Machine learning–based classification of spatially resolved diffuse reflectance and autofluorescence spectra for human skin cancer diagnostics aid - Research paper

Q: What accuracy ranges were reported for different diagnostic groupings?

Binary tests discriminating BCC or SCC versus healthy skin exceeded 80%, while actinic keratosis versus other classes achieved about 65%–75% accuracy, and multiclass accuracy reached about 57%.

Sep 23, 2026

Uploaded by Oliver

Incidence of keratinocyte carcinomas continues to rise, creating urgent need for reliable early diagnostic support. The study evaluates an optical spectroscopy device combined with machine-learning classification to aid detection. In vivo spectra from 131 patients were classified into basal cell carcinoma, squamous cell carcinoma, actinic keratosis, and healthy skin using support vector machine, discriminant analysis, and multilayer perceptron, including binary and multiclass comparisons. Fusion across modalities improved accuracy up to 10 percentage points and showed promising performance comparable to clinicians.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages25

Reading Time1 h 3 min

AccessFree to download

Abstract

Introduction

Clinical need for improved KC diagnosis

Optical biopsy and spectroscopic approaches

RESEARCH PAPER Machine learning–based classification of spatially resolved diffuse reflectance and autofluorescence spectra acquired on human skin for actinic keratoses and skin carcinoma diagnostics aid Valentin Kupriyanov ,a,b Walter Blondel ,a Christian Daul ,a Martin Hohmann ,c Grégoire Khairallah,d Yury Kistenev ,b and Marine Amourouxa,* aUniversité de Lorraine, CNRS, CRAN UMR, Vandoeuvre-Lès-Nancy, France bTomsk State University, Laboratory of Laser Molecular Imaging and Machine Learning, Tomsk, Russia cFriedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Institute of Photonic Technologies (LPT), Erlangen, Germany dMetz-Thionville Regional Hospital, Department of Plastic, Aesthetic and Reconstructive Surgery, Ars-Laquenexy, France ABSTRACT. Significance: The incidence of keratinocyte carcinomas (KCs) is increasing every year, making the task of developing new methods for KC early diagnosis of utmost medical and economical importance. Aim: We aim to evaluate the KC diagnostic aid performance of an optical spectroscopy device associated with a machine-learning classification method. Approach: We present the classification performance of autofluorescence and diffuse reflectance optical spectra obtained in vivo from 131 patients on four histological classes: basal cell carcinoma (BCC), squamous cell carcinoma (SCC), actinic keratosis (AK), and healthy (H) skin. Classification accuracies obtained by support vector machine, discriminant analysis, and multilayer perceptron in binaryand multi-class modes were compared to define the best classification pipeline. Results: The accuracy of binary classification tests was >80% to discriminate BCCor SCC from H. For AK versus other classes, the classification achieved a 65% to 75% accuracy. In multiclass (three or four classes) classification modes, accuracy reached 57% . Fusion of decisions increased classification accuracies (up to 10 percentage point-increase), proving the interest of multimodal spectroscopy compared with a single modality. Conclusions: Such levels of classification accuracy are promising as they are comparable to those obtained by general practitioners in KC screening. © The Authors. Published by SPIE under a Creative Commons Attribution 4.0 International License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI. [DOI: 10.1117/1.JBO.30.3.035001] Keywords: skin cancer; optical biopsy; multimodal spectroscopic methods; diffuse reflectance spectroscopy; autofluorescence spectroscopy; machine learning Paper 240286GR received Oct. 8, 2024; revised Dec. 22, 2024; accepted Jan. 21, 2025; published Mar. 4, 2025 . *Address all correspondence to Marine Amouroux, [marine.amouroux@univ-lorraine.fr](marine.amouroux@univ-lorraine.fr) Journal of Biomedical Optics 035001-1 March 2025 • Vol. 30(3) Kupriyanov et al.: Machine learning–based classification of spatially resolved diffuse. . . 1 Introduction In 2022, keratinocyte carcinoma (KC) was the fifth most common cancer type in the world.1 Basal cell carcinoma (BCC) and squamous cell carcinoma (SCC) are the most common types of KC, accounting for 70% and 25% of cases, respectively. Metastasis occurs in only 1 out of 14,000,000 cases for BCC and in <10% for SCC,2 inducing an overall 90% 5-year survival rate. The age-standardized incidence rate of KC increased from 54.08/100,000 in 1990 to 79.10/ 100,000 in 2019 (þ 1.78% per year), and the number of reported cases is increasing every 3–5 year. This increasing incidence entails an important medico-economic burden. Indeed, the standard procedure for suspicious skin lesion diagnosis includes surgical biopsy and histopathological analysis of the removed tissue samples.6,7 The ratio of biopsies to treatment increased from 1. 1 in 1993 to 2 . 1 in 2016 .8 In addition, biopsy is an invasive, long, and expensive procedure, which induces stress

FAQs

What accuracy ranges were reported for different diagnostic groupings?

Binary tests discriminating BCC or SCC versus healthy skin exceeded 80%, while actinic keratosis versus other classes achieved about 65%–75% accuracy, and multiclass accuracy reached about 57%.

Sep 23, 2026

Uploaded by Oliver