Lung Image Patch Classification with Automatic Feature Learning

Image patch classification is an important task in many different medical imaging applications. The classification performance is usually highly dependent on the effectiveness of image feature vectors. While many feature descriptors have been proposed over the past years, they can be quite complicated and domain-specific. Automatic feature learning from image data has thus emerged as a different trend recently, to capture the intrinsic image features without manual feature design. In this paper, we propose to create multi-scale feature extractors based on an unsupervised learning algorithm; and obtain the image feature vectors by convolving the feature extractors with the image patches. The auto-generated image features are data-adaptive and highly descriptive. A simple classification scheme is then used to classify the image patches. The proposed method is generic in nature and can be applied to different imaging domains. For evaluation, we perform image patch classification to differentiate various lung tissue patterns commonly seen in interstitial lung disease (ILD), and demonstrate promising results.