Deep learning for detecting and early predicting chronic obstructive pulmonary disease from spirogram time series

Abstract

Chronic Obstructive Pulmonary Disease (COPD) is a chronic lung condition characterized by airflow obstruction. Current diagnostic methods primarily rely on identifying prominent features in spirometry (Volume-Flow time series) to detect COPD, but they are not adept at predicting future COPD risk based on subtle data patterns. In this study, we introduce a novel deep learning-based approach, DeepSpiro, aimed at the early prediction of future COPD risk. DeepSpiro consists of four key components: SpiroSmoother for stabilizing the Volume-Flow curve, SpiroEncoder for capturing volume variability-pattern through key patches of varying lengths, SpiroExplainer for integrating heterogeneous data and explaining predictions through volume attention, and SpiroPredictor for predicting the disease risk of undiagnosed high-risk patients based on key patch concavity, with prediction horizons of 1-5 years, or even longer. Evaluated on the UK Biobank dataset, DeepSpiro achieved an AUC of 0.8328 for COPD detection and demonstrated strong predictive performance for future COPD risk (p-value < 0.001). In summary, DeepSpiro can effectively predict the long-term progression of COPD disease.

Fig. 1. Our input module uses the raw Time–Volume curve time series collected from hospitals and patient demographic data, which are then passed into our AI-based model module.

The AI-based model module is divided into four tasks (see the section “Overview” for details). After processing through the AI-based model module, the output data is handled by the output module. If the AI-based model diagnoses the individual as a COPD patient, we will output their diagnosis results and the interpretability figure of the model. If the AI-based model diagnoses the individual as a non-COPD patient, we will output their risk of developing COPD over the next 1-5 years.

Fig. 2. Evaluation comparison.

a–c Comparison of receiver operating characteristic (ROC) curves for three prediction methods—DeepSpiro COPD risk predictions, Volume-Flow ResNet18 COPD risk predictions, and FEV1/FVC ratio-based risk—across three evaluation tasks: All COPD patients (left), future COPD-related hospitalization (center), and COPD-related death (right). d-f Comparison of precision-recall (PR) curves across three evaluation tasks—All COPD patients (left), future COPD-related hospitalization (center), and COPD-related death (right)—for three prediction methods: DeepSpiro COPD risk predictions, Volume-Flow ResNet18 COPD risk predictions, and FEV1/FVC ratio-based risk. The error bars represent bootstrapped 95% confidence intervals (n = 100 bootstrapping samples).

Fig. 3. The nomogram of COPD detection.

The nomogram illustrates the contribution of demographic information and the FEV1/FVC diagnostic gold standard to the model’s diagnostic accuracy for COPD. The nomogram allows for visual estimation of the probability of COPD diagnosis by assigning weighted scores to each variable. The “threshold” in the figure represents the cutoff value at which the predicted probability indicates a positive diagnosis for COPD. For instance, a threshold of 0.5 means that a predicted probability greater than 0.5 would be considered indicative of COPD.

Fig. 4. Future COPD prediction overview.

a The left vertical axis represents the predicted probability of COPD, while the right vertical axis indicates the concavity degree based on the directed area metric for each phase. Each plot displays the mean concavity degree for the respective phase, illustrating how the concavity changes over time in relation to COPD risk. b This figure illustrates the probabilities of not being diagnosed with COPD over time for the high-risk and low-risk groups, as predicted by the model. The X-axis represents the time since the pulmonary function test, while the Y-axis shows the probability of not being diagnosed with COPD at each time point. Due to right censoring, not all high-risk patients are diagnosed within the observation period, resulting in probabilities that remain above zero. c As the onset time progresses, the concavity of the patient’s Volume-Flow curve decreases year by year.

Fig. 5. Subgroup analysis overview.

a The subgroup analysis for smoking. b The subgroup analysis by sex. c The subgroup analysis by age. d As the onset time progresses, an individual’s concavity measure gradually decreases. Compared to non-smokers, smokers show significantly higher lung function concavity measures. e As the onset time progresses, an individual’s concavity measure gradually decreases. Compared to females, males show significantly higher lung function concavity measures. f As the onset time progresses, an individual’s concavity measure gradually decreases. Compared to younger patients, older patients show significantly higher lung function concavity measures.

Fig. 6. SHAP analysis overview.

a Brighter colors mean that the feature has a more positive impact on the model’s predictions. The blow ratio represents the FEV1/FVC value, an important indicator of lung function. b Relationship between smoking status and predicted risk of COPD. The figure illustrates how the risk value predicted by DeepSpiro correlates with smoking status, indicating a higher risk value is associated with a larger proportion of smokers. This result is consistent with clinical findings, as smokers are more likely to develop COPD. c Relationship between age and predicted risk of COPD. The figure demonstrates that the predicted risk value of DeepSpiro increases with age, indicating that older individuals have a higher risk of developing COPD. This observation is consistent with clinical findings, which show that older patients are more susceptible to COPD.

Fig. 7. The explainer of the model.

The brighter the color, the more attention the model pays. a A model interpretability figure for cases where non-COPD patients detected as non-COPD by the model. b A model interpretability figure for cases where non-COPD patients detected as COPD by the model. c A model interpretability figure for cases where COPD patients detected as non-COPD by the model. d A model interpretability figure for cases where COPD patients detected as COPD by the model.

Fig. 8. Framework overview.

It is divided into four modules. In the first module, we process the varied-length original volume curve and convert it into a smoothed Volume-Flow curve. In the second module, we extract features from the varied-length Volume-Flow curve. In the third module, combined with demographic information, we output the COPD detection results and model explainer. In the fourth module, based on concavity information, we output the risk values for COPD at different future periods.

Fig. 9. Spirometry curve and smoothing effects overview.

a Examples of Volume-Flow diagrams for different populations. b The conversion of the original Time-Volume curve. To obtain the degree of airflow limitation, we need to use finite difference methods on the original Time-Volume curve to calculate the corresponding Volume-Flow curve. c Smoothing comparison, the original curve shows fluctuations in some areas when converting to Time-Flow and Volume-Flow curves. After smoothing, the curves in the same positions become more stable.

Fig. 10. Volume-Flow curves illustrating phase-specific concavity and airway collapse.

a A representative Volume-Flow curve divided into four phases: early (PEF–FEF25), mid-early (FEF25–FEF50), mid-late (FEF50–FEF75), and late (FEF75+), highlighting concave and convex segments relative to the baseline. b Volume-Flow curve of a healthy individual, showing airway collapse occurring in the late phase (FEF75+). c Volume-Flow curve of a COPD patient, showing early-stage airway collapse in the early phase (PEF–FEF25).