD can slow its progression, improving patients’ quality of life and reducing healthcare costs.
The team utilized a deep convolutional neural network to train an ML model to distinguish healthy lungs from those with COPD. The training data was obtained from medical records, spirograms, and potential diagnosis classification systems. Spirograms measure lung strength by asking patients to blow into a tube-like device connected to a machine. Once the system could differentiate healthy lungs from those with COPD, the team added liability score data compiled over many years to identify early COPD in patients.
The researchers then tested their ML-based system on data from 325,000 patients in the UK Biobank, which included spirograms. They also included risk data from participants in several other healthcare-related initiatives. The ML-based liability score accurately predicted COPD-related hospitalization, discriminated against COPD cases and controls, and was associated with overall survival and exacerbation events. Additionally, the system conducted a genome-wide association study on the ML-based liability score, replicating existing COPD and lung function loci and identifying 67 new loci.
The team’s findings suggest that their system could be used to screen patients for COPD by feeding it spirogram data. It could also aid in further research to understand how COPD develops in the lungs and why it progresses so rapidly in some patients. The ML-based liability score could identify patients at high risk of developing COPD, leading to earlier detection and treatment, and thus, improving patients’ quality of life and reducing healthcare costs.
This study represents a significant advancement in COPD research, providing a framework for using ML methods and medical-record-based labels that do not require domain knowledge or expert curation to improve disease prediction and genomic discovery for drug design.
The system could help to improve COPD detection and aid in the development of new therapies that slow its progression. With the increasing availability of patient data and advancements in ML technology, the use of ML-based systems in disease prediction and genomic discovery is expected to become more common in the future.
