PhD Scientific Days 2025
Budapest, 7-9 July 2025
Surgical Medicine
The Role of Artificial Intelligence in the Diagnosis of Obstructive Sleep Apnea
Előadó neve
Török Eszter
Neptun code
I7XFCN
Előadó munkahelye
Semmelweis Egyetem Fül-Orr-Gégészeti és Fej-Nyaksebészeti Klinika
Előadó telefonszáma
+36307126616
Előadó e-mail címe
torok.eszter51@gmail.com
Az előadás címe
The Role of Artificial Intelligence in the Diagnosis of Obstructive Sleep Apnea
Szerző(k) neve és munkahelye
Eszter Török1
1: Semmelweis Egyetem Fül-Orr-Gégészeti és Fej-Nyaksebészeti Klinika
Bemutatás módja
Szóbeli
Szekció
Surgical Medicine
Language of the presentation
Hungarian
Preferred session
Surgical Medicine
Összefoglaló szövege
Introduction: Obstructive sleep apnea (OSA) is characterized by partial or complete upper airway obstruction during sleep, leading to hypoxia, hypercapnia, and sleep fragmentation. Untreated OSA is associated with cardiovascular disease, stroke, and metabolic disorders, highlighting the importance of early diagnosis. Although nearly one billion people worldwide are affected, the rate of undiagnosed cases remains high due to limitations in current diagnostic methods.
Aims: Our research aimed to improve the diagnosis of OSA using soft tissue MRI of the neck, supported by artificial intelligence (AI).
Method: In this prospective study, 92 patients (68 men, 24 women; mean age ± SD: 40.63 ± 12.7 years) with suspected sleep apnea or snoring were examined at the Department of Otorhinolaryngology and Head and Neck Surgery, Semmelweis University. All participants underwent polysomnography, soft tissue neck MRI, and anthropometric measurements. Based on the sleep study results, patients were divided into control (n=30) and OSA (n=62) groups. On the MRI scans, eight predefined parameters were measured, including upper airway length, upper airway volume, the antero-posterior and latero-lateral diameters of both the retropalatal and retroglossal regions, and the length of the retropalatal and retroglossal regions.
Results: Using conventional statistical analysis, significant differences were found between the OSA and control groups in upper airway length (p<0.05), retropalatal length (p<0.05), and the latero-lateral diameters of the retropalatal and retroglossal regions (p<0.05). Using machine learning AI and the eight previously defined parameters, we were able to identify the presence of OSA with 90% accuracy, achieving a specificity of 92% and a sensitivity of 87%
Conclusion: Measurements of upper airway structures on soft tissue neck MRI can be used to reliably predict the presence of OSA with the help of AI. This method could represent a significant step forward in OSA screening, as it offers high predictive value for detecting the condition. It may allow OSA to be screened by reusing existing neck MRI scans, without placing further burden on the healthcare system. This approach offers a cost-effective and rapid screening alternative in primary care, with significantly greater accuracy than currently used questionnaires.
University
Semmelweis University
Supervisor
Dr. Molnár Viktória PhD
Publication of my abstract
I do not give consent to the publication of my abstract on the website of the congress.
phd.section.field
before finishing undergraduate studies (TDK, MD-PhD)
Kind
Szabad
Status
elfogadva
Accepted presentation method
szóbeli
Előadás fájl jóváhagyás
nem hagyta jóvá
Előadó
9086
Start
10:15
End
10:30