Combined Naïve Bayes and logistic regression for quantitative breast sonography

Abstract

Sonography is commonly used as an adjunct to mammography for early detection of breast cancer. We are developing methods to classify solid breast masses in sonograms as malignant or benign. The goal of this study was to combine two independent probabilistic classifiers to improve computer-aided diagnosis of breast masses. Naïve Bayes and logistic regression were used for supervised classification of masses from extracted morphological sonographic features, in combination with mammographic BI-RADS (categories 1 to 5) and patient age. Solid masses with biopsy-proven diagnoses were analyzed. Training and testing were performed using leave-one-out cross validation. Diagnostic performance was evaluated by the area under the curve (AUC) of the receiver operating characteristic (ROC). Agreement between predictions from the two classifiers was used to differentiate benign and malignant masses. The resultsshow that logistic regression and Naïve Bayes performed with ROC area of 0.902 ± 0.023 and 0.865 ± 0.027, respectively. The combined use of logistic regression and Naïve Bayes demonstrated reduction in biopsies by 48%, with malignancy missed in 2% of cases (false negative rate of 6.4%).Index Terms -quantitative breast ultrasound; computer-aided diagnosis; breast cancer; machine learning.

Color Doppler Ultrasound Improves Machine Learning Diagnosis of Breast Cancer

“…These nine features were motivated by the conventional visual features used in the clinic to assess lesions on ultrasound, described by Stavros et al [ 26 ]. They include the angular variation at the margin (AVM), angular variation of the interior (AVI), brightness difference (BD), margin sharpness (MS), tortuosity, depth-to-width ratio (DWR), axis ratio (AR), radius variation (RV), and ellipse-normalized skeleton (ENS) [ 9 , 14 , 23 ]. The formulas for deriving the features are given in the literature [ 9 , 10 ] and in the Appendix A Table A1 .…”

“…Although canonical AdaBoost boosts weak homogeneous learners, the motivation for boosting already-strong heterogeneous classifiers is that if one classifier already has very high performance, the optimal classification strategy is likely close to the already-strong strategy, so it is boosted with a learner that provides a second opinion. Previous studies found that naïve Bayes and logistic regression both performed well on the breast mass classification problem and, then, that they could be boosted even though their predictions were close case by case [ 10 , 20 , 21 , 23 ]. Though stacking is usually used instead of boosting with heterogeneous learners, the close agreement by case between our constituent learners makes them poor candidates for stacking.…”

“…The two classifiers have been combined into a higher-performing model trained on grayscale (GS) features alone [ 10 ] using serial adaptive boosting (AdaBoost [ 22 ]). Instead of boosting a homogenous weak learner as in canonical AdaBoost, the method boosts heterogeneous classifiers, such as naïve Bayes and logistic regression, that are already known to both perform well and to even closely agree on cases [ 10 , 20 , 21 , 23 ].…”

Athlete Health Prediction Using Machine Learning Methods

“…ML methods have also been used in many areas of medical science including breast cancer detection, cardiovascular risk prediction, stroke prediction, Parkinson's disease prediction, and medical imaging, to name a few [23,24,25,26,27,28,29]. For example, researchers in [29] use naive Bayes and logistic regression models for early detection of breast cancer. In the paper, the authors claim that computer-based quantitative methods improve diagnosis on breast ultrasound and have the potential to reduce the number of biopsies.…”

The diagnostic performance of leak-plugging automated segmentation versus manual tracing of breast lesions on ultrasound images

“…The grayscale and morphological features were: brightness difference at the margin, margin sharpness, angular variation in brightness at the margin, depth-to-width ratio, axis ratio, tortuosity, radius variation, and elliptically normalized skeleton. 7,21 The features, F, were used with logistic regression to determine the probability of malignancy (P(MjF)). A round-robin (leave-oneout) approach was used to assess the discriminating capability of the extracted lesion features: N-1 samples of the N samples in the data were trained to predict the behavior of the remaining sample, and the process was repeated until each sample had been the test case.…”