Scholarlink

Background <p>High prevalence and serious complications associated with obstructive sleep apnea (OSA) urge the need for methods that support non-invasive examinations, diagnostic accuracy and pathophysiological insights of OSA. This study aims to develop a Synthetic MRI-based radiomics model for diagnosing and stratifying OSA severity, and to explore its pathophysiological basis through integrated histopathological-transcriptomic analyses.</p> Methods <p>This cross-sectional study prospectively collected data at the tertiary hospital between July 2023 and October 2024. Individuals with and without OSA were recruited from the Otolaryngology Head and Neck Inpatient and Outpatient Department, all of whom underwent upper airway SyMRI. All participants were randomly divided into training and a validation set in an 8:2 ratio. The four-class (non-, mild, moderate and severe OSA) radiomics diagnostic model was developed to diagnose OSA and evaluate its severity across three algorithms (logistic regression, support vector machine, and random forest). Model performance was assessed using the receiver operating characteristic (ROC) curve. The underlying pathophysiological connections between radiomic features and OSA were explored by creating radiogenomic maps linking radiomic features with relevant biological pathways using the weighted correlation network analysis (WGCNA).</p> Results <p>230 participants were enrolled and divided into four groups by the apnea-hypopnea index (AHI): non-OSA (76 participants, AHI &lt; 5 events/h), mild OSA (28 participants, 5 ≤ AHI &lt; 15 events/h), moderate OSA (53 participants, 15 ≤ AHI &lt; 30 events/h), and severe OSA (73 participants, AHI ≥ 30 events/h). The four-class radiomics diagnostic model demonstrated excellent performance across the three algorithms. In the training set, the average area under curves (AUCs) of the four classes were 0.999, 0.997, 0.988, and 0.995, respectively. In the test set, the average AUCs were 0.968, 0.942, 0.893, and 0.932, respectively. WNGCA and histopathological staining analysis of the soft palates indicated that radiomic features were significantly associated with muscle function and the myocyte areas in soft palates.</p> Conclusions <p>SyMRI-based radiomic analysis provides an effective and supportive tool for diagnosing OSA, which can differentiate its severity from a pathophysiological perspective. The pathophysiological relevance of radiomics lies in its ability to capture subtle tissue alterations that are not readily apparent on conventional imaging.</p> Clinical trial number <p>Not applicable.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-regional radiomics of obstructive sleep apnea based on synthetic MRI

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月12日

查看全文
  • Jia Chen,
  • Shiyi Wen,
  • Ge Si,
  • Weixing Liu,
  • Zhiyuan Wang,
  • Weijin Li,
  • Jiaqi Chen,
  • Yaohong Deng,
  • Tufeng Chen,
  • Yanhua Zhu,
  • Mengyin Cai,
  • Wu Yuan,
  • Yanming Chen,
  • Yan Zou,
  • Guojun Shi,
  • Jin Ye

摘要

Background

High prevalence and serious complications associated with obstructive sleep apnea (OSA) urge the need for methods that support non-invasive examinations, diagnostic accuracy and pathophysiological insights of OSA. This study aims to develop a Synthetic MRI-based radiomics model for diagnosing and stratifying OSA severity, and to explore its pathophysiological basis through integrated histopathological-transcriptomic analyses.

Methods

This cross-sectional study prospectively collected data at the tertiary hospital between July 2023 and October 2024. Individuals with and without OSA were recruited from the Otolaryngology Head and Neck Inpatient and Outpatient Department, all of whom underwent upper airway SyMRI. All participants were randomly divided into training and a validation set in an 8:2 ratio. The four-class (non-, mild, moderate and severe OSA) radiomics diagnostic model was developed to diagnose OSA and evaluate its severity across three algorithms (logistic regression, support vector machine, and random forest). Model performance was assessed using the receiver operating characteristic (ROC) curve. The underlying pathophysiological connections between radiomic features and OSA were explored by creating radiogenomic maps linking radiomic features with relevant biological pathways using the weighted correlation network analysis (WGCNA).

Results

230 participants were enrolled and divided into four groups by the apnea-hypopnea index (AHI): non-OSA (76 participants, AHI < 5 events/h), mild OSA (28 participants, 5 ≤ AHI < 15 events/h), moderate OSA (53 participants, 15 ≤ AHI < 30 events/h), and severe OSA (73 participants, AHI ≥ 30 events/h). The four-class radiomics diagnostic model demonstrated excellent performance across the three algorithms. In the training set, the average area under curves (AUCs) of the four classes were 0.999, 0.997, 0.988, and 0.995, respectively. In the test set, the average AUCs were 0.968, 0.942, 0.893, and 0.932, respectively. WNGCA and histopathological staining analysis of the soft palates indicated that radiomic features were significantly associated with muscle function and the myocyte areas in soft palates.

Conclusions

SyMRI-based radiomic analysis provides an effective and supportive tool for diagnosing OSA, which can differentiate its severity from a pathophysiological perspective. The pathophysiological relevance of radiomics lies in its ability to capture subtle tissue alterations that are not readily apparent on conventional imaging.

Clinical trial number

Not applicable.