| Summary: | Automatic lung segmentation and lung nodule detection through High-
Resolution Computed Tomography (HRCT) image is a new and exciting research
in the area of medical image processing and analysis. In this research, two new
techniques for segmentation of lung regions and extraction of nodules on the
HRCT image are proposed. An automatic lung segmentation system is proposed for
identifying the lungs in HRCT lung images. First, lung regions are extracted from
the HRCT images by grey-level thresholding. The lung background information is
eliminated by linear scans originating from border pixels. Finally, lung boundaries
are smoothed along the mediastinum. The lung nodule extraction from the HRCT
image is processed based on a set of continuous HRCT slices of lung images. In the
first stage, the abnormal areas are extracted based on nodule pixel collection and
combination. In the final stage, the abnormal area is extracted by comparing the
density and shape profile. Both of the systems have been tested by processing data
sets from 10 continuous image sets (100 images). Lung segmentation results are
presented by comparing our automatic method to manually traced borders.
Averaged over all results, the accuracy of lung segmentation is 96.10%. The
proposed nodule detection method has been tested on image sets containing healthy
and unhealthy lung images. Statistical analysis has been done and the results show
the overall nodule detection rate is 88.44% along with the false positive rate of 0.18.
|
|---|
