| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Show document Help

Title:Detekcija zamrznjenega koraka in stimulacija v realnem času s personaliziranim nosljivim sistemom za bolnike s Parkinsonovo boleznijo : doktorska disertacija
Authors:ID Slemenšek, Jan (Author)
ID Šafarič, Riko (Mentor) More about this mentor... New window
ID Pirtošek, Zvezdan (Comentor)
ID Geršak, Jelka (Comentor)
Files:.pdf Doktorska_Disertacija_Slemensek_Jan.pdf (6,57 MB)
MD5: 71BEB687444043D6808EC5EA1E1170EE
 
Language:Slovenian
Work type:Dissertation
Typology:2.08 - Doctoral Dissertation
Organization:FERI - Faculty of Electrical Engineering and Computer Science
Abstract:Analiza in razumevanje človeškega gibanja odpirata nova vrata na različnih področjih, kot so šport, robotika, virtualna resničnost, medicina in rehabilitacija. Detekcija specifičnih aktivnosti človeškega gibanja omogoča razvoj naprednih, personaliziranih naprav, orodij in pripomočkov za medicinske namene. Zbiranje in analiza gibalnih podatkov omogočata ustvarjanje objektivnejše ocene o dejanskem motoričnem stanju posameznika ali bolnika, kar lahko poveča učinkovitost treninga, okrevanja in rehabilitacije. Disertacija predlaga robusten, nosljiv merilni sistem za zajemanje in analizo človeškega gibanja v realnem času, namenjen bolnikom s Parkinsonovo boleznijo, ki doživljajo epizode zamrznitve koraka. Gibalni podatki so pridobljeni s pomočjo pospeškometrov, giroskopov in merilnikov mišične aktivnosti, vgrajenih v elastičen pas, nameščen pod kolenom na obeh nogah. Gibalni podatki so uporabljeni za učenje in testiranje algoritmov strojnega učenja. Z obširno primerjalno analizo desetih uveljavljenih klasifikacijskih algoritmov strojnega učenja za namene detekcije petih aktivnosti smo identificirali kombinacijo konvolucijskih in rekurentnih nevronskih mrež z dodanim mehanizmom pozornosti kot najbolj učinkovit klasifikacijski model, ki je nove instance klasificiral s točnostjo 98.9 %, natančnostjo 96.8 %, senzitivnostjo 97.8 %, specifičnostjo 99.1 % ter F1 oceno 97.3 %. Enostavnejša, čeprav zanemarljivo manj učinkovita kombinacija konvolucijskih in rekurentnih nevronskih mrež z dodatkom preteklih podatkovnih instanc, je implementirana na mikrokrmilniku in uporabljena za klasifikacijo novih instanc s frekvenco 40 Hz. Sistem je v realnem času detektiral zamrznjen korak pri bolnikih s Parkinsonovo boleznijo s točnostjo 95.1 % in povprečno zakasnitvijo 261 ms, pri čemer so bolniki prejemali stimulacijo po potrebi. Ritmični vibracijski stimulatorji so uspešno zmanjšali povprečno trajanje zamrznjenega koraka za 38 %.
Keywords:Analiza gibanja, Parkinsonova bolezen, zamrznitev koraka, strojno učenje, aktivna stimulacija.
Place of publishing:Maribor
Place of performance:Maribor
Publisher:[J. Slemenšek]
Year of publishing:2024
Number of pages:XI, 67 f.
PID:20.500.12556/DKUM-88690 New window
UDC:004.85.021:616.858-009.1(043.3)
COBISS.SI-ID:212623875 New window
Publication date in DKUM:22.10.2024
Views:190
Downloads:78
Metadata:XML DC-XML DC-RDF
Categories:KTFMB - FERI
Misc.
:
Copy citation
  
Average score:(0 votes)
Your score:Voting is allowed only for logged in users.
Share:Bookmark and Share



Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Secondary language

Language:English
Title:Online freezing of gait detection and cueing with personalizable wearable system for Parkinson's disease patients
Abstract:The analysis and understanding of human movement open new doors in various fields such as sports, robotics, virtual reality, medicine, and rehabilitation. The detection of specific human movement activities enables the development of advanced, personalized devices, tools, and aids for medical purposes. The collection and analysis of movement data allow for the creation of a more objective assessment of an individual's or patient's actual motoric condition, which can enhance the effectiveness of training, recovery and rehabilitation. This dissertation proposes a robust, wearable measurement system for capturing and analyzing human movement in real time, designed for patients with Parkinson's disease experiencing freezing of gait episodes. Movement data were obtained using accelerometers, gyroscopes, and muscle activity sensors integrated into an elastic belt worn below the knee on both legs. Gait data was used for training and testing machine learning algorithms. Through an extensive comparative analysis of ten established machine learning classification models for detecting five gait activities, we identified a combination of convolutional and recurrent neural networks with an added attention mechanism as the most effective classification model, which classified new instances with 98.9 % accuracy, 96.8 % precision, 97.8 % sensitivity, 99.1 % specificity and 97.3 % F1 score. A simpler, albeit neglegable less effective, combination of convolutional and recurrent neural networks with the addition of past data instances was implemented on a microcontroller and used for classifying new instances at a rate of 40 Hz. In real time, the system detected freezing of gait episodes in patients with Parkinson's disease with an accuracy of 95.1 % and an average detection delay of only 261 ms, where the patientsreceived 'on demand'stimulation. Using rhythmic vibratory stimulators, the system successfully reduced the duration of freezing episodes by an average of 38 %.
Keywords:Movement analysis, Parkinson's disease, freezing of gait, machine learning, active stimualtion.


Comments

Leave comment

You must log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica