An adaptive morphology based segmentation technique for lung nodule detection in thoracic CT image.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Radiology

Background:

• Lung cancer diagnosis relies on High-Resolution Computed Tomography (HRCT) for nodule detection.
• Manual nodule identification is time-consuming, necessitating automated solutions.
• Computer-aided detection (CADe) systems enhance radiologist efficiency and diagnostic accuracy.

Purpose of the Study:

• To introduce an adaptive morphology-based segmentation technique (AMST) for improved lung nodule segmentation.
• To develop an automated CADe system for accurate and efficient lung nodule detection.
• To reduce false positives in nodule detection from CT images.

Main Methods:

• Designed an adaptive morphological filter with an adaptive structuring element (ASE) for nodule candidate detection.
• Employed morphological, texture, and intensity-based features for classification.
• Utilized a Support Vector Machine (SVM) classifier.
• Evaluated performance using 10-fold cross-validation on LIDC/IDRI and private datasets.

Main Results:

• The proposed CADe system achieved 94.88% sensitivity, 93.45% specificity, and 94.27% accuracy on the LIDC/IDRI dataset.
• On a private dataset, the system achieved 91.43% sensitivity, 90.45% specificity, and 92.83% accuracy.
• Demonstrated superior performance compared to state-of-the-art methods in automatic nodule detection.

Conclusions:

• The AMST provides improved segmentation and accurate lung nodule detection.
• The developed CADe system offers a significant advancement for early lung cancer diagnosis.
• The system effectively reduces false positives, enhancing diagnostic reliability.