Machine learning-powered audio-omics processing method as an auxiliary diagnostic approach for advanced
Ting You1, Fangna Huan2, Zhenzhen Luo3
1The First Affiliated Hospital, Department of Emergency, Hengyang Medical School, University of South China, Hengyang, 421001, China.
BMC Cancer
|June 25, 2026
Summary
A new machine learning model analyzes acoustic features in speech to help detect advanced nasopharyngeal carcinoma (NPC). This non-invasive approach shows promise for screening, complementing existing methods for this head and neck cancer.
Area of Science:
- Oncology
- Artificial Intelligence
- Bioacoustics
Background:
- Nasopharyngeal carcinoma (NPC) is a malignant head and neck tumor, often diagnosed at advanced stages.
- Epstein-Barr virus (EBV) is linked to NPC, but serological detection has limitations.
- Current screening methods for NPC have limitations, necessitating novel diagnostic approaches.
Purpose of the Study:
- To develop and evaluate a machine learning-based acoustic signal processing model for auxiliary diagnosis of advanced NPC.
- To explore the potential of non-invasive acoustic analysis for NPC screening.
Main Methods:
- Collected audio data from 359 advanced NPC patients and 304 healthy controls.
- Developed a machine learning-powered Nasopharyngeal Carcinoma Screening (ML-NPCS) system.
- The ML-NPCS system involves speech acquisition, acoustic feature extraction, and classification.
Main Results:
- The ML-NPCS system achieved 84.2% accuracy in an independent test set.
- Sensitivity was 88.9% and specificity was 78.7% for distinguishing advanced NPC patients.
- The model demonstrated effectiveness using voice-derived acoustic features.
Conclusions:
- The ML-NPCS model shows preliminary potential for identifying advanced NPC patients.
- Voice analysis offers a non-invasive auxiliary diagnostic tool for NPC.
- Further validation in diverse cohorts and early-stage disease is recommended.
