Comparison of digital filtering methods for diffuse reflectance spectra of the oral mucosa

Abstract

In 2023, 11 625 new cases of malignant neoplasm (MN) of the oral cavity and 9 055 deaths from MN of the oral cavity and pharynx were recorded in Russia [1]. Early diagnosis of oral mucosa MN increases survival rates by 50–60% [2], which underscores the urgency of developing methods for their early detection. Diffuse reflectance spectroscopy, being a non-invasive optical technique, demonstrates high diagnostic efficacy when combined with machine learning algorithms [3–5]. Raw diffuse reflectance spectra contain instrumental noise and operator-dependent variability associated with changes in probe positioning, angle, and applied pressure. Filtering, as a part of spectral data preprocessing, is aimed at suppressing random fluctuations while preserving the informative signal component. However, excessive smoothing can alter local spectral features. Therefore, the appropriate choice of filtering algorithm affects the performance of machine learning models and the diagnostic efficacy of the method. [6]. Aim of the study is to conduct a comparative analysis and select a digital filtering algorithm for diffuse reflectance spectra of the oral mucosa. Diffuse reflectance spectra of the oral mucosa in the 500–900 nm range, recorded from areas of MN development, filtered using six methods and evaluated according to visual and quantitative criteria, including the degree of original signal modification, spectral angle, smoothness metrics, residual characteristics, and the extent of preservation of local features. The proposed criteria are systematized into a hierarchy based on the results of evaluating the quality of automatic spectral classification before and after filtering.
References:
1. Kaprin A.D., Starinskiy V.V., Shakhzadova A.O., editors. Zlokachestvennye novoobrazovaniya v Rossii v 2023 godu (zabolevaemost i smertnost) [Malignant neoplasms in Russia in 2023 (morbidity and mortality)]. Moscow: MNIOI im. P.A. Gertsena – filial FGBU «NMITs radiologii» Minzdrava Rossii/ – 2024. – 276 p. (In Russ.).
2. Tiziani S., Lopes V., Günther U.L. Early-stage diagnosis of oral cancer using 1H NMR-based metabolomics // Neoplasia. – 2009. – Vol. 11 – No 3. – P. 269. DOI: 10.1593/neo.81396
3. Kolpakov A.V. Moshkova A.A., Melikhova E.V. Diffuse Reflectance Spectroscopy of the Oral Mucosa: In Vivo Experimental Validation of the Precancerous Lesions Early Detection Possibility // Diagnostics. – 2023. – Vol. 13. – No. 9. – Art. 1633. – DOI: 10.3390/diagnostics13091633.
4. Jayanthi J.L., Nisha G.U., Manju S. Diffuse Reflectance Spectroscopy: Diagnostic Accuracy of a Non-Invasive Screening Technique for Early Detection of Malignant Changes in the Oral Cavity // BMJ Open. – 2011. – Vol. 1. – Art. e000071. – DOI: 10.1136/bmjopen-2011-000071.
5. Rossberg N., Li C.L., Innocente S. Machine Learning Applications to Diffuse Reflectance Spectroscopy in Optical Diagnosis: A Systematic Review // Applied Spectroscopy Reviews. – 2025. – DOI: 10.1080/05704928.2025.2525817.
6. Sánchez-Maroño N., Alonso-Betanzos A., Tombilla-Sanromán M. Filter Methods for Feature Selection – A Comparative Study. In: Yin, H., Tino, P., Corchado, E., Byrne, W., Yao, X. (eds) Intelligent Data Engineering and Automated Learning - IDEAL 2007. IDEAL 2007. Lecture Notes in Computer Science, vol 4881. Springer, Berlin, Heidelberg. DOI: 10.1007/978-3-540-77226-2_19

Speaker

Maxim Mamaev
Bauman Moscow State Technical University
Russia

Discussion

Ask question