|
|
Graph Neural Networks in Epilepsy Surgery
Hrtonová, Valentina ; Filipenská, Marina ; Klimeš, Petr
Epilepsy surgery presents a viable treatment option for patients with drug-resistant epilepsy, necessitating precise localization of the epileptogenic zone (EZ) for optimal outcomes. As the limitations of currently used localization methods lead to a seizure-free postsurgical outcome only in about 60% of cases, this study introduces a novel approach to EZ localization by leveraging Graph Neural Networks (GNNs) for the analysis of interictal stereoelectroencephalography (SEEG) data. A GraphSAGE-based model for identifying resected seizure-onset zone (SOZ) electrode contacts was applied to a clinical dataset comprising 17 patients from two institutions. This study uniquely focuses on the use of interictal SEEG recordings, aiming to streamline the presurgical monitoring process and minimize risks and costs associated with prolonged SEEG monitoring. Through this innovative approach, the GNN model demonstrated promising results, achieving an Area Under the Receiver Operating Characteristic (AUROC) score of 0.830 and an Area Under the Precision-Recall Curve (AUPRC) of 0.432. These outcomes along with the potential of GNNs in leveraging the patient-specific electrode placement highlight their potential in enhancing the accuracy of EZ localization in drug-resistant epilepsy patients.
|
|
|
Atrial fibrillation localization for burden assessment
Martinásková, Klára ; Ředina, Richard (referee) ; Filipenská, Marina (advisor)
The diploma thesis deals with the problem of detection of atrial fibrillation from ECG recordings and localization of given fibrillation segments in signals with paroxysmal fibrillation. A research is done on atrial fibrillation, the origin of this pathology and methods of fibrillation detection from ECG recordings using deep learning. Subsequently, a convolutional neural network model with residual blocks is implemented in Python to classify short (3 s) segments of the ECG signal. Subsequently, the classification results are processed and the segments with paroxysmal fibrillation are localized in the signals with fibrillation. With the classification and localization, the burden assessment of fibrillation is further evaluated. The implemented classifier on the test set achieves an F1 score of 96,15 %. When the sections with fibrillation are localized by the algorithm, MAE of 0,95 s for detecting the beginnings and 1,29 s for detecting the ends with respect to the reference positions is achieved. The estimated patient's burden assessment is compared with the actual values and achieves MAE of 3 %
|
|
|
ECG arrhythmia detection
Pchálková, Aneta ; Filipenská, Marina (referee) ; Novotná, Petra (advisor)
This thesis describes the principles of ECG, the physiology of arrhythmias, their origin, and manifestations in the ECG, focusing on ventricular extrasystoles and bundle branch blocks. It examines contemporary methods for detecting these arrhythmias and acquiring the necessary features for their implementation. The work also covers data handling, including data preprocessing. Classification of ventricular extrasystoles and bundle branch blocks is implemented using k-nearest neighbors models.
|
|
|
Predicting the success of football players using machine learning methods
Janeček, Jan ; Filipenská, Marina (referee) ; Ředina, Richard (advisor)
This bachelor thesis focuses on the implementation of an artificial neural network in the Python programming language using the Keras library. The aim of the work is the numerical prediction of a football player’s match readiness on a scale from 0 to 1. The prediction is based on five physiological-kinematic data obtained from three training sessions preceding a given match. The reference data for training the artificial neural network includes technical data on the number of successful and total actions during the match. The data used in this work was collected from Sigma Olomouc U19 football club players using Polar Team Pro and Wyscout software. The lowest recorded model error, which was 0.1046, was achieved using a single hidden layer containing 15 perceptrons.
|
|
|
Graph Neural Networks in Epilepsy Surgery
Hrtoňová, Valentina ; MSc, Daniel Uher, (referee) ; Filipenská, Marina (advisor)
Úspěch epileptochirurgického zákroku závisí na přesné lokalizaci epileptogenní zóny (EZ), avšak pouze 60% pacientů je po operaci bez záchvatů, což je často způsobeno nepřesnou identifikací EZ. Tato práce představuje novou metodu lokalizace EZ využívající grafové neuronové sítě (GNN) k analýze interiktálních biomarkerů - konkrétně interiktálních spiků a relativní entropie. Modely GNN byly využity pro lokalizaci kontaktů elektrod v resekované zóně vzniku záchvatu na základě dat z interiktální stereoelektroencefalografie a validovány na souboru klinických dat 37 pacientů ze dvou institucí. Nejlépe hodnocený model GNN - Graph Attention Network - dosáhl mediánu Area Under the Receiver Operating Characteristic (AUROC) 0,971 a mediánu Area Under the Precision-Recall Curve (AUPRC) 0,525 v souboru 19 pacientů s dobrým pooperačním výsledkem, přičemž v obou metrikách statisticky významně překonal referenční model založený na četnosti spiků (Wilcoxon Signed Rank test, p
|
|
|
Health assessment using smart devices
Vargová, Enikö ; Filipenská, Marina (referee) ; Němcová, Andrea (advisor)
This thesis deals with the possibilities of non-invasive determination of blood glucose from photoplethysmographic signals. Elevated blood sugar is often associated with disease called diabetes mellitus. Diabetes is one of the world’s major chronic diseases. Untreated diabetes is often a cause of death. The aim of the work is to propose methods for glycemic classification and prediction. Two datasets have been created by recording the PPG signals using two smart devices (a smart wristband and a smartphone), along with their blood glucose levels measured in an invasive way. The PPG signals were preprocessed, and suitable features were extracted from them. Various machine-learning models for glycemic classification and prediction were created.
|
|
|
Advanced sleep quality estimation
Benáček, Petr ; Ředina, Richard (referee) ; Filipenská, Marina (advisor)
This thesis deals with the assessment of sleep quality using modern deep learning methods. The thesis describes metrics for automatic classification of sleep stages. A selected database of sleep data is discussed. Due to the low number of data in the wakefulness phase, different methods of data augmentation are described and implemented. Models based on 1D convolutional networks are the basis for the classification. As a result, models for binary classification and classification of 3 and 4 sleep phases are prepared. Finally, sleep quality metrics are calculated using these models and the results are compared with the literature.
|
|
|
Automatické měření efektivní refrakterní periody srdeční tkáně
Ředina, Richard ; Filipenská, Marina
Elektrofyziologické vyšetření jako jedna z možností léčby arytmií je i v dnešní době stále časově náročný výkon. S cílem snížení této časové zátěže prezentujeme vyvinutý algoritmus, který z měřeného EKG automaticky určuje efektivní refrakterní periodu (ERP) tkáně. Algoritmus sestává z především filtrace a detekce lokálních extrémů v signálech. Algoritmus byl testován na interní databázi signálů získaných od deseti pacientů, kteří podstoupili elektrofyziologické vyšetření. Výstup algoritmu se shodoval s elektrofyziologem stanovenou ERP v devíti z deseti případů (σ = 6 ms). Pro svou relativní úspěšnost a nenáročnou implementaci slibuje možné využití v real-time aplikaci během vyšetření, při kterém by mohl plně automatizovat a tím i urychlit stimulační protokoly.
|
|
|
ECG arrhythmia detection
Šoltés, Tomáš ; Filipenská, Marina (referee) ; Novotná, Petra (advisor)
This bachelor thesis describes commonly present arrhytmias such as premature ventricular complex, bundle branch blocks and their detection using conventional methods and modern methods, utilising neural networks. Practical part includes: Detection of premature ventricular complexes and detection of bundle branch blocks using statistical analysis of QRS complex and their classification with K-nearest neighbors
|
|
|
ECG arrhythmia detection
Šoltés, Tomáš ; Filipenská, Marina (referee) ; Novotná, Petra (advisor)
This bachelor thesis describes commonly present arrhytmias such as premature ventricular complex, bundle branch blocks and their detection using conventional methods and modern methods, utilising neural networks. Practical part includes: Detection of premature ventricular complexes by their RR intervals and detection of bundle branch blocks using statistical analysis of QRS complex and their clustering with K-nearest neighbors
|
guest ::
