izpis_h1_title_alt

Analiza biomehanskih značilnosti ortopedskih pacientov za napovedovanje degenerativnih obolenj
ID Ravnik, Šimen (Author), ID Šubelj, Lovro (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (23,26 MB)
MD5: CFE719796C5AFE00E087D8DB4780754A

Abstract
Analiza biomehanskih značilnosti ortopedskih pacientov za napovedovanje degenerativnih obolenj je raziskovalno delo na področju medicine oziroma bolj natančno ortopedije. Sagitalno ravnovesje pomeni harmonično obliko hrbtenice in pravilna sagitalna usmeritev medenice je pri zagotavljanju le-tega ključnega pomena. Sagitalno usmeritev medenice določimo na podlagi merjenja biomehanskih oziroma geometrijskih parametrov, s katerimi se nato ugotavlja nepravilnosti, ki lahko povzročijo degenerativna obolenja, kot so na primer hernia diska in spondilolisteza. Ugotavljanje nepravilnosti usmeritve medenice je običajno ročno opravilo in je posledično rezultat relativno subjektivna odločitev. S sodobnimi metodami strojnega učenja je mogoče paciente klasificirati in s tem omogočiti lažjo in bolj natančno odločitev. V delu je najprej predstavljena analiza geometrijskih parametrov, s katerimi se diagnosticira paciente z degenerativnimi obolenji ledvene hrbtenice, nato pa so parametri uporabljeni tudi v različnih napovednih modelih strojnega učenja, s katerimi izvedemo klasifikacijo tovrstnih pacientov. Ugotovili smo, da je najuspešnejši napovedni model strojnega učenja za naše podatke logistična regresija. Z njo dosežemo uspešnost diagnosticiranja hernijskih pacientov – občutljivost: 83.3% (95%CI 72.1 - 96.8%) in specifičnost: 75.0% (95%CI 64.7 - 85.2%), ki je primerljiva uspešnosti kliničnega diagnosticiranja. Uspešnost diagnosticiranja pacientov s spondilolistezo z našim napovednim modelom pa znaša – občutljivost: 96,4% (95%CI 93,6 - 99,1%) in specifičnost: 97.5% (95%CI 94,3 - 1.0%), kar izboljša občutljivost kliničnega diagnosticiranja pacientov s spondilolistezo od 8% do 36%, specifičnost pa do približno 10%.

Language:Slovenian
Keywords:strojno učenje, podatkovno rudarjenje, računalništvo v medicini, sagitalna usmeritev medenice
Work type:Bachelor thesis/paper
Typology:2.11 - Undergraduate Thesis
Organization:FRI - Faculty of Computer and Information Science
Year:2021
PID:20.500.12556/RUL-124847 This link opens in a new window
COBISS.SI-ID:53028611 This link opens in a new window
Publication date in RUL:23.02.2021
Views:1572
Downloads:223
Metadata:XML DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:English
Title:Sagittal pelvic orientation feature analysis for predicting degenerative diseases
Abstract:
Sagittal pelvic orientation feature analysis for predicting degenerative diseases is a research in the field of medicine or more specifically orthopedics. Sagittal balance means a harmonious shape of the spine and proper sagittal orientation of the pelvis is crucial in ensuring this. Sagittal orientation of the pelvis is determined based on biomechanical or geometric parameters, which are then used to identify abnormalities that can cause degenerative diseases, such as disc herniation and spondylolisthesis. Determining pelvic orientation irregularities is usually a manual task and the result is a relatively subjective decision. With modern machine learning techniques, patients can be classified and thus enable an easier and more accurate decision. The paper first presents an analysis of geometric parameters used to diagnose patients with degenerative diseases of the lumbar spine, and then the parameters are also used in various predictive models, which are used to classify such patients. We found that the most successful predictive machine learning model for our data is logistic regression. It achieves the accurateness of diagnosing hernia – sensitivity: 83.3% (95%CI 72.1 - 96.8%) and specificity: 75.0% (95%CI 64.7 - 85.2%), which is comparable to the accuracy of clinical diagnosis. The accurateness of diagnosing spondylolisthesis with our model is – sensitivity: 96,4% (95% CI 93,6 - 99,1%) and specificity: 97.5% (95%CI 94.3 - 1.0%), which improves the sensitivity of the clinical diagnosis of spondylolisthesis from 8% to 36%, and the specificity to about 10%.

Keywords:machine learning, data mining, computer science in medicine, sagittal pelvic orientation

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back