Work

Wavelet-Based 1D-to-2D Encoding of SESI-MS Breathomics Profiles for Disease-Group Classifi- cation

Year: 2026
Type: article
Abstract: Complete analysis code and experiment outputs supporting the manuscript “Wavelet-Based 1D-to-2D Encoding of SESI-MS Breathomics Profiles for Disease-Group Classification” (Kavouras et al.). The repository provides a permanent, citable snapshot of the code published as a Kaggle notebook (version 17), together with its complete execution outputs: cross-validation protocol and fold assignments, hyperparameter search spaces and selected configurations, classification results and pooled out-of-fold metrics with bootstrap confidence intervals, statistical comparisons (Nadeau–Bengio corrected test, paired and label-permutation tests), explainability (occlusion-sensitivity) outputs, and all tables and figures reproduced in the manuscript. A synthetic SESI-MS dataset, matching the structure of the original measurements, is included so that the full pipeline can be executed end-to-end by anyone. The original raw SESI-MS measurements were collected within the VOCORDER clinical trial (ClinicalTrials.gov NCT06711939) and constitute sensitive health data under the GDPR; they are not included here and remain available on request from the corresponding authors, subject to applicable ethical, institutional, legal, and project requirements. Identifiers appearing in the included outputs are internal, re-coded tracking codes unrelated to hospital patient records. Keywords: breathomics, SESI-MS, wavelet transform, mass spectrometry, machine learning, disease classification, explainable AI, nested cross-validation
Language: English
Open Access status: green
OpenAlex ID: W7234420893
FWCI:
Cites:
Cited by:
Related to: 10