A multi-omics method for breast cancer diagnosis based on metabolites in exhaled breath, ultrasound imaging, and basic clinical information

Yuan Yang,Huiling Long,Yong Feng,Shuangming Tian,Haibin Chen,Ping Zhou

DOI: https://doi.org/10.1016/j.heliyon.2024.e32115

IF: 3.776

2024-05-28

Heliyon

Abstract:Background and aims: Through a nested cohort study, we evaluated the diagnostic performance of breath-omics in differentiating between benign and malignant breast lesions, and assessed the diagnostic performance of a multi-omics approach that combines breath-omics, ultrasound radiomics, and clinic-omics in distinguishing between benign and malignant breast lesions. Materials and methods: We recruited 1,723 consecutive patients who underwent an automated breast volume scanner (ABVS) examination. Breath samples were collected and analyzed by high-pressure photon ionization time-of-flight mass spectrometry (HPPI-TOF-MS) to obtain breath-omics features. 238 of 1,723 enrolled participants have received pathological confirmation of breast nodules finally. The breast lesions of the 238 participants were contoured manually based on ABVS images for ultrasound radiomics feature calculation. Then, single- and multi-omics models were constructed and evaluated for breast nodules diagnosis via five-fold cross-validation. Results: The area under the curve (AUC) of the breath-omics model was 0.855. In comparison, the multi-omics model demonstrated superior diagnostic performance for breast cancer, with sensitivity, specificity, and AUC of 84.1 %, 89.9 %, and 0.946, respectively. The multi-omics performance was comparable to that of the Breast Imaging Reporting and Data System (BI-RADS) classification via senior ultrasound physician evaluation. Conclusion: The multi-omics approach combining metabolites in exhaled breath, ultrasound imaging, and basic clinical information exhibits superior diagnostic performance and promises to be a non-invasive and reliable tool for breast cancer diagnosis.