Lightweight CNN for Detecting Microcalcifications Clusters in Digital Mammograms

Ricardo Salvador Luna-Lozoya, Humberto de Jesús Ochoa-Domínguez, Juan Humberto Sossa-Azuela, Vianey Guadalupe Cruz-Sánchez, Osslan Osiris Vergara-Villegas


Digital mammogram plays a key role inbreast cancer screening, with microcalcifications beingan important indicator of an early stage. However,these injuries are difficult to detect. In this paper,we propose a lightweight Convolutional Neural Network(CNN) for detecting microcalcifications clusters indigital mammograms. The architecture comprises twoconvolutional layers with 6 and 16 filters of 9×9,respectively at a full scale, a global pooling layer thateliminates the flattening and dense layers, and a sigmoidfunction as the output layer for binary classification. Totrain the model, we utilize the public INbreast databaseof digital mammograms with labeled microcalcificationclusters. We used data augmentation techniques toartificially increase the training set. Furthermore, wepresent a case study that encompasses the utilizationof a software application. After training, the resultingmodel yielded an accuracy of 99.3% with only 8,301parameters. This represents a considerable parameterreduction as compared to the 67,797,505 used inMobileNetV2 with 99.8 % accuracy.


Microcalcifications clusters detection, shallow convolutional neural network, deep learning

Full Text: PDF