Automatic Detection of Masses From Mammographic Images via Artificial Intelligence Techniques

A novel computer-aided tool for automated sensing of normal tissue and abnormal masses from mammographic X-ray images is described. The pre-processing technique was firstly adopted for noise elimination on mammographic images. The automatic initialization of active contour was then placed on the pre-processed image for segmentation followed by deep convolutional neural networks to extract the features. Principal component analysis was then applied to choose the most significant features as input to the support vector machine classifier. Lastly, k-fold cross-validation techniques were executed for results validation. The developed tool was tested on public available datasets, namely Mammographic Image Analysis Society, and Digital Database for Screening Mammogram, based on eight evaluation methods: accuracy, sensitivity, specificity, receiver operating characteristic curve, area under curve (AUC), F1-score, precision, and recall. The outcome demonstrated the proposed system as a competitive tool in assisting radiologists as it attains an average of 95.24%, 93.94%, 96.61%, 94.66, 93.00%, 94.34%, and 0.98 for accuracy, sensitivity, specificity, precision, recall, F1-score, and AUC, respectively for testing on a combination of the aforementioned two datasets.