https://sphere.acg.edu/jspui/handle/123456789/2296 2026-04-15T14:50:31Z https://sphere.acg.edu/jspui/handle/123456789/2366 Title: From pixels to clinical insight: Computer vision and deep learning for automated left ventricular segmentation and ejection fraction prediction in pediatric echocardiography videos Authors: Kostiris, Konstantinos Abstract: This study aims to explore the fusion of medical imaging and advanced machine learning in echocardiography, an essential diagnostic tool for heart assessments, especially in the pediatric domain. While echocardiography is widely used, challenges specific to the pediatric population emphasize the need for innovation. Beginning with investigating the heart’s complex structures, the journey of this research dives into reviewing the existing literature on pediatric echocardiography, addressing its advantages and limitations. Central to the methodology is the first extensive pediatric echocardiographic video dataset available as of early 2023 from Stanford's Center for Artificial Intelligence in Medicine & Imaging, named EchoNet-Pediatric. Expert-annotated echocardiograms lay the foundation for computer vision algorithms and deep learning models on which they are trained and validated. The results, based on key performance metrics, demonstrate the capabilities of integrating Artificial Intelligence in pediatric echocardiography, showcasing the synergy between traditional medical practices and cutting-edge advancements. 2023-12-12T00:00:00Z https://sphere.acg.edu/jspui/handle/123456789/2317 Title: Climate related natural disasters: A crucial challenge for port resilience. A neural network application Authors: Nomikou Lazarou, Eirini Abstract: Current consumption habits are enabled due to the various commercial ports around the world. Goods are transported and traded only due to the existence of ports since the ancient days. However, any port disruptions jeopardize the ordinary consumption patterns. A well know suspect of port operations is climate change. Climate change shifts weather patterns causing more severe and more frequent weather events very often responsible for disturbance of port operations and marine roots. In this context, we investigate how Deep Learning Neural Networks (DLNN), in contrast to the traditional Numerical Weather Prediction (NWP) processes, could offer more accurate weather predictions in port regions preventing major economic losses. This Thesis presents the relative state-of-the-art literature on deep learning weather prediction and constructs 5 days forecasts for the ten biggest US commercial ports for 2023. 2023-07-18T00:00:00Z https://sphere.acg.edu/jspui/handle/123456789/2310 Title: Deep Learning Algorithms for Early Diagnosis of Acute Lymphoblastic Leukemia Authors: Papaioannou, Dimitrios Abstract: Acute lymphoblastic leukemia (ALL) is a form of blood cancer that affects the white blood cells. ALL constitutes approximately 25% of pediatric cancers. Early diagnosis and treatment of ALL are crucial for improving patient outcomes. The task of identifying immature leukemic blasts from normal cells under the microscope can prove challenging, since the images of a healthy and cancerous cell appear similar morphologically. In this study, we propose a binary image classification model to assist in the diagnostic process of ALL. Our model takes as input microscopic images of blood samples and outputs a binary prediction of whether the sample is normal or cancerous. Our dataset consists of 10661 images out of 118 subjects. Deep learning techniques on convolutional neural network architectures were used to achieve accurate classification results. Our proposed method achieved 94.3% accuracy and could be used as an assisting tool for hematologists trying to predict the likelihood of a patient developing ALL. 2023-01-01T00:00:00Z https://sphere.acg.edu/jspui/handle/123456789/2309 Title: Graph-based inference: A case study in identifying potential drug candidates for the treatment of schizophrenia, major depressive disorder & anxiety disorders Authors: Ntokou, Eleni Abstract: Drug repurposing involves exploring new pharmaceutical purpose for already approved drugs. The drug development process comes with a high development risk, as it is demanding in terms of both time and cost. Drug repurposing tries to address these issues, as it focuses on drugs already on the market, thus alleviating the need for clinical trials to assess the already established safety profiles of drugs, requires less resources and is historically associated with higher success rates. In the present work, a knowledge graph has been created with the intention to identify potential drug candidates for schizophrenia, major depressive disorder, and anxiety disorders, with the use of ontology-based reasoning. The resulted knowledge graph involves seventy-three classes with three of them being the defined classes, designed to generate new knowledge. Twelve thousand nine hundred twenty five individuals are imported to the ontology from Kyoto Encyclopedia of Genes and Genomes and DrugBank databases containing genes and proteins, drugs, biological pathways, and the aforementioned human diseases. To this end, two potential candidate drugs for schizophrenia were identified, as well as four for major depressive disorder – two of which are already in clinical trials, emerged. For anxiety disorders, the candidates retrieved were disproportionately high in prevalence, suggesting that further investigation is required to make the inference process more selective. 2023-02-16T00:00:00Z