56016 research outputs found
Sort by
Tumor Segmentation on Orthotopic Breast Cancer Model from MR Images using U-Net with Pre-trained ResNet34
Get PDFDepartment of Biomedical EngineeringOne of the most common cancers in the world is breast cancer. To diagnose breast cancer, MRI is one of the most common screening techniques and scan tumor as small as 2mm. However, radiologists overlook tumors when the size is smaller than 2cm. Therefore, a preclinical study is required to develop novel diagnostics. Studying breast cancer in vitro has limitations because of its complexity in imitation. Therefore, an orthotopic breast cancer model, which shares many features of a human primary breast tumor, is generally used for research. Longitudinal MR images, which have various time points, and tumor volumes can be acquired when using an orthotopic breast cancer model with MRI. From these, we can diagnose small tumors at an early stage and obtain detailed information about treatment efficacy. As a result, orthotopic model research using MRI is significant.
To measure the tumor volume, tumor regions should be segmented by an expert. However, due to time consuming tasks, currently, there is various deep learning-based automatic breast tumor segmentation research. The efficiency of deep learning-based segmentation, nevertheless, has several limitations. First and foremost, longitudinal MR images were not utilized. We can detect small tumors from continuous MR images, which were difficult to detect on a single MR image. Second, the number of MR images and tumor volume were insufficient. Finally, the therapeutic effect was not considered. Most orthotopic model research utilizes chemotherapeutic agents, which affect tumor volume and shape. Therefore, it was hard to apply general orthotopic model research.
In this study, we obtained the untreated group (n=24) and treated group (n=6) with Doxorubicin, a chemotherapeutic agent, to consider the therapeutic effect. To generate a deep learning-based tumor segmentation framework, we chose U-Net + ResNet34 architecture, which replaced the encoder of U Net in pre-trained ResNet34. The untreated group (n=19) was used to train the model and the untreated group (n=5) was utilized as the first evaluation. Then, the treated group (n=6) was also utilized as a second evaluation of the trained model. As a result, the trained model had DICE of 0.912 in the untreated group, DICE of 0.927 in the treated group, and DICE of 0.920 in the overall test set, which had a successful performance. Furthermore, the model can segment small tumors, with more than 2mm3, on a single MR image. Also, the analyses of the results were performed: tumor volume growth and 3D volume rendering. By utilizing these analysis methods, we could identify the characteristics and insights of tumors.
In conclusion, first, the proposed research showed accurate tumor segmentation in the untreated and treated group. This result indicated that the model could monitor tumor growth and the therapeutic response of Doxorubicin. Second, small tumor volumes, more than 2 mm3, can be segmented by the model. Therefore, with this framework, the tumor in the early stage can be detected on a single MR image. As a result, the proposed segmentation framework can be able to apply to general orthotopic model.clos
Development of Recombinant Secondary Antibody Mimics (rSAMs) for Immunoassays using a Monomeric Alkaline Phosphatase
No full textDepartment of Biological Sciencesclos
Identification of DNA Repair Pathways Important for Temozolomide Resistance of Mismatch Repair Deficient Human Cell Lines
No full textDepartment of Biological Sciencesclos
Pioneering Approaches to Enhance Stability and Efficiency in Tin-based Perovskite Solar Cells
No full textSchool of Energy and Chemical Engineering (Energy Engineering)clos
?????? ?????? ????????? ?????? ?????????????????? ????????? ?????? ????????? ???????????? ??????????????????
No full textSchool of Energy and Chemical Engineering (Energy Engineering)clos
????????? ?????? ??? ????????? ?????? ???????????? ?????? ?????? ????????? ?????? ?????? ??????
Get PDFSchool of Energy and Chemical Engineering (Energy Engineering (Battery Science and Technology))Na ion battery has received significant interest as an alternative of Li ion battery because of abundant amount of Na resources than Li. However, several challenges disturb the commercialization of Na ion battery, which is cause by large size of Na ion. The physical limitations from Na ion size induces slow kinetic properties with large overpotential and structural distortion, causing poor cyclability. In addition, the cathode materials for Na ion battery should have large specific capacity for high energy densities because of the intrinsic low voltages of Na. Therefore, cathode materials should be designed for satisfying the requirements.
Among many candidates for cathode materials of Na ion batteries, Prussian blue analogues(PBAs) have been focused on a promising cathode material of Na ion battery. The large channel size of PBAs provides enough space for Na ion and fast diffusivity of Na ion during electrochemical reaction. In addition, the low-cost process for synthesis of PBA increases the possibility of commercialization of Na ion battery by maximizing merit of Na ion battery. However, the structural distortion of PBAs with high specific capacity causes poor cyclability and transition metal dissolution issue.
In our research, we investigated the solution of Jahn-Teller distortion which is representative cause of structural distortion of PBAs. Charge redistribution can mitigate the Jahn-Teller distortion by reducing the amount of high spin MnIII state during electrochemical reactions. In addition, we demonstrate the requirements for activating charge redistribution in PBA based on our experiments. First, structures should have enough passages for electron transfer. Second, electrons can be transferred when there is strong driving force. Lastly, unstable state should be continued for smooth electron transfer. To satisfy the conditions, we introduced water-in-salt electrolyte system with fast kinetic properties and doped different kinds of transition metal ions for inductive effect.
Based on the understanding about PBAs, we studied PBAs as a cathode material of thermally regenerative electrochemical cycle (TREC). TREC is originated battery system which energy can be stored as an electrochemical energy. Different with battery system, TREC can convert thermal energy to electrochemical energy during charge and discharge process. Therefore, we can use more energy than we are stored in the battery. One of the important variables for TREC is temperature coefficient which can decide the amount of harvested energy. In this work, we controlled intrinsic properties of PBAs for higher temperature coefficient value. As a result, our research shows the best performance among Na ion aqueous electrolyte based TREC system.clos
Local spinel transformation enabling high-rate low Li-excess Mn-rich disordered rock-salt Li-ion battery cathodes
No full textSchool of Energy and Chemical Engineering (Energy Engineering (Battery Science and Technology))clos
Understanding of Growth Retardation in phaC-deleted Methylorubrum extorquens AM1 through Adaptive Laboratory Evolution
No full textSchool of Energy and Chemical Engineering (Chemical Engineering)Methylorubrum extorquens AM1 has gained attention as a model microorganism for methanol- and formate-based biorefineries because it has a particular ability of growing on one carbon molecules. The AM1 exhibits remarkable metabolic diversity, allowing it to survive in harsh conditions. However, certain metabolic pathways, such as carotenoids and poly(3-hydroxybutyrate) (PHB) synthetic pathways, may need to be excluded for biorefinery applications of this strain due to the fairly many genes and cellular resources involved in these processes. In this study, we eliminated the PHB synthesis ability by deleting phaC in AM1. However, the phaC-deleted AM1 shows growth retardation (?? = 0.08 h-1) compared with wild-type (?? = 0.18 h-1) under the succinate supplied condition. To investigate the retardation mechanism of the knockout mutant, we employed adaptive laboratory evolution. Mutants were selected based on two criteria: 1) an increased specific growth rate and 2) higher maximum OD600nm compared to the un-evolved strain. Subsequently, these isolated strains were subjected to re-sequencing to understand the growth retardation mechanism in phaC-deleted AM1. Further investigations need to characterize the activities and mechanisms the introduced mutations, also their effect on growth and biochemical production in AM1. This study would contribute to a better understanding microbial metabolism and provide an insight for the development of AM1 as a biotechnological chassis.clos
Resilience and social change: Findings from research trends using association rule mining
No full textThis study analyzed the historical development of resilience with respect to multidisciplinary aspects using association rule mining (ARM). ARM is a rule-based machine-learning approach tailored to identify validated relations among multiple variables in a large dataset. This study collected author keywords from all resilience-related literature in the Web of Science database and examined the changes in validated resilience-related topics using ARM. We found that resiliencerelated research tends to diversify and expand over time. Although topics and their academic fields related to engineering and complex adaptive systems were prominent in the early 2000s, psychosocial resilience and social-ecological resilience have received significant attention in recent years. The increasing interest in resilience-related topics linked to psychological and ecological factors, as well as social system components, can be attributed to the impact of a series of complex and global events that occurred in the late 2000s. Recently, resilience has been conceived as a way of thinking, perspective, or paradigm to address emergent complexity and uncertainty with vague concepts. Resilience is increasingly being regarded as a boundary spanner that promotes communication and collaboration among stakeholders who share different interests and scientific knowledge
Effects of Cr and Mo contents on the flow-accelerated corrosion behavior of low alloy steels in the secondary side of pressurized water reactors
No full textCarbon steels have been extensively used as structural materials in the secondary side of pressurized water reactors, which are susceptible to flow-accelerated corrosion (FAC). To overcome this issue, low-alloy steels containing 2.25Cr-1Mo (P22) have been introduced, as Cr and Mo are effective alloying elements in mitigating FAC. However, the behavior of these alloying elements in the secondary water chemistry is not well-understood, and therefore, there is a need to explore alternative materials that can address the issue of FAC. In this study, three model alloys were manufactured based on Ducreux's model on FAC rate to examine the effects of Cr and Mo on the corrosion behavior of low-alloy steels compared to that of commercial P22. Microstructure of P22 was ferritic/pearlite structure while the model alloys was ferritic/tempered bainite structure due to the existence of Mn and Si. The difference in microstructure led to the difference in hardness values, and elimination of Mo from the model alloys resulted in reduction in strain. The FAC rate was mainly influenced by the Cr content in the oxide layer rather than Mo. The passive film layer hinders the exposure of the inner layer to H2O in the solution, resulting in Cr3+ dissolution into the Fe oxide layer. The continuous passivation leads to the formation of two compact layers, one amorphous and one crystalline Cr, creating Fe oxide substitutes. It is also revealed that the solubility of Cr species is much lower than that of Fe species resulting in the enrichment of Cr in the outer layer, and thus the higher Cr content reinforces the passivity of the steels. The Fe-Cr alloy exhibited promising corrosion resistance, suggesting that it could be a potential substitute for Fe-Cr-Mo alloy