@article{LIU2026, 
author = {Mei LIU and Xin CHEN and Chao LIU and Ruiwei WANG and Yanzhou WANG},
title = {Diagnostic efficacy of ultrasound combined with exhaled volatile organic compounds for epithelial ovarian carcinoma},
year = {2026},
journal = {Journal of Army Medical University},
volume = {48},
number = {11},
pages = {1569-1576},
keywords = {ovarian neoplasms, ultrasonography, volatile organic compounds, cross-sectional study, random forest},
url = {https://www.sciopen.com/article/10.16016/j.2097-0927.202604055},
doi = {10.16016/j.2097-0927.202604055},
abstract = {ObjectiveOvarian cancer (OC) has the highest case fatality rate among gynecological malignancies, among which epithelial ovarian carcinoma (EOC) represents the most common pathological subtype. Due to its atypical early symptoms, approximately 70% of patients are diagnosed at advanced stages (Stage Ⅲ or Ⅳ), resulting in an extremely poor prognosis. Conversely, the 5-year survival rate for early-stage patients may exceed 90%. Therefore, efficient early diagnostic methods are of critical clinical significance for improving patient outcomes. Ultrasound serves as the preferred imaging modality for ovarian tumors, and its non-invasiveness, favorable reproducibility and cost-effectiveness have rendered it widely applied in clinical practice. However, relying solely on ultrasound for diagnosing OC entails certain limitations in sensitivity and specificity. Exhaled volatile organic compounds (VOC) analysis, as a non-invasive metabolomic technique, has demonstrated potential application value in tumor screening and diagnosis. This study aims to investigate whether the integration of exhaled VOC characteristics with ultrasound could enhance the diagnostic efficacy for EOC, thereby exploring a more precise non-invasive diagnostic paradigm.MethodsA single-center, cross-sectional study was carried out from July 30, 2024 to April 30, 2025, employing thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS) to analyze exhaled breath samples obtained from 236 patients with ovarian tumors (162 benign ovarian tumors and 74 malignant ovarian tumors) admitted to our department between July 30, 2024 and April 30, 2025. Differential VOCs were identified and selected for model development. Two prediction models were constructed using random forests: an ultrasound-based model (incorporating tumor long-axis diameter and cystic-solid composition) and a combined ultrasound-VOC model. Subgroup analyses based on pathology and stage were performed. Diagnostic performance of both models for early-stage EOC was evaluated through sensitivity, specificity and area under the receiver operating characteristic curve (AUC).ResultsThe ultrasound model achieved an AUC value of 0.865 (95%CI: 0.812 to 0.918), a sensitivity of 0.92 and a specificity of 0.82. The combined ultrasound-VOC model demonstrated superior performance with an AUC value of 0.896 (95%CI: 0.845 to 0.947), a sensitivity of 0.92, and a specificity of 0.84. Comparative analysis indicated that no change in sensitivity between the 2 models, and neither the AUC increase of 0.031 (95%CI: -0.008 to 0.070) nor the specificity improvement of 0.02 in the combined model reached statistical significance. Subgroup analysis revealed that the combination model improved the sensitivity of early-stage (Stage Ⅰ) ovarian cancer from 89% to 94%, yet the overall diagnostic efficacy gain remained modest across various pathological subtypes and stages.ConclusionThe ultrasound-based model demonstrates favorable diagnostic efficacy for EOC. The ultrasound-VOC model exhibits some improvement in diagnostic efficacy, but the gain is modest and lacks statistical significance, which is insufficient to challenge the position of ultrasound as a first-line diagnostic tool. Notably, the combined model demonstrates high sensitivity for early-stage EOC, suggesting that it may be more suitable as a screening tool for high-risk populations rather than replacing ultrasound as a first-line diagnostic modality.}
}