RESUMO
It is crucial for understanding mechanisms of drug action to quantify the three-dimensional (3D) drug distribution within a single cell at nanoscale resolution. Yet it remains a great challenge due to limited lateral resolution, detection sensitivities, and reconstruction problems. The preferable method is using X-ray nano-computed tomography (Nano-CT) to observe and analyze drug distribution within cells, but it is time-consuming, requiring specialized expertise, and often subjective, particularly with ultrasmall metal nanoparticles (NPs). Furthermore, the accuracy of batch data analysis through conventional processing methods remains uncertain. In this study, we used radioenhancer ultrasmall HfO2 nanoparticles as a model to develop a modular and automated deep learning aided Nano-CT method for the localization quantitative analysis of ultrasmall metal NPs uptake in cancer cells. We have established an ultrasmall objects segmentation method for 3D Nano-CT images in single cells, which can highly sensitively analyze minute NPs and even ultrasmall NPs in single cells. We also constructed a localization quantitative analysis method, which may accurately segment the intracellularly bioavailable particles from those of the extracellular space and intracellular components and NPs. The high bioavailability of HfO2 NPs in tumor cells from deeper penetration in tumor tissue and higher tumor intracellular uptake provide mechanistic insight into HfO2 NPs as advanced radioenhancers in the combination of quantitative subcellular image analysis with the therapeutic effects of NPs on 3D tumor spheroids and breast cancer. Our findings unveil the substantial uptake rate and subcellular quantification of HfO2 NPs by the human breast cancer cell line (MCF-7). This revelation explicates the notable efficacy and safety profile of HfO2 NPs in tumor treatment. These findings demonstrate that this 3D imaging technique promoted by the deep learning algorithm has the potential to provide localization quantitative information about the 3D distributions of specific molecules at the nanoscale level. This study provides an approach for exploring the subcellular quantitative analysis of NPs in single cells, offering a valuable quantitative imaging tool for minute amounts or ultrasmall NPs.
RESUMO
PURPOSE: Convolution operator-based neural networks have shown great success in medical image segmentation over the past decade. The U-shaped network with a codec structure is one of the most widely used models. Transformer, a technology used in natural language processing, can capture long-distance dependencies and has been applied in Vision Transformer to achieve state-of-the-art performance on image classification tasks. Recently, researchers have extended transformer to medical image segmentation tasks, resulting in good models. METHODS: This review comprises publications selected through a Web of Science search. We focused on papers published since 2018 that applied the transformer architecture to medical image segmentation. We conducted a systematic analysis of these studies and summarized the results. RESULTS: To better comprehend the benefits of convolutional neural networks and transformers, the construction of the codec and transformer modules is first explained. Second, the medical image segmentation model based on transformer is summarized. The typically used assessment markers for medical image segmentation tasks are then listed. Finally, a large number of medical segmentation datasets are described. CONCLUSION: Even if there is a pure transformer model without any convolution operator, the sample size of medical picture segmentation still restricts the growth of the transformer, even though it can be relieved by a pretraining model. More often than not, researchers are still designing models using transformer and convolution operators.