Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images
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 ligh...
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-09-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/22/19/7139 |
| _version_ | 1797477006105903104 |
|---|---|
| author | Marco La Salvia Emanuele Torti Raquel Leon Himar Fabelo Samuel Ortega Francisco Balea-Fernandez Beatriz Martinez-Vega Irene Castaño Pablo Almeida Gregorio Carretero Javier A. Hernandez Gustavo M. Callico Francesco Leporati |
| author_facet | Marco La Salvia Emanuele Torti Raquel Leon Himar Fabelo Samuel Ortega Francisco Balea-Fernandez Beatriz Martinez-Vega Irene Castaño Pablo Almeida Gregorio Carretero Javier A. Hernandez Gustavo M. Callico Francesco Leporati |
| author_sort | Marco La Salvia |
| collection | DOAJ |
| description | 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. |
| first_indexed | 2024-03-09T21:11:49Z |
| format | Article |
| id | doaj.art-29cef25421c34983b8b190a74d80bdf9 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-09T21:11:49Z |
| publishDate | 2022-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-29cef25421c34983b8b190a74d80bdf92023-11-23T21:44:05ZengMDPI AGSensors1424-82202022-09-012219713910.3390/s22197139Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal ImagesMarco La Salvia0Emanuele Torti1Raquel Leon2Himar Fabelo3Samuel Ortega4Francisco Balea-Fernandez5Beatriz Martinez-Vega6Irene Castaño7Pablo Almeida8Gregorio Carretero9Javier A. Hernandez10Gustavo M. Callico11Francesco Leporati12Department of Electrical, Computer and Biomedical Engineering, University of Pavia, 27100 Pavia, ItalyDepartment of Electrical, Computer and Biomedical Engineering, University of Pavia, 27100 Pavia, ItalyResearch Institute for Applied Microelectronics (IUMA), University of Las Palmas de Gran Canaria (ULPGC), 35001 Las Palmas de Gran Canaria, SpainResearch Institute for Applied Microelectronics (IUMA), University of Las Palmas de Gran Canaria (ULPGC), 35001 Las Palmas de Gran Canaria, SpainResearch Institute for Applied Microelectronics (IUMA), University of Las Palmas de Gran Canaria (ULPGC), 35001 Las Palmas de Gran Canaria, SpainDepartment of Psychology, Sociology and Social Work, University of Las Palmas de Gran Canaria, 35001 Las Palmas de Gran Canaria, SpainResearch Institute for Applied Microelectronics (IUMA), University of Las Palmas de Gran Canaria (ULPGC), 35001 Las Palmas de Gran Canaria, SpainDepartment of Dermatology, Hospital Universitario de Gran Canaria Doctor Negrín, Barranco de la Ballena, s/n, 35010 Las Palmas de Gran Canaria, SpainDepartment of Dermatology, Complejo Hospitalario Universitario Insular-Materno Infantil, Avenida Maritima del Sur, s/n, 35016 Las Palmas de Gran Canaria, SpainDepartment of Dermatology, Hospital Universitario de Gran Canaria Doctor Negrín, Barranco de la Ballena, s/n, 35010 Las Palmas de Gran Canaria, SpainDepartment of Dermatology, Complejo Hospitalario Universitario Insular-Materno Infantil, Avenida Maritima del Sur, s/n, 35016 Las Palmas de Gran Canaria, SpainResearch Institute for Applied Microelectronics (IUMA), University of Las Palmas de Gran Canaria (ULPGC), 35001 Las Palmas de Gran Canaria, SpainDepartment of Electrical, Computer and Biomedical Engineering, University of Pavia, 27100 Pavia, ItalyCancer 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.https://www.mdpi.com/1424-8220/22/19/7139skin cancerhyperspectral imagingdeep learningdisease diagnosishigh-performance computing |
| spellingShingle | Marco La Salvia Emanuele Torti Raquel Leon Himar Fabelo Samuel Ortega Francisco Balea-Fernandez Beatriz Martinez-Vega Irene Castaño Pablo Almeida Gregorio Carretero Javier A. Hernandez Gustavo M. Callico Francesco Leporati Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images Sensors skin cancer hyperspectral imaging deep learning disease diagnosis high-performance computing |
| title | Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images |
| title_full | Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images |
| title_fullStr | Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images |
| title_full_unstemmed | Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images |
| title_short | Neural Networks-Based On-Site Dermatologic Diagnosis through Hyperspectral Epidermal Images |
| title_sort | neural networks based on site dermatologic diagnosis through hyperspectral epidermal images |
| topic | skin cancer hyperspectral imaging deep learning disease diagnosis high-performance computing |
| url | https://www.mdpi.com/1424-8220/22/19/7139 |
| work_keys_str_mv | AT marcolasalvia neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT emanueletorti neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT raquelleon neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT himarfabelo neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT samuelortega neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT franciscobaleafernandez neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT beatrizmartinezvega neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT irenecastano neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT pabloalmeida neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT gregoriocarretero neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT javierahernandez neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT gustavomcallico neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages AT francescoleporati neuralnetworksbasedonsitedermatologicdiagnosisthroughhyperspectralepidermalimages |