RESUMO
PURPOSE: To develop a computerized detection system for the automatic classification of the presence/absence of mass lesions in digital breast tomosynthesis (DBT) annotated exams, based on a deep convolutional neural network (DCNN). MATERIALS AND METHODS: Three DCNN architectures working at image-level (DBT slice) were compared: two state-of-the-art pre-trained DCNN architectures (AlexNet and VGG19) customized through transfer learning, and one developed from scratch (DBT-DCNN). To evaluate these DCNN-based architectures we analysed their classification performance on two different datasets provided by two hospital radiology departments.DBT slice images were processed following normalization, background correction and data augmentation procedures. The accuracy, sensitivity, and area-under-the-curve (AUC) values were evaluated on both datasets, using receiver operating characteristic curves. A Grad-CAM technique was also implemented providing anindication of the lesion position in the DBT slice. RESULTS: Accuracy, sensitivity and AUC for the investigated DCNN are in-line with the best performance reported in the field. The DBT-DCNN network developed in this work showed an accuracy and a sensitivity of (90% ± 4%) and (96% ± 3%), respectively, with an AUC as good as 0.89 ± 0.04. Ak-fold cross validation test (withk = 4) showed an accuracy of 94.0% ± 0.2%, and a F1-score test provided a value as good as 0.93 ± 0.03. Grad-CAM maps show high activation in correspondence of pixels within the tumour regions. CONCLUSIONS: We developed a deep learning-based framework (DBT-DCNN) to classify DBT images from clinical exams. We investigated also apossible application of the Grad-CAM technique to identify the lesion position.
Assuntos
Neoplasias da Mama , Aprendizado Profundo , Área Sob a Curva , Neoplasias da Mama/diagnóstico por imagem , Feminino , Humanos , Mamografia , Redes Neurais de Computação , Curva ROCRESUMO
The origin recognition complex (ORC) regulates DNA replication. However, some members of the ORC core, such as ORC3 and ORC5, have been implicated in neuronal maturation. In cultured cerebellar granule cells (CGCs), ORC3 mRNA and protein levels increased from 6 to 8days in vitro, a time that coincided with the maximal development of the dendritic arbor. In contrast, expression of ORC5 remained low throughout CGC maturation. Activation of type-4 metabotropic glutamate receptors with the selective enhancer, PHCCC, during a critical time-window (from 4 to 6days in vitro) anticipated the developmental peak of ORC3, increased the expression of two proteins associated with neuronal maturation, i.e. the mitogen-associated protein-2 (MAP-2) and postsynaptic density-95 (PSD-95), as well as dendritic length. siRNA-induced ORC3 knockdown reduced MAP-2 and PSD-95 expression on its own and abrogated the action of PHCCC. We examined whether the maturational effects of ORC3 were mediated by changes in the activity of the monomeric GTP-binding protein, Rho, which is known to regulate granule cell morphology. ORC3 knockdown increased the levels of the GTP-bound active form of Rho, whereas exposure to PHCCC reduced Rho activation. The action of PHCCC was largely attenuated in cultures deprived of ORC3. Finally, granule cell exposure to the Rho-associated kinase inhibitor, Y-27632, abolished the lowering effect of ORC3 knockdown on MAP-2 expression, and increased dendritic length. These data suggest that ORC3 supports neuronal maturation by inhibiting the Rho signaling pathway, and mediates the differentiating activity of mGlu4 receptors in cultured cerebellar granule cells.