Ege University Institutional Repository
Not a member yet
116239 research outputs found
Sort by
Türkçe klinik raporlarda isimlendirilmiş varlık tanıma
No full textİsimlendirilmiş Varlık Tanıma (İVT), dilin karmaşık sondan eklemeli mor- folojisi ve açıklamalı, etki alanına özgü veri kümelerinin süregelen eksikliği nedeniyle Türkçe klinik metinlere uygulanması zordur. Bu tez, bir "Geri Ge- tirme ve Düşünce Zinciri" çerçevesini uygulayarak ve genişleterek, bu bağlamda az atımlı İVT görevini iyileştirmeye yönelik geniş ölçekli bir çalışma sunmak- tadır. Deneysel değerlendirme için kullanılan iki veri kümesi, 2010 i2b2/VA Challenge veri kümesinin profesyonel olarak çevrilmiş Türkçe versiyonunu ve orijinal İngilizce versiyonunu içermektedir. İVT için odak noktası problem, test ve tedavi varlıklarıdır. Tasarım metodolojimiz, bir Büyük Dil Modelinin (BDM), bu durumda GPT-4o'nun muhakeme gücünü ve dinamik ve bağlama bağlı örnekleri analojik olarak geri getirme sürecini içermektedir. Birkaç örneğin geri getirilmesiyle, BİÖ, BDM'den faydalanır ve varlıkların sınırlarını doğru bir şekilde tanımlamak için daha açık bir akıl yürütme süreci boyunca modele rehberlik etmek üzere DZİ'yi sistematik olarak araştırır. Bu çalışma, RT'nin Geri Getirici bileşenini özellikle dikkate değer olarak tanımlamıştır. Geri Getirici, en alakalı örnekleri almak için çeşitli Cümle Gömüleri modellerini karşılaştıran oldukça kapsamlı bir deneyin odak noktasıdır. İlk örnek olan Türkçe veri kümesinde, RT boru hattını genel bir BERTurk tabanlı Cümle-BERT (SBERT) modeli ile birlikte kullandık ve 0,6656'lık bir Makro-F1 skoru elde ettik. Bunun ardından özel bir SimCSE (Simple Karşıt Öğrenme Cümle Gömüleri) modeli geliştirildi ve bu model daha sonra Türkçe Doğal Dil Çıkarımı (DDÇ) ve BERTurk kullanılarak etki alanından klinik cümle çiftleri kullanılarak ince ayar yapıldı. Bu özel model 0,6525'lik bir Makro-F1 puanı üretmiştir. Bu, özellikle düşük kaynaklı, morfolojik olarak karmaşık bir ortamda çalışırken, genel modellere karşı özel geri getiricileri optimize etmenin son derece incelikli zorluğunu yansıttığı için önemlidir. Ayrıca, orijinal İngilizce i2b2 2010 Veri Kümesi; BioBERT'e dayalı bir Geri Getirici modeline ek olarak aynı RT Boru Hattı ile 0,7475'lik bir Macro-F1 puanı elde edebildik ve iyi kaynaklara sahip bir dil için performans ölçütü olarak hizmet ettik. Sonuçlar, her bir yapılandırma için 500 test örneğinin işlenmesinden elde edilen nicel özellikleri (Kesinlik, Geri Çağırma ve F1 puanları) göstermektedir. Genel olarak, RT çerçevesinin performansı umut verici görünmektedir. Türkçe için SimCSE modeli de özel olarak ayarlanmış olsa da umut vericidir. Ayrıca veriler, genel SBERT modellerinin rekabetçi performansının yanı sıra Türkçe klinik metinler için etki alanı Karşıt Öğrenmenin daha fazla araştırılmasına yönelik fırsatları da göstermektedir. Bu çalışma, çeşitli geri getirici stratejilerinin kullanımının etkilerine ve Düşünce Zinciri İsteminin avantajlarına ilişkin değerli nicel kanıtlar içermekte ve Türkçe klinik metinler gibi klinik açıdan zengin, morfolojik açıdan zengin, düşük kaynaklı etki alanları için daha güvenilir İVT sistemlerinin nihai tasarımına katkıda bulunmaktadır.Named Entity Recognition (NER) is difficult to apply to Turkish clinical texts due to the language's complex agglutinative morphology and an ongoing lack of annotated, domain-specific datasets. This thesis presents a large-scale study on improving the few-shot NER task in this context by implementing and extending a "Retrieving and Chain-of-Thought" framework. The two datasets used for empirical evaluation include a professionally translated Turkish version of the 2010 i2b2/VA Challenge dataset and the original English version. The focus is on NER for the problem, test, and treatment entities. Our design methodology incorporates the reasoning power of a Large Language Model (LLM), in this case, GPT-4o and a process of analogically retrieving examples that are dynamic and context-dependent. By retrieval of a few examples, the ICL exploits the LLM and systematically investigated CoT prompting to guide the model through a more overt reasoning process to accurately delineate the boundaries of the entities. This study identified the retriever component of RT as especially notable. The retriever is the focus of a quite thorough experiment comparing various sentence embedding models for retrieving the most relevant examples. In the first example, the Turkish dataset, we used the RT pipeline with a general BERTurk-based Sentence-BERT (SBERT) model for retrieving and achieved a Macro-F1 score of 0.6656. This was followed by the development of a custom SimCSE (Simple Contrastive Learning of Sentence Embeddings) model, which was then fine-tuned using Turkish Natural Language Inference (NLI) and clinical sentence pairs from the domain using BERTurk. This specific model generated a Macro-F1 score of 0.6525. This is important because it reflects the highly nuanced challenge of optimizing specialized retrievers against general models, particularly when working in a low-resource, morphologically complex environment. Furthermore, on the original English i2b2 2010 dataset; with the same RT pipeline in addition to a retriever model based on BioBERT we were able to achieve a Macro-F1 score of 0.7475, and serve as a performance benchmark for a well-resourced language. The results show the quantitative aspects obtained (Precision, Recall and F1-scores) from processing 500 test samples for each configuration. Overall, the performance of the RT framework appears promising. The SimCSE model for Turkish, though custom-tuned, is also promising. Additionally, the data also shows the competitive performance of general SBERT models along with opportunities for further investigating domain specific contrastive learning for Turkish clinical text. The work includes valuable quantitative evidence of the effects of using the various retrieval strategies and the advantages of CoT prompting, contributing to the eventual design of more reliable NER systems for clinically-resourced, morphologically-rich, low-resource, domains such as Turkish clinical texts
Predicting immune-related adverse events in patients with melanoma: the role of interleukin-7 rs16906115 polymorphism and lymphocyte dynamics
No full textIntroduction: Immune checkpoint inhibitors (ICIs) have transformed the therapeutic landscape of malignant melanoma; however, they are frequently associated with immune-related adverse events (irAEs). Emerging evidence suggests that genetic predispositions, including interleukin-7 (IL-7) gene variants, may influence the risk of these toxicities. Methods: In this single-center retrospective study, we investigated the potential utility of IL-7 rs16906115 polymorphism and lymphocyte stability index (LSI) in predicting susceptibility to irAEs among 96 melanoma patients treated with ICIs. Results: Genotyping revealed a minor allele frequency of 8.3% for rs16906115. Logistic regression analysis indicated that carriers of the minor allele had a significantly increased risk of all-grade irAEs compared to reference allele carriers (adjusted OR: 3.93; 95%CI:1.13–13.64; p=0.031). Subgroup analyses revealed a significant increase in risk across endocrine, non-cutaneous, multiple, low-grade, and early onset (<3 months) irAEs. While neither baseline lymphocyte count nor LSI predicted overall irAE incidence, an elevated LSI emerged as a key risk factor for early steroid-requiring irAEs (adjusted OR:3.79; 95% CI: 1.14–12.61; p =0.030). Discussion: These findings from a Turkish cohort corroborate earlier European studies suggesting that rs16906115 minor allele carriage may be a genetic risk factor for irAEs. Furthermore, LSI may serve as a dynamic biomarker for predicting early steroid-requiring irAEs. Prospective multicenter studies among diverse populations are warranted to validate these findings. © 2025 Elsevier B.V., All rights reserved
Can Asclepius in Mythology Be a Role Model for Today's Physicians?
No full textTo be a good physician, role models are needed who have the characteristics of a 'good physician' and have professional, ethical and professional values. The purpose of this study is to seek an answer to the question of whether history and medical symbols can be used to determine what characteristics a role model has and who should be taken as a role model. In mythology, Asclepius is known as the god of health and healing. When the characteristics of Asclepius with his snake staff, that he brought from history are investigated, we come across the portrait of a physician, who is well-educated, reliable, respected, careful, quick to make decisions, open to innovation, a lifelong learner, and interested in health. In fact, these characteristics are the basic characteristics desired in a 'good physician' and are valid for all physicians today, including neuroscientists such as neurologists, neurosurgeons, and neuroradiologists. © 2025 Elsevier B.V., All rights reserved
Borax pentahydrate as a promising boron-based angiogenesis inhibitor
No full textBackground: Boron, a trace element, is involved in various physiological and metabolic processes, and recent studies suggest that boron compounds may have potential in cancer prevention and treatment. In this study, the antiangiogenic effects of a boron compound, borax pentahydrate (BPH), were investigated. Angiogenesis is a tightly regulated biological process responsible for the formation of new blood vessels from existing vasculatures. This process plays a critical role in cancer progression, making it an important target for cancer therapy. Pancreatic and kidney cancers are difficult to treat because they are aggressive and resistant to chemotherapy. Methods: The antiproliferative activity was evaluated using the MTT assay, while antiangiogenic effects were tested through in vitro tube formation assays and in ovo chick chorioallantoic membrane (CAM) assay. The effect of BPH on VEGF levels was determined using Western blot analysis in HUVEC, ACHN, PANC-1 cells. The effect of BPH on tumor angiogenesis was investigated with an in vivo Ehrlich ascites carcinoma model (EAC). Results: BPH exhibited potent antiproliferative and antiangiogenic activities, inhibiting the proliferation of ACHN, PANC-1, and HUVECs, disrupting endothelial tube formation, and inhibiting vascular formation on the CAM surface in a dose-dependent manner. VEGF levels were significantly decreased in ACHN, PANC-1 and HUVECs. There was also a decrease in VEGF and TGF-beta 1 levels in BPH-treated tumor groups. In addition, BPH caused a decrease in tumor size. Conclusion: These findings suggest that BPH may be a new antiangiogenic and antitumoral agent. BPH may contribute to drug development studies targeting angiogenesis-related diseases as a promising new therapeutic agent.Ministry of National Education of the Republic of TurkeyThis study was funded by Ministry of National Education of the Republic of Turkey. (Since the responsible author works as a teacher at the Ministry of National Education of the Republic of Turkey, she was financially supported for the realization of this project without any contract)
Türkiye balıkçılık ve yetiştiricilik sektöründe iş sağlığı ve güvenliğine ilişkin parametreler üzerine değerlendirmeler
No full textBu tezde Su ürünleri sektörü günümüzde hala gelişmekte olup, son derece önemli bir öneme sahiptir. Su Ürünleri sektörü artan üretim ile beraber birtakım iş sağlığı ve Güvenliği açısından sorunları beraberinde getirmektedir. Su ürünleri sektörü avcılık, yetiştiricilik ve işlemeyle beraber yoğun istihdamı olan bir sektördür. Nitelikli ve niteliksiz olarak çalışan sayısının fazlalığı iş sağlığı ve güvenliği nezdinde tehlikeleri ve kaza oluşumlarını arttırmaktadır. Balıkçılık ve ilgili sektörler yaygın olarak en tehlikeli mesleklerden biri olarak kabul edilmektedir. Mevcut çalışmanın amacı, Türk balıkçılık sektöründeki bazı iş sağlığı ve güvenliği parametrelerini (olay, geçici iş göremezlik ve ölüm oranları) hakkında istatiki anlamda değerlendirmek ve 2013'ten 2022'ye kadar sektörler arasında bir karşılaştırma yapmaktır. Türk balıkçılık sektöründe 10 yıllık süre zarfında 27 ölüm ve 76798 gün geçici iş göremezlik vakasıyla sonuçlanan toplam 14585 iş kazası bildirilmiştir. Su ürünleri sektöründe yaşanılan iş kazası Olay oranları son dokuz yılda 87,9 (balıkçılık, 2017) ile 32955,22 (işleme, 2022) arasında değişiklik göstermiştir. Su ürünleri yetiştiriciliği ve işleme sektörlerinin ölüm oranları, Türkiye'deki tüm sektörler için hesaplanan oranlardan daha yüksektir. Su ürünleri yetiştiriciliği ve işleme sektörlerinde balıkçılığa kıyasla daha küçük bir iş gücü olmasına rağmen, su ürünleri yetiştiriciliği ve işleme alanındaki kümülatif iş kazası ve ölüm sayısı balıkçılıktakinden fazladır. Bu çalışmanın bulguları, su ürünleri yetiştiriciliği ve işleme sektörlerinin, olay sayısı, geçici iş göremezlikler ve ölümler nedeniyle balıkçılıktan daha riskli olduğunu göstermiştir. Bununla birlikte, ölüm biçimleri, yaralanma türleri, aylara göre olaylar ve her sektöre göre yaralanan vücut parçaları gibi önemli ayrıntıların eksikliği nedeniyle, su ürünleri sektöründe güvenlik prosedürleri hala tartışılmalıdır. İşyeri olaylarının kapsamlı bir şekilde raporlanmasına yönelik iş veren ve kamu istatistikleri doğru giriş sağlanması için iş verenlerin teşvik edilmesi, Türkiye'nin balıkçılık sektöründe güvenlik bilincini artırmak için hayati öneme sahiptir. Veri toplamaya yönelik daha uygun bir yaklaşım benimsenmesi, hem yetkililer hem de şirketler için sektörde kazaları, yaralanmaları ve ölümleri önlemeye yardımcı olacak önemli bir kaynak olacaktır.In this thesis, the aquaculture sector is still developing today and has a very important importance. The aquaculture sector brings with it some problems in terms of occupational health and safety with increasing production. The aquaculture sector is a sector with intensive employment together with hunting, breeding and processing. The high number of qualified and unqualified employees increases the dangers and accident occurrences in terms of occupational health and safety. Fishing and related sectors are widely accepted as one of the most dangerous professions. The aim of the present study is to statistically evaluate some occupational health and safety parameters (incident, temporary disability and death rates) in the Turkish fisheries sector and to make a comparison between the sectors from 2013 to 2022. A total of 14585 occupational accidents were reported in the Turkish fisheries sector in a 10-year period, resulting in 27 deaths and 76798 days of temporary disability. Occupational accidents in the aquaculture sector The incident rates have varied between 87.9 (fisheries, 2017) and 32955.22 (processing, 2022) in the last nine years. The mortality rates of the aquaculture and processing sectors are higher than the rates calculated for all sectors in Türkiye. Although the aquaculture and processing sectors have a smaller workforce compared to fishing, the cumulative number of occupational accidents and deaths in aquaculture and processing is higher than in fishing. The findings of this study showed that the aquaculture and processing sectors are riskier than fishing due to the number of incidents, temporary disabilities, and deaths. However, safety procedures in the aquaculture sector are still debatable due to the lack of important details such as the types of deaths, types of injuries, incidents by month, and injured body parts by sector. Encouraging employers to ensure accurate entry of employer and public statistics for comprehensive reporting of workplace incidents is vital to increasing safety awareness in Türkiye's fisheries sector. Adopting a more tailored approach to data collection will be an important resource for both authorities and companies to help prevent accidents, injuries and deaths in the industry
Bi-allelic pathogenic variants in TRMT1 disrupt tRNA modification and induce a neurodevelopmental disorder
No full textThe post-transcriptional modification of tRNAs plays a crucial role in tRNA structure and function. Pathogenic variants in tRNA-modification enzymes have been implicated in a wide range of human neurodevelopmental and neurological disorders. However, the molecular basis for many of these disorders remains unknown. Here, we describe a comprehensive cohort of 43 individuals from 31 unrelated families with bi-allelic variants in tRNA methyltransferase 1 (TRMT1). These individuals present with a neurodevelopmental disorder universally characterized by developmental delay and intellectual disability, accompanied by variable behavioral abnormalities, epilepsy, and facial dysmorphism. The identified variants include ultra-rare TRMT1 variants, comprising missense and predicted loss-of-function variants, which segregate with the observed clinical pathology. Our findings reveal that several variants lead to mis-splicing and a consequent loss of TRMT1 protein accumulation. Moreover, cells derived from individuals harboring TRMT1 variants exhibit a deficiency in tRNA modifications catalyzed by TRMT1. Molecular analysis reveals distinct regions of TRMT1 required for tRNA-modification activity and binding. Notably, depletion of Trmt1 protein in zebrafish is sufficient to induce developmental and behavioral phenotypes along with gene-expression changes associated with disrupted cell cycle, immune response, and neurodegenerative disorders. Altogether, these findings demonstrate that loss of TRMT1-catalyzed tRNA modifications leads to intellectual disability and provides insight into the molecular underpinnings of tRNA-modification deficiency caused by pathogenic TRMT1 variants.Health & Care Research Wales, Epilepsy Research UK; Swansea University PhD fundingThe authors thank the affected individuals and their families for their support of this study. One of the authors of this publication (Z.T.) is a member of the European Reference Network on Rare Congenital Malformations and Rare Intellectual Disability, ERN-ITHACA (EU Framework Partnership Agreement ID: 3HP-HP-FPA ERN-01-2016/739516) . B.V. is a member of the European Reference Network on Rare Congenital Malformations and Rare Intellectual Disability (ERN-ITHACA) (EU Framework Partnership Agreement ID: 3HP-HP-FPA ERN-01-2016/739516) .The research in this paper was supported by NIH GM141038 to D.F. Studies performed in the lab of G.K.V. was funded by NIH/ORIP R24OD034438. The clinic-genetic research was funded in part by the Wellcome Trust (WT093205MA and WT104033AIA) . This study was funded by the Medical Research Council (MR/S01165X/1, MR/S005021/1, and G0601943) , The National Institute for Health Research University College London Hospitals Biomedical Research Centre, Rosetrees Trust, Ataxia UK, Multiple System Atrophy Trust, Brain Research United Kingdom, Sparks Great Ormond Street Hospital Charity, Muscular Dystrophy United Kingdom (MDUK) , Muscular Dystrophy Association (MDA USA) , and the King Baudouin Foundation. S.E. and H.H. were supported by an MRC strategic award to establish an International Centre for Genomic Medicine in Neuromuscular Diseases (ICGNMD) MR/S005021/1. B.V. was supported by the Deutsche Forschungsgemeinschaft (DFG) DFG VO 2138/7-1 grant 469177153. J.S. is supported by Cancer Research UK and University College London. A.F. and S.C. were supported by Health & Care Research Wales, Epilepsy Research UK, and Swansea University PhD funding
A Comparative Review of Fuzzy Reinforced Search Algorithms: Methods and Applications
No full textEngineering optimization provides efficient designs that balance performance with resource demand. Metaheuristic algorithms excel at this task, but their lack of adaptability across different problems limits their search capability. In this regard, integrating these methods with auxiliary decision-making mechanisms based on fuzzy logic can considerably improve their search ability. Fuzzy logic empowers these algorithms to adapt their search behavior dynamically based on specific problem characteristics. The current study assesses how this integration improves search efficiency and adaptability to complex and uncertain scenarios, ultimately leading to more effective solutions in engineering optimization. To this end, different fuzzy-reinforced metaheuristic approaches are evaluated, and their search capabilities are compared among themselves and against their standard versions. The selected methods were thoroughly assessed from diverse aspects, including search performance, behavioral process, computational cost, and stability across various problems (e.g., mathematical, mechanical, and structural problems). The acquired results are reported and discussed in detail. Consequently, the attained outcomes indicate that a proper fuzzy-based decision mechanism can considerably improve the search capability of metaheuristic algorithms.Scientific and Technological Research Council of Turkiye (TUBITAK)Open access funding provided by the Scientific and Technological Research Council of Turkiye (TUBITAK). No funding was received for conducting this study
Global Roadkill Data: a dataset on terrestrial vertebrate mortality caused by collision with vehicles
No full textRoadkill is widely recognized as one of the primary negative effects of roads on many wildlife species and also has socioeconomic impacts when they result in accidents. A comprehensive dataset of roadkill locations is essential to evaluate the factors contributing to roadkill risk and to enhance our comprehension of its impact on wildlife populations and socioeconomic dimensions. We undertook a compilation of roadkill records, encompassing both published and unpublished data gathered from road surveys or opportunistic sources. GLOBAL ROADKILL DATA includes 208,570 roadkill records of terrestrial vertebrates from 54 countries across six continents, encompassing data collected between 1971 and 2024. This dataset serves to minimise the collection of redundant data and acts as a valuable resource for local and macro scale analysis regarding rates of roadkill, road- and landscape-related features associated with risk of roadkill, vulnerability of species to road traffic, and populations at risk of local extinction. The objective of this dataset is to promote scientific progress in infrastructure ecology and terrestrial vertebrate conservation while limiting the socio-economic costs. © 2025 Elsevier B.V., All rights reserved
The Urogenital System Microbiota: Is It a New Gamechanger in Urogenital Cancers?
No full textThe human microbiome, which encompasses microbial communities and their genetic material, significantly influences health and disease, including cancer. The urogenital microbiota, naturally present in the urinary and genital tracts, interact with factors such as age, lifestyle, and health conditions to affect homeostasis and carcinogenesis. Studies suggest that alterations in this microbiota contribute to the development and progression of genitourinary cancers, emphasizing the concept of oncobiome, which refers to microbial genetic contributions to cancer. Similarly, gut microbiota can influence hormone levels and systemic inflammation, impacting cancers such as cervical and prostate cancer. Advanced studies indicate that microbial communities in genitourinary cancers have distinct profiles that may serve as diagnostic biomarkers or therapeutic targets. Dysbiosis of the urinary microbiota correlates with bladder and kidney cancer. Additionally, gut microbiota influence the effectiveness of cancer treatments. However, further research is necessary to clarify causality, the role of microbial metabolites, and hormonal regulation. The aim of this review is to understand that these dynamics present opportunities for innovative cancer diagnostics and therapies, highlighting the need for integration of microbiology, oncology, and genomics to explore the role of microbiota in genitourinary cancers. For this, a comprehensive search of relevant databases was conducted, applying specific inclusion and exclusion criteria to identify studies examining the association between microbiota and urogenital cancers. Research into the mechanisms by which microbiota influence urogenital cancers may pave the way for new diagnostic and therapeutic approaches, ultimately improving patient outcomes.KLIMUD (Society for Clinical Microbiologists of Turkiye); KLIMUD-Microbiome Restoration and Bioinformatics Study GroupWe would like to acknowledge the KLIMUD (Society for Clinical Microbiologists of Turkiye) and the KLIMUD-Microbiome Restoration and Bioinformatics Study Group for critical organization and support
Does optic nerve sheath fenestration alleviate headaches in patients with idiopathic intracranial hypertension? A retrospective study
No full textObjective: To evaluate headache course after optic nerve sheath fenestration (ONSF) in idiopathic intracranial hypertension (IIH) patients. Methods: In this retrospective cohort study, IIH patients' data who had undergone ONSF due to progressive visual field loss despite medical treatment or advanced visual field loss at onset, including age, sex, cerebrospinal fluid (CSF) opening pressure, laterality of ONSF (unilateral or bilateral), best corrected visual acuity (BCVA), visual field mean deviation (MD) preoperatively and after ONSF in the worse and fellow eyes, were noted. Headache severity assessed using the visual analog scale, and headache frequency defined as the number of headache days per month before and after surgery were considered. Results: 20 patients (17 females, 3 males) with a mean age of 32.4 years and a mean CSF opening pressure of 492.79 mm H2O were studied. ONSF was unilateral in 12 (60 %) and bilateral in 8 (40 %) patients. Median LogMAR BCVA and MD in the worse eye improved significantly after surgery (p = 0.04 and p = 0.02, respectively), whereas improvement in the fellow eye was not significant (p > 0.05). Though headache severity and frequency decreased after surgery (p = 0.01 and p = 0.001, respectively) it was still present in 12 patients (60 %). Age, sex, CSF opening pressure, uni/bilateral ONSF, BCVA or MD in the worse or fellow eye had no effect on the course of headache severity or frequency. Conclusion: Specific headache treatment is essential in IIH, since lowering intracranial pressure by ONSF is not associated with headache relief in over half of the patients