Lung Cancer Detection Using Fuzzy Auto-Seed Cluster Means Morphological Segmentation and SVM Classifier

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Artificial Intelligence in Medicine

Background:

• Lung cancer detection relies on accurate segmentation of nodules in Computed Tomography (CT) images.
• Existing methods may struggle with differentiating benign nodules from other structures like blood vessels and calcifications.

Purpose of the Study:

• To develop and validate an effective fuzzy auto-seed clustering morphological algorithm for lung nodule segmentation and lung cancer detection.
• To improve the accuracy of distinguishing malignant lung nodules using image features and a Support Vector Machine (SVM) classifier.

Main Methods:

• A fuzzy auto-seed clustering morphological algorithm was employed for nodule segmentation from CT slices.
• Image processing techniques, including centroid shift analysis and texture feature extraction (contrast, homogeneity, auto-correlation), were used for feature engineering.
• A Support Vector Machine (SVM) classifier was utilized to differentiate malignant nodules based on extracted features.

Main Results:

• The algorithm demonstrated high performance with sensitivity, specificity, and accuracy of 100%, 93%, and 94%, respectively.
• The system effectively eliminated non-nodule structures like blood vessels and calcifications.
• A low False Positive (FP) rate of 0.38 per patient was achieved.

Conclusions:

• The proposed fuzzy clustering algorithm is effective for accurate lung nodule segmentation and lung cancer detection in CT images.
• The combination of morphological segmentation, feature analysis, and SVM classification offers a robust approach for computer-aided diagnosis of lung cancer.
• This method shows significant potential for improving early lung cancer diagnosis and patient outcomes.