Please use this identifier to cite or link to this item:
Title: Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images
Authors: La Salvia, Marco
Torti, Emanuele
León Martín, Sonia Raquel 
Fabelo Gómez, Himar Antonio 
Ortega, Samuel 
Balea Fernandez, Francisco Javier 
Martínez Vega, Beatriz 
Castaño González, Irene
Almeida Martín, Pablo Julio 
Carretero Hernández, Gregorio
Hernández, Javier A.
Marrero Callicó, Gustavo Iván 
Leporati, Francesco
UNESCO Clasification: 3314 Tecnología médica
320106 Dermatología
Keywords: Skin cancer
Hyperspectral imaging
Deep learning
Disease diagnosis
High-performance computing
Issue Date: 2022
Project: Talent Imágenes Hiperespectrales Para Aplicaciones de Inteligencia Artificial 
Journal: Sensors (Switzerland) 
Abstract: Cancer originates from the uncontrolled growth of healthy cells into a mass. Chromophores, such as hemoglobin and melanin, characterize skin spectral properties, allowing the classification of lesions into different etiologies. Hyperspectral imaging systems gather skin-reflected and transmitted light into several wavelength ranges of the electromagnetic spectrum, enabling potential skin-lesion differentiation through machine learning algorithms. Challenged by data availability and tiny inter and intra-tumoral variability, here we introduce a pipeline based on deep neural networks to diagnose hyperspectral skin cancer images, targeting a handheld device equipped with a low-power graphical processing unit for routine clinical testing. Enhanced by data augmentation, transfer learning, and hyperparameter tuning, the proposed architectures aim to meet and improve the well-known dermatologist-level detection performances concerning both benign-malignant and multiclass classification tasks, being able to diagnose hyperspectral data considering real-time constraints. Experiments show 87% sensitivity and 88% specificity for benign-malignant classification and specificity above 80% for the multiclass scenario. AUC measurements suggest classification performance improvement above 90% with adequate thresholding. Concerning binary segmentation, we measured skin DICE and IOU higher than 90%. We estimated 1.21 s, at most, consuming 5 Watts to segment the epidermal lesions with the U-Net++ architecture, meeting the imposed time limit. Hence, we can diagnose hyperspectral epidermal data assuming real-time constraints.
ISSN: 1424-8220
DOI: 10.3390/s22197139
Source: Sensors (Switzerland) [ISSN 1424-8220], v. 22 (19), 7139, (Septiembre 2022)
Appears in Collections:Artículos
Adobe PDF (2,44 MB)
Show full item record

Google ScholarTM




Export metadata

Items in accedaCRIS are protected by copyright, with all rights reserved, unless otherwise indicated.