Incoporating Weighted Board Learning System for Accurate Occupational Pneumoconiosis Staging
This work addresses inaccurate staging for occupational pneumoconiosis patients due to data imbalance, but it is incremental as it adapts existing methods to a specific medical imaging task.
The authors tackled the problem of imbalanced data in occupational pneumoconiosis staging from chest X-rays by proposing a model using gray level co-occurrence matrix for feature extraction and a weighted broad learning system for classification, achieving better performance than state-of-the-art classifiers on six hospital data cases.
Occupational pneumoconiosis (OP) staging is a vital task concerning the lung healthy of a subject. The staging result of a patient is depended on the staging standard and his chest X-ray. It is essentially an image classification task. However, the distribution of OP data is commonly imbalanced, which largely reduces the effect of classification models which are proposed under the assumption that data follow a balanced distribution and causes inaccurate staging results. To achieve accurate OP staging, we proposed an OP staging model who is able to handle imbalance data in this work. The proposed model adopts gray level co-occurrence matrix (GLCM) to extract texture feature of chest X-ray and implements classification with a weighted broad learning system (WBLS). Empirical studies on six data cases provided by a hospital indicate that proposed model can perform better OP staging than state-of-the-art classifiers with imbalanced data.