Dermatologists are starting to make use of Computer-Aided Diagnosis based on deep learning algorithms, which can provide them with an objective judgement during evaluation of equivocal lesions. DL algorithms can be trained to classify skin lesions with datasets of diverse nature like traditional RGB, clinical and dermoscopic images, or more experimentally, with images from other modalities, such as multispectral imaging. In this work, we have evaluated and customized the different DL approaches that exist in the state of the art to classify a dataset of +500 images acquired on skin lesions. The images were acquired with a staring multispectral imaging prototype in the visible and near-infrared ranges. The best results were obtained for a customized model VGG-16 that combined 3D convolutional layers, 3D maxpooling layers and dropout regularization, leading to an overall accuracy of 71%.
Spectral reflectance of the eye fundus was evaluated in adult healthy patients through a fast visible and near-infrared multispectral fundus camera. Spectral signatures were analyzed for different ocular structures of the retina and the choroid.
We present a multispectral fundus camera that performs fast imaging of the ocular posterior pole in the visible and near-infrared (400 to 1300 nm) wavelengths through 15 spectral bands, using a flashlight source made of light-emitting diodes, and CMOS and InGaAs cameras. We investigate the potential of this system for visualizing occult and overlapping structures of the retina in the unexplored wavelength range beyond 900 nm, in which radiation can penetrate deeper into the tissue. Reflectance values at each pixel are also retrieved from the acquired images in the analyzed spectral range. The available spectroscopic information and the visualization of retinal structures, specifically the choroidal vasculature and drusen-induced retinal pigment epithelium degeneration, which are hardly visible in conventional color fundus images, underline the clinical potential of this system as a new tool for ophthalmic diagnosis.
The effective and non-invasive diagnosis of skin cancer is a hot topic in biophotonics since the current gold standard, biopsy followed by histological examination, is a slow and costly procedure for the healthcare system. Therefore, authors have put their efforts in characterizing skin cancer quantitatively through optical and photonic techniques such as 3D topography and multispectral imaging. Skin relief is an important biophysical feature that can be difficult to appreciate by touch, but can be precisely characterized with 3D imaging techniques, such as fringe projection. Color and spectral features given by skin chromophores, which are routinely analyzed by the naked eye and through dermoscopy, can also be quantified by means of multispectral imaging systems. In this study, the outcomes of these two imaging modalities were combined in a machine learning process to enhance classification of melanomas and nevi obtained from the two systems when operating isolately. The results suggest that the combination of 3D and multispectral data is relevant for the medical diagnosis of skin cancer.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.