English

Suomi

tulokset

Silmäile

Julkaisuarkisto

UTUPub on Turun yliopiston avoin julkaisuarkisto, joka sisältää Turun yliopiston tiedekuntien ja laitosten julkaisuja ja opinnäytteitä kuten väitöskirjoja, lisensiaatintöitä, pro gradu -tutkielmia ja niiden tiivistelmiä sekä kandidaatintöitä.

Kokoelmat

Näytetään 1 - 4 / 4

Kirjat ja opinnäytteet

UTUCRIS-artikkelit

Digitoidut aineistot

Viimeksi lisätyt

Prognostic role of histological depth of invasion in T1-2 oropharyngeal squamous cell carcinoma
Nissi, Linda; Elseragy, Amr; Wahab, Awais; Wennerstrand, Pia; Knuutila, Maisa; Routila, Johannes; Huusko, Teemu; Irjala, Heikki; Salo, Tuula; Ventelä, Sami (Elsevier)
Background
Depth of invasion (DOI) is a well-established prognostic factor in oral cavity squamous cell carcinoma (OSCC). Incorporating DOI into the OSCC staging system has substantially refined T-classification and now informs key management decisions in patients with early-stage (T1–T2) disease. In contrast, the prognostic significance of DOI in oropharyngeal squamous cell carcinoma (OPSCC) remains relatively unclear. The aim of this study was to evaluate the prognostic value of DOI in T1–2 OPSCC.
Materials and methods
A population-based cohort representative of 1033 head and neck squamous cell carcinoma patients was used to identify T1–2 OPSCC patients treated with a curative intent and a tumor sample available for DOI measurement.
Results
In this retrospective cohort study of 74 patients, of whom 63.8% (n = 44) were p16-positive, 71.6% (n = 53) had a DOI ≥ 5 mm. While high DOI correlated with heavy alcohol consumption, it did not predict survival outcomes (5-year disease-specific survival HR 1.06; 95% CI 0.94–1.20; p = 0.315) or locoregional metastasis (HR 1.54; 95% CI 0.51–4.63; p = 0.444) in T1–2 tumors of the oropharynx. Instead, within the p16-negative cohort, increased DOI (≥ 5 mm) was associated with an observable tendency toward poorer survival.
Conclusion
DOI was not a prognostic factor in p16-positive T1–2 OPSCC. Nevertheless, DOI may hold prognostic relevance for p16-negative disease. The patterns of local invasion and locoregional spread in OPSCC may reflect distinct biological mechanisms compared to those in other head and neck subsites.
Feasibility of multimodal metabolic analysis for detecting early changes in acute neuroinflammation
Grist, James T.; Evstafev, Ilia; Olesova, Dominika; Nynäs, Signe E.; Orešič, Matej; Dickens, Alex M.; Tyler, Damian J.; Couch, Yvonne (Springer Science and Business Media LLC)
Given the prevalence of metabolic perturbations in a variety of neurological and neurodegenerative diseases, understanding and monitoring brain metabolism is a key step in our advancement of therapies. The details of the citric acid cycle were established at the beginning of the last century but only recently have its metabolic intermediates been observed in vivo in the brain. In this study, we employed orthogonal analyses to investigate metabolic alterations in response to acute neuroinflammation in vivo, demonstrating a multi-technique approach that could be used for future studies.Hyperpolarized [1-¹³C] pyruvate spectroscopy revealed an early decline in pyruvate metabolism via pyruvate dehydrogenase (PDH), leading to reduced ¹³C-bicarbonate formation. This metabolic disruption occurred despite the absence of structural or perfusion changes on conventional MRI. Further analysis of polar metabolites in the ipsilateral hemisphere confirmed ongoing inflammatory processes. These findings highlight the potential of this dual technique approach to inform upon metabolic changes due to neuroinflammation.Combining methods to probe metabolism in invasive (metabolomics) and non-invasive (hyperpolarized MRI) manners, this represents a promising translational approach for real-time metabolic assessments in an area of the body, the brain, where studying processes such as metabolism has traditionally been challenging. This study has demonstrated the approach to monitor changes in metabolism in response to inflammation in the brain.
Adolescents’ Intentions for Upper Secondary Education: The Role of Metacognitive Awareness
Kallio Heli; Kallio Manne; Virta Kalle; Iiskala Tuike (EJournal Publishing)
Application of the REINFORCE Algorithm in Real-Time Inverted Pendulum Control
Paldanius, Kimmo (2026-07-02)
Diplomityö
This thesis investigates the application of the REINFORCE policy-gradient algorithm to upright stabilization of the Quanser Qube-Servo 2 rotary inverted pendulum. Since the balancing task is approximately linear near the upright equilibrium, classical PD control already provides an effective solution. The aim is therefore not to show that REINFORCE outperforms classical control, but to examine whether a simple model-free reinforcement learning method can learn a local balancing policy, how the resulting controller compares with a PD reference controller, and how it behaves outside the training simulator. A Gym-compatible simulation environment was developed for the Furuta pendulum and used as the training environment. A continuous-action Gaussian policy was trained using a Monte Carlo REINFORCE update without a value-function baseline or critic network. The policy receives a compact two-dimensional feature representation formed from the arm and pendulum angles and angular velocities, and outputs the mean control action and state-dependent standard deviation. A simple survival reward encourages the policy to maximize balancing time. Both the learned policy and the PD reference controller were evaluated in the custom simulation, the Quanser virtual environment, and on the physical Qube-Servo 2 hardware. The results show that REINFORCE learns a local stabilizing policy that works in the virtual environment and on the physical device without retraining. Settled-state angular precision was broadly comparable to that of the PD controller, but the learned controller used a more active voltage signal and was less consistently reliable. The work demonstrates the feasibility of REINFORCE for this real-time control task while highlighting its sensitivity to task formulation and the limitations of transferring from simulation to the physical device.Tässä diplomityössä tutkitaan, voidaanko REINFORCE-vahvistusoppimisalgoritmia hyödyntää Quanser Qube-Servo 2 -pyörivän käänteisheilurin pystyasennon tasapainottamiseen. Koska tasapainotustehtävä on likimain lineaarinen pystyasennon tasapainopisteen lähellä, klassinen PD-säätö tarjoaa siihen jo tehokkaan ratkaisun. Työn tavoitteena ei siten ole osoittaa REINFORCE-algoritmin paremmuutta klassiseen säätöön nähden, vaan selvittää, voiko yksinkertainen mallivapaa vahvistusoppimismenetelmä oppia paikalliseen tasapainotukseen soveltuvan ohjauspolitiikan. Lisäksi työssä tarkastellaan, miten opittu säädin vertautuu PD-vertailusäätimeen ja miten se toimii koulutussimulaattorin ulkopuolella. Työtä varten kehitettiin Gym-yhteensopiva simulaatioympäristö pyörivälle käänteisheilurille. Säädin toteutettiin jatkuva-arvoisena stokastisena ohjauspoliti-ikkana, jossa ohjaustoiminto kuvataan gaussisen todennäköisyysjakauman avulla. Ohjauspolitiikka koulutettiin Monte Carlo -REINFORCE-päivityksellä ilman arvofunktioon perustuvaa vertailutasoa tai kriitikkoverkkoa. Ohjauspolitiikka saa syötteenään kaksi piirrettä, jotka muodostetaan varren ja heilurin kulmista sekä kulmanopeuksista, ja tuottaa ohjaustoiminnon keskiarvon sekä tilasta riippuvan keskihajonnan. Yksinkertainen tasapainossa pysymiseen perustuva palkkio ohjaa oppimista kohti mahdollisimman pitkää tasapainotusaikaa. Opittua säädintä ja PD-vertailusäädintä arvioitiin tätä työtä varten kehitetyssä simulaatioympäristössä, Quanserin virtuaaliympäristössä sekä fyysisellä Qube-Servo 2 -laitteella. Tulokset osoittavat, että REINFORCE oppii paikalliseen stabilointiin soveltuvan ohjauspolitiikan, joka toimii virtuaaliympäristössä ja fyysisellä laitteella ilman uudelleenkoulutusta. Vakiintuneessa tilassa kulmatarkkuus oli verrattavissa PD-säätimeen, mutta opitun säätimen ohjausjännite vaihteli enemmän ja sen toimintavarmuus oli heikompi. Työ osoittaa, että REINFORCE-algoritmia voidaan soveltaa tähän reaaliaikaiseen säätötehtävään, mutta tuo samalla esiin menetelmän herkkyyden oppimistehtävän määrittelylle sekä simulaatiosta fyysiselle laitteelle siirrettävyyden rajoitteet.
Design Framework for Trustworthy User Interfaces
Pääkkönen, Iiro (2026-06-30)
Diplomityö
This thesis studies the relationship between user interface (UI) design, microinteractions and perceived trustworthiness in software systems. The thesis combines a narrative literature review with an empirical user study. The literature review gathers together research from trust psychology, human-computer interaction (HCI) and UI design to identify the most critical elements and decisions in UI design that contribute to trust formation. Based on already established trust models, such as the work of Mayer et al. and Hancock et al. the review finds functionality, helpfulness, predictability, visual trust signals and trust reinforcement actions as the core aspects of trustworthy UI design. Based on the literature review, the Trustworthy User Interface Design Framework (TUIDF) was created as a practical checklist for UI designers. To validate the framework, two mobile banking application prototypes were created using Figma Make, an artificial intelligence model made for UI design. The control prototype was designed using traditional UI design principles while the experiemental prototype was designed according to the TUIDF. Ten participants divided into two focus groups completed tasks to evaluate the prototypes with a questionnaire combining the STS and the System Usability Scale (SUS). The TUIDF-prototype scored higher on most dimensions that were measured, including purpose, process, visual appeal and general trust. This suggests that the framework can improve perceived trustworthiness. The findings show that UI design based on the TUIDF produces measurable improvements in user trust, although further validation with a larger sample size is required.

Turun yliopiston kirjasto
Turun yliopisto

julkaisut@utu.fi
Tietosuojaseloste
Saavutettavuusseloste
Evästeasetukset