Breast　cancer　seems　to　be　the　most　deadly　cancer　for　females　globally.　Mammographic　masse　detection　is　essential　for　the　early　diagnosis　of　breast　cancer.　The　manual　detection　of　breast　masses　using　texture　analysis　from　digital　mammograms　is　hard　because　of　its　diverse　patterns.　Automatic　detection　of　breast　masses　from　mammograms　with　computer　algorithms　at　early　phases　could　help　physicians　to　avoid　unnecessary　biopsies.　In　the　current　study,　we　investigated　the　limitations　of　texture　analysis　for　mass　detection　by　creating　an　MR　taxon　filter　bank　using　Laplacian　of　Gaussian.　We　created　a　deep　taxon　feature　fusion　bank　for　the　EfficinetNet　DCNN　to　classify　benign　and　malignant　masses.　The　experimental　results　were　examined　on　the　public　CBIS-DDSM　and　the　local　PINUM　dataset.　Our　method　yielded　0.96　and　0.98　accuracies,　respectively.　The　results　findings　illustrated　that　our　method　significantly　outperformed　than　the　MobileNet,　AlexNet,　DenseNet,　and　previous　studies.　©　2022,　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.