Πλοήγηση ανά Συγγραφέας "Tsiknakis, Emmanouil"

Τώρα δείχνει 1 - 4 of 4
  • Μικρογραφία εικόνας
    Τεκμήριο
    Design and implementation of an application in Flutter for monitoring and analysis of physical activity and stess levels using Fitbit.
    (ΕΛΜΕΠΑ, Σχολή Μηχανικών (ΣΜΗΧ), Τμήμα Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών, 2024-10-21) Melakis, Nikolaos; Μελάκης, Νικόλαος; Tsiknakis, Emmanouil; Τσικνάκης, Εμμανουήλ
    A greater understanding of the significance of holistic well-being has emerged as a result of the modern lifestyle, which is characterized by quick technology improvements and the demands of contemporary society. Built on the adaptable Flutter framework, Stress.io emerges as a trailblazing mobile application that smoothly integrates with Fitbit smartwatches. In addition to monitoring heart rate and physical activity, this application goes beyond traditional well-being tools by introducing a ground-breaking stress calculation capability. Stress.io is proof of the technological prowess made possible by the Flutter platform. Flutter's hot reload functionality, which was designed with cross-platform compatibility in mind, permitted quick iterations and made sure that the user experience was consistent across different platforms. The user interface, created using a wealth of Flutter widgets, offers not just a pleasing appearance but also a simple and straightforward navigating experience. Fitbit smartwatch integration is a distinguishing feature of Stress.io, enhancing its potential as a complete health companion. Users of Stress.io can synchronize their heart rate and physical activity data in real-time thanks to Fitbit, a major participant in the wearable technology market. By utilizing the accuracy of Fitbit's sensors, this integration gives customers a thorough overview of their health. The partnership between Stress.io and Fitbit places the app at the forefront of technologically advanced well-being solutions. Finally, Stress.io serves as an example of a technology vision for the future of well-being. It highlights how technology may be used to track data as well as gain insightful knowledge and make useful decisions. With the combination of Flutter's adaptability, Fitbit's accuracy, and revolutionary stress computation, Stress.io is positioned as a leader at the nexus of technology and well-being. The app's ultimate objective will always be to empower people on their path to holistic health and wellbeing.
  • Μικρογραφία εικόνας
    Τεκμήριο
    Multimodal pain intensity assessment based on physiological biosignals.
    (ΕΛΜΕΠΑ, Σχολή Μηχανικών (ΣΜΗΧ), ΠΜΣ Μηχανικών Πληροφορικής, 2024-07-15) Pavlidou, Elisavet; Παυλίδου, Ελισάβετ; Tsiknakis, Emmanouil; Τσικνάκης, Εμμανουήλ
    Pain is a multifaceted and subjective experience that can affect individuals both physically and emotionally. According to the most widely accepted definition, it is characterized as the “unpleas ant sensory and emotional experience associated with, or resembling that associated with actual or potential tissue damage”. Visual analog scales, numeric rating scales, and various questionnaires, all relying on patient-reported outcome measurements, are prevalent methods within healthcare and research domains for assessing the presence, incidence, and severity of pain. Nevertheless, self-report evaluation tools require cognitive, linguistic, and social abilities, which may manifest variations in certain populations such as neonates, individuals with intellectual disabilities, and those affected by dementia. Pain can be classified as either acute or chronic. Acute pain is sudden, sharp, and usually, temporary. Conversely, persistent pain can result in prolonged suffering. Chronic pain is defined as enduring discomfort lasting beyond six months, often representing a disease state. Diagnosing and managing chronic pain can be challenging due to its sustained nature. In the case of individ uals hospitalized with chronic conditions, precise pain assessment is imperative for the effective application of pain management strategies. Currently, pain detection and accurate classification continue to pose a significant challenge in both the medical field and research endeavors. The objective of this master thesis is to automate pain recognition and assessment through physiological-data-driven predictive models. This study emphasizes the substantial importance of integrating multimodal physiological signals and explor ing their most efficient combination. Furthermore, we explored the influence of using multiple types of signal (multimodal approach) on any improvements in performance compared to using just one type of signal (unimodal approach). Additionally, we aimed to scrutinize the influence of demographic factors such as gender and age on pain perception and how they may affect the out comes of our experiments. We used the BioVid Part A dataset as input in our pipeline, incorporating the demographic data with age groups divided into: 20-35 years, 36-50 years, and 51-65 years. The Pan & Tompkins algorithm was applied for ECG feature extraction, while statistical analysis was employed for both GSR and EMG signals. Following data pre-processing, an early fusion approach was implemented in all unimodal and multimodal experiments encompassing binary (no pain vs. 1 pain) and multiclass tasks (all pain levels, no pain vs. pain levels). The Long Short-Term Memory (LSTM) approach demonstrated superior performance in pain intensity classification compared to Support Vector Machines (SVM) and Decision Trees, whether applied in unimodal or multimodal tasks. Galvanic skin response emerged as the most effective standalone modality, while the combination of ECG, EMG, and GSR proved to be the most ef fective pair of fused modalities. The combination of all physiological signals resulted in further improvement in our outcomes, achieving 81.75% accuracy in the binary classification task of no pain (BL) versus the highest level of pain (PA4) and 86.05% in the female population of the same classification task. It is noteworthy that in machine learning experiments, the polynomial kernel consistently yielded higher accuracy values in both binary and multiclass classification tasks. To the best of our knowledge, our study reported the highest accuracy (82.83%) in gender-age-based experiments using all physiological data from the BioVid dataset. This performance was observed in a binary task distinguishing between baseline and the highest intensity of pain within the female population aged 20-35.
  • Μικρογραφία εικόνας
    Τεκμήριο
    Parkinson’s disease prediction using artificial intelligence
    (ΕΛΜΕΠΑ, Σχολή Μηχανικών (ΣΜΗΧ), ΠΜΣ Μηχανικών Πληροφορικής, 2024-04-08) Panagiotakis, Georgios; Παναγιωτάκης, Γεώργιος; Tsiknakis, Emmanouil; Τσικνάκης, Εμμανουήλ
    Precise diagnosis of Parkinson’s disease (PD) is crucial for effective treatment and management of this progressive neurological condition. Existing diagnostic methods face obstacles due to overlapping symptoms with other neurological disorders and the lack of a conclusive diagnostic test. Sleep disorders are common among PD patients, and nocturnal sleep electroencephalography (EEG) data hold significant insights into the connection between PD and sleep disturbances, providing opportunities for early diagnosis and disease tracking. This research utilizes deep learning methodologies to examine nocturnal sleep EEG data for the differentiation of PD subjects and healthy individuals. An extensive review of current literature is performed to evaluate the state-of-the-art in PD, sleep disorders, EEG data analysis, and deep learning applications for neurological disorder classification. A dataset of nocturnal sleep EEG recordings from PD patients and healthy subjects is obtained, preprocessed, and divided into sleep stages, followed by the extraction of pertinent features. Several deep learning models, such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Gated Recurrent Units (GRU) networks, are explored for their appropriateness in the classification task. The chosen models are developed, executed, and optimized to differentiate PD subjects from healthy controls using nocturnal sleep EEG data. Model performance is assessed using relevant metrics (e.g., accuracy, precision, recall, F1-score) and compared with existing methods found in the literature.
  • Μικρογραφία εικόνας
    Τεκμήριο
    Το διαδίκτυο των πραγμάτων στο τομέα της υγείας.
    (ΕΛΜΕΠΑ, Σχολή Μηχανικών (ΣΜΗΧ), Τμήμα Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών, 2024-07-22) Στρατάκη, Δέσποινα; Strataki, Despoina; Τσικνάκης, Εμμανουήλ; Tsiknakis, Emmanouil
    Η παρούσα πτυχιακή εργασία διερευνά την ενσωμάτωση του Διαδικτύου Ιατρικών Πραγμάτων (ΔτΙΠ) με προηγμένες τεχνικές μηχανικής μάθησης για τη βελτίωση των αποτελεσμάτων της υγειονομικής περίθαλψης. Εμβαθύνει στην αξιοποίηση των δεδομένων βιοϊατρικών αισθητήρων, αναδεικνύοντας τον μετασχηματισμό της υγειονομικής περίθαλψης μέσω της συλλογής και ανάλυσης δεδομένων σε πραγματικό χρόνο. Σημαντική έμφαση δίνεται στο ρόλο της υπολογιστικής ομίχλης στη διαχείριση του τεράστιου όγκου δεδομένων από αισθητήρες υγειονομικής περίθαλψης, δίνοντας έμφαση στην αποτελεσματικότητά της στην προεπεξεργασία για εφαρμογές μηχανικής μάθησης. Η μελέτη εξετάζει διάφορους αλγορίθμους μηχανικής μάθησης, από τα βαθιά νευρωνικά δίκτυα έως την ομαδοποίηση, για την εξαγωγή σημαντικών μοτίβων σε βιοϊατρικά δεδομένα, βοηθώντας έτσι στην παρακολούθηση των ασθενών και την πρόβλεψη ασθενειών. Επιπλέον, η πτυχιακή ασχολείται με τις κρίσιμες προκλήσεις της ασφάλειας δεδομένων και της ιδιωτικότητας στο Διαδίκτυο των Ιατρικών Πραγμάτων προτείνοντας λύσεις όπως η ομοσπονδιακή μάθηση για τη διασφάλιση ευαίσθητων πληροφοριών υγείας. Εξετάζει επίσης τους ηθικούς προβληματισμούς στην αυτοματοποιημένη λήψη αποφάσεων στο πλαίσιο της υγειονομικής περίθαλψης. Επιπλέον, η έρευνα συζητά τη σημασία των εργαλείων οπτικοποίησης στο να γίνουν κατανοητά τα πολύπλοκα αποτελέσματα της Μηχανικής Μάθησης στους επαγγελματίες του τομέα της υγείας, ενισχύοντας τις διαδικασίες λήψης αποφάσεων. Η πτυχιακή επιχειρεί επίσης να εξετάσει τις αρχιτεκτονικές υλικού και βάσεων δεδομένων που απαιτούνται για την υποστήριξη πολυτροπικών δεδομένων αισθητήρων, τονίζοντας την ανάγκη για ισχυρά συστήματα που να χειρίζονται αποτελεσματικά διαφορετικούς τύπους δεδομένων υγειονομικής περίθαλψης. Τέλος, διερευνά τον προγραμματισμό με GPU και τις παράλληλες υπολογιστικές τεχνικές, απαραίτητες για την επεξεργασία βιοϊατρικών δεδομένων μεγάλης κλίμακας, και καταλήγει με την αξιολόγηση της παρακολούθησης της υγείας με βάση σήματα για κάθε ασθενή, υπογραμμίζοντας την πορεία προς την εξατομικευμένη υγειονομική περίθαλψη.