Soft Tool Developement for Characterization of Lung Nodule From Chest X-Ray Image

Abstract

CAD (Computer Aided Diagnosis) is a concept established by taking into account equally the roles of physician and computer to comment on disease. With CAD system, the performance given by computer does not have to be comparable to or better than that by physician, but need to be complementary to that by physician. To reduce the false positive and false negative diagnosis in determining whether the tumor is malignant or benign, doctors are taking help of CAD system. CAD using image processing technique has become one of the major research subjects in medical imaging and diagnostic radiology. Radiologist uses the CAD system output as a ‘second opinion’ and make the final decision to conform the disease. In present paper, suspicious area is segmented from the chest X-ray image after doing pre-processing of the image. Characterization of lung nodule means describing distinctive essential features. By doing segmentation, the features are extracted from the segmented region. All features are calculated from their respective mathematical formulas. Calculated features values vary according to arrangement of pixels in the image. Further this information can be used as an input to the CAD system for determining whether the segmented suspicious tumor area is malignant or benign. The proposed system will not replace the doctor’s role in detection of cancer but it will help doctor to take correct decision in short time with accuracy. It will act as second opinion before conformation of cancer.

Performance Analysis of Lung Cancer Diagnosis Algorithms on X-Ray Images

“…To develop a computer algorithm for lung cancer classification based on X-Ray images, image processing, feature extraction and machine learning techniques are required. For instance, the work in [3] performs a study to determine useful features that can be used to characterize lung nodules of being benign or malignant. Several features such as area, perimeter, irregularity index, convex area, convex perimeter, solidity, convexity, deficiency, equivalent diameter, kurtosis, skewness, mean, entropy, variance and standard deviation are extracted from the segmented lung nodules.…”