A Method for Preprocessing and Batch Correction of Oral Exhalation Data Based on PTR-TOF-MS

By combining deep reinforcement learning and Bayesian inference, the signal complexity and batch effect problems of PTR-TOF-MS in processing oral exhalation data are solved, achieving high-precision peak extraction and batch correction, and improving the consistency and accuracy of data analysis.

CN121834140BActive Publication Date: 2026-06-30SHANGHAI NINTH PEOPLES HOSPITAL SHANGHAI JIAO TONG UNIV SCHOOL OF MEDICINE
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SHANGHAI NINTH PEOPLES HOSPITAL SHANGHAI JIAO TONG UNIV SCHOOL OF MEDICINE
Filing Date
2026-03-11
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing PTR-TOF-MS technology faces challenges in processing oral exhalation data, including signal complexity and instability, baseline drift and asymmetric tailing, as well as severe batch effects. This makes it difficult to accurately identify low abundance peaks and overlapping peaks. Furthermore, the existing processing procedures lack end-to-end standardization and cannot effectively eliminate systematic biases caused by changes in instrument status.

Method used

A deep reinforcement learning-based agent is constructed to walk on the mass spectrum to adaptively identify spectral peaks. Batch effect correction is performed by combining Bayesian inference. Batch effects are eliminated through peak extraction, integration, monotonic transformation, and the ComBat algorithm. An automated data preprocessing pipeline is constructed to achieve adaptive identification of complex spectral peaks and effective removal of batch effects.

Benefits of technology

It achieves high-precision adaptive extraction of PTR-TOF-MS data, preserves trace biomarker characteristics, improves the consistency of multi-center large-scale clinical data analysis, and eliminates systematic bias caused by changes in instrument status.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure SMS_23
    Figure SMS_23
  • Figure SMS_25
    Figure SMS_25
  • Figure SMS_9
    Figure SMS_9
Patent Text Reader

Abstract

This invention discloses a preprocessing and batch correction method for oral exhalation data based on PTR-TOF-MS. The method selects spectra with peak height and stable waveforms in the middle segment, and removes abnormal instrument segments with no peak height and / or below the sample inspection threshold. It calculates the intersection-union ratio (IUR) of the intervals marked by the intelligent agent with those marked manually or generated with high confidence. If the IUR exceeds a preset threshold, a positive reward is given; if the IUR does not exceed the preset threshold, it continues to determine whether there are missed or duplicate markings of the peak start, apex, and end points. If so, a penalty is given; otherwise, it continues to determine whether noise intervals are marked as peak start, apex, and end points. If so, an additional penalty is given. The method trains a policy network to form a structured peak list for each sample's output peak start, apex, and end point m / z intervals. It then performs integral calculations, monotonic transformations, and ComBat correction to obtain the final data matrix, achieving adaptive identification of complex peaks and effectively eliminating batch effects.
Need to check novelty before this filing date? Find Prior Art