A Texture-based Morphologic Enhancement Filter in Two-dimensional Thoracic CT scans

This paper presents a novel enhancement filter as a preprocessing step in the early detection of lung cancer. The identification and enhancement of the nodular structures is the initial stage in computer-aided diagnosis (CAD) for improving the sensitivity of nodule detection and reducing the number of false positives. Based on nodular texture feature and mathematical morphology, our proposed enhancement filter is simpler and automatic to extract and enhance the contrast of the region of interests (ROI) in thoracic computer tomography (CT) images. The proposed algorithm consists of the segmentation methods using gray-scale threshold, mathematical morphologic analysis and texture-based segmentation, and the enhancement method using contrast limiting adaptive histogram equalization (CLAHE). In our preprocessing stage, the automated segmentation and reconstruction of the pulmonary parenchyma has been performed. Then the ROI extraction based on nodular texture has been processed. Finally, the contrast of the ROI is enhanced by CLAHE. We applied our enhancement filter to two-dimensional (2D) CT images from LIDC using DICOM standards to show its effectiveness in the enhancement of the ROI. We believe that the enhancement filter developed in this study would be useful in the automated detection of nodules in 2D medical images