34254 research outputs found
Sort by
Ovarian Cancer Metastasis Segmentation using CT image based on Deep Learning
No full textOvarian cancer is a type of cancer that originates from the female ovaries or nearby tissues. Tumor cells can invade and metastasize to other organs. There is currently no effective method for early detection of ovarian cancer, so it\ue2s necessary to undergo regular examinations for timely detection and treatment. These examinations may include physical examination, blood tests, and imaging examination.
Imaging examinations for ovarian cancer include transvaginal ultrasound (TVS), computed tomography (CT), and magnetic resonance imaging (MRI). Because the imaging scope and accuracy of CT images surpass TVS images, and MRI images are more costly and time-consuming than CT images, clinicians usually select CT images for diagnosing ovarian cancer. Therefore, this study selects computed tomography scan images as the research data.
When observing the images, the size, location, and morphology of the tumor are important for clinicians as they influence the treatment plan and choice of surgical methods. To assist clinicians in clinical diagnosis and judge, this study trained a deep learning semantic segmentation model on CT images for segmenting seven types of tumors and metastatic symptoms: ovarian malignant tumor, liver metastasis, lymph node metastasis, omentum carcinomatosis, tumor seeding, spleen metastasis and pancreatic metastasis.
The images were annotated by professional physicians and radiologists, totaling 10542 CT slices. After pre-processing such as removing CT bed, denoising, resizing, and normalization, the processed image size is 512\uc3512. Due to the imbalance in the distribution of metastatic symptoms, divisions were made based on each patient's combination of classes to balance the number of classes in the training set, validation set, and test set, while reducing the number of images without symptoms. This approach aimed to mitigate the impact of data imbalance. Subsequently, data augmentation was employed to enhance the diversity of the dataset. The deep learning semantic segmentation model used in this study is DeepLabV3+. The loss function adopts Categorical Cross Entropy, and the evaluation metrics include MIoU (Mean Intersection over Union) and a custom function, Hit rate. The MIoU on the test set is measured at 0.709. The segmentation performance is more favorable for classes with sufficient numbers, relatively fixed distribution, or larger areas, providing assistance to clinicians in diagnosis. Examples include liver metastasis, ovarian malignant tumor, and omentum metastasis
Is ESG a Poison, an Antidote, or Merely a Placebo?- Take the Changes in Market Capitalization of the 'Coalition of Movers and Shakers on Sustainable Finance' During the COVID-19 Pandemic as an Example.
No full textThis research primarily investigates whether actively promoting sustainable development and implementing ESG initiatives in alignment with government policies is an unavoidable burden or an opportunity for Taiwan's listed financial holding companies to wield their investment influence while seizing the future of sustainable investments.
The report employs the Difference-in-Difference Model for analysis, comparing market value changes before and after the Covid-19 outbreak between members of the " Coalition of Movers and Shakers on Sustainable Finance " and non-members.
Empirical results show that Taiwanese financial holding companies have achieved significant market value growth post-pandemic. Companies that joined the Coalition also saw significant growth in market value. However, the impact on market value post-pandemic for members of the coalition was not significant. A possible reason is that the Coalition was established towards the end of the pandemic when most countries had returned to pre-pandemic norms by 2022.
The international surge in ESG and actions corresponding to the SDGs (Sustainable Development Goals) have become increasingly prominent in corporate development. Although Taiwan is not a member of the United Nations or other international organizations, the determination of its regulatory authorities to align with international standards is imperative. In addition to gradually formulating relevant regulations and actively guiding the financial sector as providers of capital to follow responsible investment principles and use their investment influence, organizations like this have been established to influence peers and change the environment for sustainable development
A Comparative Study of the Performance of Robo-Advisors and ETFs that Track Global Indices
No full textIn recent years, robo-advisor has gradually become a new type of investment and financial services, attracting the attention of many investors with advantages such as automation, low cost, and convenience. This study compares the investment performance of six robo-advisor services in Taiwan with ETFs tracking global indices.
The research results show that the cumulative investment return rates of the six roboadvisor services during the same period are all better than traditional passive investment benchmarks. This result is in stark contrast to the fact that most actively managed large funds have long lagged behind the benchmark indices. From the performance indicators of Treynor's ratio, Jensen's index, Sharpe's ratio, and information ratio, the performance
of robo-advisor is also better than the benchmark of passive investment.
Robo-advisor performs better than traditional passive investment benchmarks in the short term. However, the long-term performance still needs further verification. If the sample observation period is expanded, the long-term performance advantage of roboadvisor needs further verification. But preliminary results show that intelligent investment decision-making tools do indeed demonstrate specific value. If this trend
continues, robo-advisor is expected to gradually replace manual finance and become the new normal of investment management in the future
Interactive and Interpretable Topic Refinement for Analyzing Online Vaccine-Related Narratives
No full textThis research aims to develop highly interpretable models that help generate easy-to-explain data representations of social media texts, which will enhance the interpretability of the online measurement extracted from social media user-generated texts. Such a capacity can benefit our research seeking to measure online engagement and its connection to collective decision-making on societal changes. In this research, we develop an interactive and interpretable framework that allows analysts to identify text with similar or distinct narratives. We use social media text related to the Coronavirus disease 2019 (COVID-19) vaccines as a case study and test the capability of our framework in identifying the Anti-vaccine and Pro-vaccine narratives. Our framework offers two major advantages. First, it leverages semi-supervised topic modeling with deep learning architecture to identify topics that distinguishes between Anti-vaccine and Pro-vaccine posts. Second, it incorporates a constrained hierarchical clustering method that allows human-in-the-loop topic refinement through the system interface, where analysts can explore the relationship of topics via visual representation, verify the labels of post instances, or update labels that are more likely to be incorrect or less certain. Our evaluation shows that the results with refinement significantly improve the topics' coherence and allow for exploring the relationship between Anti-vaccine and Pro-vaccine topics
Interpretations of Classical Liberalism by Scholars in China Taiwan and Hong Kong
No full textThis article intends to explore the issue of the boundaries of freedom through the study of classical liberalism, and uses the works of Yin Hai-guang, a well-known scholar on both sides of the Taiwan Strait, as an example, and the books of Qian Man-su and Lu Xue-kun as reference for interpretation. Firstly, it clarifies the basic meaning of liberalism, its principles and ideological connotations, and takes Yin Hai-guang's ideological context as the research basis to make a holistic exposition of Yin Hai-guang's liberal democratic thoughts, scientific ideas and philosophical thinking.
Afterwards, analyzes the open attitude towards freedom of thought and speech adopted by the Chinese scholar Qian Man-su when he examines the United States on the other side of the Pacific Ocean, gives a bird's-eye view of China' scurrent development of liberalism in the United States, and adopts the role of observation or learning. Then, it is clarified that Hong Kong scholars start from the perspective of European political philosophy, and explore whether the research of Kant's theory of freedom in the Asia-Pacific region has influenced the changes in the liberal thinking of Hong Kong intellectuals.
\ue3 Then it discusses the value and critical theory of liberalism put forward by Chinese scholars, and analyzes the future situation of liberalism, so as to discuss the influence on Taiwan. How will Taiwan respond to China's political thinking in the future? The conclusion is that in an atmosphere between war and peace, the motivation and focus of liberalism, and the reasons for the impact on the interests of both sides, are summarized and analyzed with liberal thinking, and the pursuit of the spirit and influence of classical liberalism
Tsung-Chen Lin Violin Recital Program Notes
No full textThis report is the concert program notes of Tsung-Chen, Lin\ue2s Violin Recital. The repertoire selected for this recital are all for solo violin, included Eug\uc3\ua8ne-Auguste Ysa\uc3\ubfe\ue2s Sonata for Solo Violin in E Minor, Op. 27, No. 4, Johann Sebastian Bach\ue2s \ue2Ciaccona\ue2, from Partita for Solo Violin No.2 in D Minor, BWV 1004 and Sonata for Solo Violin No.2 in A Minor, BWV 1003.
This report is divided into 4 chapters. Chapter 1 explains the motive, the purpose and the approach of this research. Chapter 2 introduces Ysa\uc3\ubfe\ue2s Sonata for Solo Violin in E Minor, Op. 27, No. 4, including the composer\ue2s life, compositional background, compositional techniques and the analysis of the work. Chapter 3 introduces Bach\ue2s Violin Partita and Sonata, including the composer\ue2s life, compositional background, compositional techniques and the analysis of the two works. Chapter 4 is the conclusion
Modern Benthic Foraminiferal Assemblages in the Southern Taiwan Strait in Response to the Hydrological Environment
No full textThe purpose of this study is to investigate the distribution of benthic foraminiferal assemblages in the southern region of the Taiwan Strait, particularly around the Taiwan Banks. The strategy is using statistical analysis to explore the relationship among the distribution of foraminiferal assemblages, population structures and hydrographic settings. The Taiwan Banks is a convex terrain located in the southern Taiwan Strait, where hydrological conditions strongly influenced by monsoon driving currents, the Pearl River plume, and the upwelling in summer. Surface sediments were collected around Taiwan Banks during four cruises: OR1-1071 (April 22nd-24th, 2014), NOR1-0029 (April 20th-23rd, 2022), LGD-2207 (June 29th- August 3rd, 2022), and NOR3-0113(August 16th-17th, 2022). All samples were stained with Rose-Bengal to distinguish shells between living (fresh) and dead (empty) foraminifera.
The results indicated that the population structure of benthic foraminifera exhibits uneven spatial distribution, in terms of total shell abundance and stained shell concentration. Generally, there is a higher value toward the eastern part of Taiwan Banks, where is also within the influence of the upwelling system. Cluster analysis classified all the sampling stations into four clusters, in which the cluster B2 includes stations showing a distribution pattern similar to the upwelling region near the eastern Taiwan Banks. Canonical Correspondence Analysis (CCA) results revealed the association between foraminiferal assemblages of each cluster and upwelling hydrological signature. CCA also characterized cluster B2 with higher bottom water fluorescence values, reflecting the impact of the upwelling on foraminiferal assemblages. The results of Pearson correlation analysis indicated that benthic foraminifera were simultaneously influenced by river plume transported from Pearl River estuary and upwelling during summer, resulting in a significant increase in the population of benthic foraminifera.
The aforementioned results revealed a positive association between upwelling around the eastern Taiwan Banks and benthic foraminiferal communities. This creates an important baseline for future studies about the relationship between environmental fluctuations and benthic foraminifera in the Taiwan Strait
Recommendation Systems based on Non-negative Matrix Factorization-A Case Study of Pet E-commerce Platforms
No full textAbstract
This research focuses on the flourishing development of the e-commerce sector and the significance of personalized recommendation systems. With the rapid expansion of e-commerce, businesses are challenged to shift from traditional mass production to personalized customization. This transformation requires companies to offer products that cater to diverse customer needs, and recommendation systems have become an effective way to address the challenges posed by the abundance of information. Customers often deal with a vast amount of information when selecting products that meet their needs, and recommendation systems assist them in filtering and suggesting items that align with their preferences, providing a personalized shopping experience.
The application of recommendation systems is crucial in the realm of e-commerce, not only enhancing customer satisfaction but also contributing to increased sales. Through recommendation systems, businesses can gain a better understanding of customer preferences and needs, offering more precise product recommendations to strengthen customer loyalty. Therefore, recommendation systems have become a vital business tool capable of reshaping the e-commerce landscape. Major e-commerce websites widely implement recommendation systems to improve customer shopping experiences and boost sales. Streaming platforms such as Netflix, YouTube, Spotify, and others also leverage recommendation systems to enhance user engagement.
Given the thriving development in related industries, such as pet food and supplies, and the increasing maturity of e-commerce technology, the application of recommendation systems in this field becomes a topic worthy of in-depth research. This paper aims to utilize the "approach based on sparse non-negative matrix factorization." Using examples of NMF (Non-negative Matrix Factorization) and nsNMF (Nonsmooth Non-negative Matrix Factorization), we explore the practical application results of recommendation systems and analyze the variations in their applications
The Impact of Symbolic Meanings and Functional Values of Battery Electric Vehicles on Attitude and Purchase Intention through Self Images Set Up
No full textWidespread attention and heated discussions have been attracted in global warming and air pollution caused by transportation carbon emissions, in which the choice of private vehicles directly affects the carbon emissions of personal travel, and many countries are trying to promote electric vehicles to replace traditional fuel vehicles to alleviate the problem of transportation carbon emissions. In the past, the promotion of electric vehicles (EV) has touted their green features, but with the withdrawal of subsidies and the high price of battery electric vehicles (BEV), it is necessary to find new incentives for consumers to purchase electric vehicles.
A survey method has been used in this research with existing BEV users and BEV non-users in Taiwan as sample subjects, to explore whether consumers' perceptions of the symbolic value and functional value of BEV, and after they perceive the symbolic value and functional value of BEV, whether they will generate the perceptions of self-image, and change attitude and purchase intention on three kinds of self-image influence.
The results of the study showed that the technical performance of BEV had the most significant influence on attitudes and purchase intention through self-image, while different symbolic values had different degrees of influence on self-image. In terms of BEV user groups, users of BEV clearly perceive the symbolic values more than non-users, while non-users perceive the technical performance of BEV more than users, and further influence attitudes and purchase intentions through self-image. It is hoped that the perception of function and symbolism by existing BEV users can be used to identify new favorable drivers of EV consumption. This study provides theoretical and practical support for the psychological segmentation of the target group in the marketing of pure EVs, which is expected to contribute to the marketing aspect of EV and promote the popularity of EV in the future
Salinization-induced changes in the growth and rhizosphere bacterial communities of Chenopodium formosanum
No full textEnvironmental changes have become severe global issues, among which soil salinization is one of the critical changes. Searching for potential salt-tolerant crops and studying these crops are essential.
Chenopodium formosanum Koidz., djulis, is one of the important indigenous crops in Taiwan and is a potential salt-tolerant crop. My research aims to explore the changes in the growth of djulis under different salinization environments, to establish a BBCH scale for djulis as a reference for future research, and to investigate the impacts in rhizosphere bacterial communities in different salinization levels.
My results show that the seed germination rate of djulis dropped from 70% to 13% while the salinization levels increased from 0 mM to 500 mM. This indicates that the salt tolerance of djulis is excellent. My data also report that the growth rates of djulis are slower at higher salinization levels. A BBCH scale for dulis was established based on my growth data. In addition, my results suggested that the rhizosphere bacterial diversity decreases when the soil salinization level increases. In highly saline soil, the abundance of salt-tolerant and halophilic bacteria, like Proteobacteria and Actinobacteria, increases. Further studies would be needed to see how the rhizosphere bacterial communities in saline soil shape the growth and development of djulis