RESUMEN
Previously, convolutional neural networks mostly used deep semantic feature information obtained from several convolutions for image classification. Such deep semantic features have a larger receptive field, and the features extracted are more effective as the number of convolutions increases, which helps in the classification of targets. However, this method tends to lose the shallow local features, such as the spatial connectivity and correlation of tumor region texture and edge contours in breast histopathology images, which leads to its recognition accuracy not being high enough. To address this problem, we propose a multi-level feature fusion method for breast histopathology image classification. First, we fuse shallow features and deep semantic features by attention mechanism and convolutions. Then, a new weighted cross entropy loss function is used to deal with the misjudgment of false negative and false positive. And finally, the correlation of spatial information is used to correct the misjudgment of some patches. We have conducted experiments on our own datasets and compared with the base network Inception-ResNet-v2, which has a high accuracy. The proposed method achieves an accuracy of 99.0% and an AUC of 99.9%.
Asunto(s)
Procesamiento de Imagen Asistido por Computador , Redes Neurales de la Computación , Mama/diagnóstico por imagen , Procesamiento de Imagen Asistido por Computador/métodosRESUMEN
Vacuole membrane protein 1 (VMP1) was recently found to be involved in the process of tumor metastasis and is also considered to play a vital role in balancing apoptosis and autophagy. In the present study, the expression of VMP1 in colorectal cancer and matched adjacent noncancerous tissues was evaluated by immunohistochemistry (IHC) for studying the role of VMP1 in the process of colorectal cancer. KaplanMeier analysis and the log-rank test were used to calculate the correlation of classic clinicopathological characteristics related to survival and the expression of VMP1. In vitro, a VMP1 stable gene silencing cell model was constructed using a lentiviral vector. The invasive ability and proliferation of colorectal cancer cells were evaluated by Transwell and MTT assays, respectively, and the underlying signaling pathway was explored by western blotting. Additionally, drug susceptibility to cisplatin, oxaliplatin and 5-FU was tested before and after VMP1 knockout. Finally, an animal model was constructed to explore the role of VMP1 in the physiopathologic process of colorectal cancer. Our results indicated that VMP1 showed increased expression in the adjacent non-cancer tissues compared with that in the colorectal cancer tissues. For different stages of colorectal cancer, expression of VMP1 had a negative correlation with the malignancy of the cancer. In clinical research, we also found that the median survival of patients with low VMP1 expression was much shorter than the survival of patients with high expression. In vitro, after infection with the lentivirus, cells with VMP1 knockout gained significant aggressive properties in regards to invasion and proliferation, and the mechanisms may be related to the activation of the PI3K/Akt/ZO-1/E-cadherin pathway. We also found that shVMP1 cells were more sensitive to 5-FU, but not cisplatin and oxaliplatin. Finally, we found a higher number of formed nodules in nude mice after intraperitoneal injection with shVMP1 cells in the in vivo study.