Ensemble-based hybrid probabilistic sampling for imbalanced data learning in lung nodule CAD

• Published in: Computerized medical imaging and graphics Vol. 38; no. 3; pp. 137 - 150
• Main Authors: Cao, Peng, Yang, Jinzhu, Li, Wei, Zhao, Dazhe, Zaiane, Osmar
• Format: Journal Article
• Language: English
• Published: New York, NY Elsevier Ltd 01.04.2014; Elsevier
• Subjects: Algorithms; Applied sciences; Artificial Intelligence; Biological and medical sciences; Computer science; control theory; systems; Data Interpretation, Statistical; Data processing. List processing. Character string processing; Ensemble classifier; Exact sciences and technology; False positive reduction; Humans; Imbalanced data learning; Internal Medicine; Investigative techniques, diagnostic techniques (general aspects); Lung Neoplasms - diagnostic imaging; Lung nodule detection; Mathematics; Medical sciences; Memory organisation. Data processing; Other; Pattern Recognition, Automated - methods; Probability and statistics; Radiographic Image Enhancement - methods; Radiographic Image Interpretation, Computer-Assisted - methods; Radionuclide investigations; Random subspace method; Re-sampling; Reproducibility of Results; Respiratory system; Sample Size; Sampling theory, sample surveys; Sciences and techniques of general use; Sensitivity and Specificity; Software; Solitary Pulmonary Nodule - diagnostic imaging; Statistics; Tomography, X-Ray Computed - methods; Re-sampling; Lung nodule detection; False positive reduction; Imbalanced data learning; Random subspace method; Ensemble classifier; Data analysis; Dimensionality; Probabilistic approach; Random sampling; Lung; Small signal; Subspace method; Unbalanced conditions; Respiratory system; Modeling; Aggregate model; Experimental result; Vector space; Receiver operating characteristic curves; Data structure; Computer aided design; Multiple classification
• ISSN: 0895-6111; 1879-0771; 1879-0771
• DOI: 10.1016/j.compmedimag.2013.12.003

Classification plays a critical role in false positive reduction (FPR) in lung nodule computer aided detection (CAD). The difficulty of FPR lies in the variation of the appearances of the nodules, and the imbalance distribution between the nodule and non-nodule class. Moreover, the presence of inherent complex structures in data distribution, such as within-class imbalance and high-dimensionality are other critical factors of decreasing classification performance. To solve these challenges, we proposed a hybrid probabilistic sampling combined with diverse random subspace ensemble. Experimental results demonstrate the effectiveness of the proposed method in terms of geometric mean (G-mean) and area under the ROC curve (AUC) compared with commonly used methods.