Automated detection of lung nodules in CT images using shape-based genetic algorithm

• Published in: Computerized medical imaging and graphics Vol. 31; no. 6; pp. 408 - 417
• Main Authors: Dehmeshki, Jamshid, Ye, Xujiong, Lin, XinYu, Valdivieso, Manlio, Amin, Hamdan
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.09.2007
• Subjects: Algorithms; Artificial Intelligence; Computer-aided detection; Genetic algorithm; Humans; Internal Medicine; Lung Neoplasms - diagnostic imaging; Lung nodule detection; Other; Pattern Recognition, Automated - methods; Phantoms, Imaging; Radiographic Image Enhancement - methods; Radiographic Image Interpretation, Computer-Assisted - methods; Reproducibility of Results; Sensitivity and Specificity; Shape index; Solitary Pulmonary Nodule - diagnostic imaging; Template matching; Tomography, X-Ray Computed - instrumentation; Tomography, X-Ray Computed - methods; CAD; FP; computer-aided detection; Template matching; HU; false positives; GATM; genetic algorithm (GA) template matching (TM); Lung nodule detection; Genetic algorithm; SI; shape index; Hounsfield Units; Shape index; Computer-aided detection
• ISSN: 0895-6111; 1879-0771
• DOI: 10.1016/j.compmedimag.2007.03.002

Abstract A shape-based genetic algorithm template-matching (GATM) method is proposed for the detection of nodules with spherical elements. A spherical-oriented convolution-based filtering scheme is used as a pre-processing step for enhancement. To define the fitness function for GATM, a 3D geometric shape feature is calculated at each voxel and then combined into a global nodule intensity distribution. Lung nodule phantom images are used as reference images for template matching. The proposed method has been validated on a clinical dataset of 70 thoracic CT scans (involving 16,800 CT slices) that contains 178 nodules as a gold standard. A total of 160 nodules were correctly detected by the proposed method and resulted in a detection rate of about 90%, with the number of false positives at approximately 14.6/scan (0.06/slice). The high-detection performance of the method suggested promising potential for clinical applications.