Neural Network Approach to Investigating the Importance of Test Items for Predicting Physical Activity in Chronic Obstructive Pulmonary Disease

Abstract

Contracting COPD reduces a patient's physical activity and restricts everyday activities (physical activity disorder). However, the fundamental cause of physical activity disorder has not been found. In addition, costly and specialized equipment is required to accurately examine the disorder; hence, it is not regularly assessed in normal clinical practice. In this study, we constructed a machine learning model to predict physical activity using test items collected during the normal care of COPD patients. In detail, we first applied three types of data preprocessing methods (zero-padding, multiple imputation by chained equations (MICE), and k-nearest neighbor (kNN)) to complement missing values in the dataset. Then, we constructed several types of neural networks to predict physical activity. Finally, permutation importance was calculated to identify the importance of the test items for prediction. Multifactorial analysis using machine learning, including blood, lung function, walking, and chest imaging tests, was the unique point of this research. From the experimental results, it was found that the missing value processing using MICE contributed to the best prediction accuracy (73.00%) compared to that using zero-padding (68.44%) or kNN (71.52%), and showed better accuracy than XGBoost (66.12%) with a significant difference (p < 0.05). For patients with severe physical activity reduction (total exercise < 1.5), a high sensitivity (89.36%) was obtained. The permutation importance showed that "sex, the number of cigarettes, age, and the whole body phase angle (nutritional status)" were the most important items for this prediction. Furthermore, we found that a smaller number of test items could be used in ordinary clinical practice for the screening of physical activity disorder.

Keywords: COPD; autoencoder; neural network; physical activity; prediction.

Figure 1

The model structure of the 3-layer autoencoder used for pre-training. There were 34 units in the input layer. In the hidden layer, a feature vector (feature values) that encoded the inputs was obtained. There were 34 units in the output layer, in which the inputs were reconstructed from the hidden layer vectors. The number of units in the hidden layer was 20, i.e., the feature values were represented by 20-dimensional vectors.

Figure 2

The model structure used for the fine-tuning, which was constructed after pre-training with the 3-layer autoencoder. The output layer in Figure 1 was removed and the model was connected to a new output layer to predict T-Ex1 and T-Ex2.

Figure 3

The model structure of the 4-layer autoencoder used for pre-training. The input layer had 34 units. In the two hidden layers, the vectors (feature values) that encoded the inputs were obtained. The output layer had 34 units and the inputs were reconstructed from the hidden layer vectors.

Figure 4

The model structure used for fine-tuning, which was constructed after pre-training with the 4-layer autoencoder. The output layer in Figure 3 was removed and the model was connected to a new output layer to predict T-Ex1 and T-Ex2.

Figure 5

The importance of each test item. The horizontal axis shows the test items and the vertical axis shows their importance. The test items in order of effect on prediction performance were sex, the number of cigarettes per day, age, whole body phase angle, DDR, mMRC, FEV₁, %RV/TLC, pack years, and BNP.