Surgical resection of skin cancer implies safety margins delineation: currently, surgeons have no diagnostic aid to narrow or widen such margins if necessary. A promising approach is the use of optical methods, which can be used non-invasively and offer real-time diagnostic assistance.
This study presents the results of classification of autofluorescence (AF) and diffuse reflectance (DR) spectra obtained in vivo on skin Basal Cell Carcinomas (BCC) and Squamous Cell Carcinomas (SCC), Actinic Keratoses (AK) and Healthy skin (H) of 140 patients. The bimodal spectroscopic instrument used in this study uses five LEDs for fluorescence excitation at wavelengths peaks between 365 and 415 nm, and a xenon lamp featuring 350-800 nm emission range to obtain AF and DR spectra for four source-detector distances (from 400 to 1000 μm).
The classification (C vs H, H vs AK) was done by support vector machine, discriminant analysis, and multilayer perceptron. Final accuracy of two-class classification tests for almost all pairs of classes was more than 80%. This study presents a comparison of the performance of these combination of methods with the standard clinical procedure.
Information on skin phototype and ages is of cosmetic and medical interest in some procedures like objective evaluation of cosmetic treatments effectiveness, laser wavelength choice, risk of skin cancer recurrence and skin evaluation before cosmetic surgeries. Phototype may be evaluated using the Fitzpatrick questionnaire whose results are impaired by patients’ subjective answers; melanometers may be used but are not always available in dermatology practice. Tewameter, corneometer or cutometer are used to evaluate skin features that may be related to skin age but they lack evaluation of skin internal structure directly related to skin age (fibrosis, elastosis, etc.). Optical spectroscopy combining autofluorescence (AF) and diffuse reflectance (DR) may be a promising and non-invasive alternative to these tests.
In the current study, a bimodal spectroscopic device was used to obtain in vivo spatially resolved AF and DR spectra of skin in the visible range. Five LEDs featuring wavelength peaks at 365, 385, 395, 400 and 415 nm and a xenon lamp featuring a 350-800 nm spectral emission were used as light sources. Four source-detector separation (SDS) were used: 400, 600, 800, and 1000 μm.
Spectra were taken in different anatomical sites on 131 patients of different age and gender during a clinical study. Spectra were analysed using classification (support vector machine and multilayer perceptron) and regression (multilayer perceptron, linear, kernel ridge and Lasso) methods. Results of skin phototype and age estimation from AF and DR spectra obtained in vivo using machine learning methods will be presented and discussed.
Kidney stones are a global problem that cause physical pain and may lead to chronic kidney disease. Recent statistics indicate the incidence of kidney stones is increasing worldwide, and usually varies from 2 to 20% depending on countries1 and especially on diabetes or obesity incidence in such countries. Intra-operative (i.e. in vivo) characterization of kidney stones is at stake for a better diagnostic management of patients. Such a goal could be achieved by optical methods. The current study aims at evaluating if absorption and scattering coefficients measurements combined to automatic classification based on machine-learning methods could be of interest in assisting urologists with kidney stones characterization. Absorption and scattering coefficients were measured using the inverse adding doubling method (IAD). This method based on solving inverse problem takes as input data measurements acquired on a double integrating spheres optical bench developed in the CRAN laboratory. The dataset is made of absorption and scattering coefficients measured every 10 nm from 535 to 845 nm on 16 kidney stones: 4 kidney stones in each diagnostic class under consideration (1a, 3a, 4c and 5a). Class 3a (5a respectively) kidney stones display the highest (lowest resp.) absorption and scattering coefficients: 3 and 30 mm-1 (1 and 10 mm-1 respectively) at 650 nm. Support-vector machine (SVM) and k-nearest neighbors (k-NN) methods were used to perform automatic classification: k-NN reached 98%-accuracy in the four-class confusion matrix when considering both absorption and scattering coefficients. Although a high intra-class variability was observed and may be seen as the main limitation of the study, this good classification rate is worth taking into account to keep on investigating this method on more kidney stones per class as a potential tool for diagnostic assistance for urologists.
KEYWORDS: Skin, Skin cancer, Feature extraction, Data processing, Principal component analysis, In vivo imaging, Diffuse reflectance spectroscopy, Autofluorescence
Non-invasive diagnosis of skin pathologies as skin cancer using optical methods has become increasingly common in recent years. However, the related skin data processing is often quite complex, and the way in which this step is carried out can significantly affect the final results. This study presents the results of diffuse reflectance spectra (with spectral range of the emission source is 300-800 nm) and autofluorescence spectra (with 7 autofluorescence excitation wavelengths in the 360-430 nm range) obtained in vivo from precancerous and benign skin lesions of various types (compensatory hyperplasia, atypical hyperplasia and dysplasia). The skin lesions were modelled using a preclinical model in mice. Spectra were taken in the range of 317 - 789 nm at three different source-detector separations: 271, 536 and 834 μm. The spectra obtained were processed using statistical feature extraction techniques, traditional machine learning (support vector machine, linear discriminant analysis, k-nearest neighbors) and deep learning methods (artificial neural network, convolutional neural network, autoencoder). This study presents a comparison of the performance of these methods and their combinations for multiclass classification of skin lesions.
Optical biopsy methods, which consists of analysing the response of tissue to light excitation, are being increasingly used in recent years for the diagnosis of skin pathologies. At the same time, the use of multimodal methods often significantly increases diagnostic efficiency as well as extending the limits of applicability of the methods. This contribution presents the results of in vivo analysis of precancerous and benign skin conditions (compensatory hyperplasia, atypical hyperplasia and dysplasia) in mice preclinical model, based on bimodal spectroscopic data, including multiply excited autofluorescence with 7 autofluorescence excitation wavelengths in the 360-430 nm range and diffuse reflectance spectroscopy with xenon lamp, that emits mainly in the 300-800 nm spectral range, as a source. The instrument used in this study provided the ability to collect spectra in the spectral range 317 - 789 nm at three different source-detector separations: 271, 536 and 834 μm. The results were processed using machine learning methods (principal component analysis, support vector machine, linear discriminant analysis, artificial neural network) and then various data fusion methods (Stacking, Begging, Boosting, Voting) were implemented to combine the results of analysis of all the modalities. This study presents a comparison of the performance of these data fusion methods. The results obtained in this work can be further applied to the diagnosis of carcinoma using optical biopsy methods.
Whether for diagnosis, therapy or surgery, the estimation of optical properties (OP) of biological tissues is now of interest in the medical context. Indeed, optical methods are increasingly used in modern medicine, and these require knowledge of the behavior of light within the tissue. The presentation contribution aims to validate the estimation process of absorption and scattering coefficients values obtained using spatially-resolved diffuse reflectance (SR-DR) spectroscopy by comparing the obtained results with those of the reference double integrating spheres (DIS) technique. A set of nine optical phantoms based on methylene blue and intralipids allowing to tune absorption and scattering properties was prepared, from which diffuse reflectance spectra and integrating spheres measurements were acquired respectively. Work presented here reports both estimations approaches developed and highlights the relative spreads of optical properties between DR, DIS and theoretical values (i.e. according scatterer and absorber concentrations introduced in phantoms). This validation on optical bench will allow to later estimate the OP from in-vivo DR spectra acquired on skin samples, to assist the surgeon in non-invasively diagnosing the health status of a tissue around a skin carcinoma
For a favorable treatment result, early diagnosis of pathological cancerous micro-areas with their subsequent removal is highly important and can be achieved by the development of new modeling techniques and conducting relevant experiments. Various models of the bladder can be developed and applied to provide a platform for studying, processing and improving the signals received from various video systems. Here, in order to study visualization properties at fluorescence endoscopy, 3D optical phantoms of urinary bladder have been developed. The phantoms simulated optical properties of the bladder wall, including localized areas that represent tumor tissues and contained PpIX photosensitizer at various concentrations for fluorescence "diagnostics". To perform bimodal fluorescence imaging, a two-channel video fluorescence system was used. First, intraoperative images of the bladder wall were obtained in a patient with bladder cancer. A video system was used to reveal and image pathological areas with increased fluorescence intensity. Fluorescence indices in tumor tissue were recorded and corresponded to different concentrations of PpIX photosensitizer. Then, a bimodal fluorescence imaging was performed on 3D phantoms. The obtained images and fluorescence intensity measurements showed the ability of the video fluorescence system to register bladder wall structures and accumulated in them photosensitizers in concentrations from 0.25 to 20 mg/kg. The developed models can serve as a useful instrument for test measurements for constructing multimodal mosaic panoramic images of the bladder surface. This will help to advance in solving problems of endoscopic image processing using bimodal imaging, which uses diagnostic (fluorescence) and color channels.
In the context of optical biopsy for the diagnosis of skin carcinoma, spatially resolved diffuse reflectance (SR-DR) spectroscopy is widely used to discern healthy from lesional tissues. The estimation of diagnostically relevant optical properties by means of inverse problem solving is one way to exploit the acquired clinical spectra. This method requires the comparison between the latter spectra collected with a medical device (MD), and the ones generated by the photons transport numerical simulations. However, this comparison is typically limited to shape comparison (spectra are normalized before a term-by-term comparison) due to non-standardization of the experimental DR spectra, for which magnitude depends on the multifiber optics probe geometry and on a preliminary calibration measurement performed on a spectralon DR standard illuminated at a given distance. This study proposes to establish a corrective factor to overcome this dependence, and thus obtain clinical spectra whose intensity unit is identical to the simulated ones, i.e., the ratio between photons sent by the emitting fiber and captured by the collecting fibers. The photometric calculations leading to a theoretical value of this factor for various calibration measurement geometries are presented. Experimental validations performed on optical phantoms (with optical properties confirmed from double integrating sphere measurements) using an existing SR-DR MD reveal encouraging fitting between experimental and simulated calculation of such corrective factor. Those results highlight the interest of the method for the standardization of clinically acquired DR spectra i.e. their comparison in terms of absolute magnitudes.
Spatially resolved diffuse reflectance spectroscopy (SR-DRS) is a widely studied optical biopsy technique to investigate and to diagnose skin tissue modifications due to pathologies such as cancers. One way to exploit clinical spectra acquired with a SR-DRS medical device consists in estimating diagnostically relevant skin optical properties that is, by solving an inverse problem based on numerical simulations to generate spectra in accordance with the technical and geometrical features of the latter device. For realistic multi-layer skin media, the simultaneous estimation of layer-wise optical properties of interest is quite challenging (difficulty of convergence or non-unicity of the solution) and time consuming, especially for one or several parameters to be estimated in more than three layers of a skin model. To tackle this problem, the work presented here proposes an improved inverse problem solving scheme, which (i) sequentially determines the parameters of interest, layer by layer, in a 5-layer skin model using (ii) a custom cost-function adapted to the layered structure of the skin, i.e. considering wavelength and source-detector distance sensitivity to each layer. In-silico validation of the proposed approach was performed through convergence analysis towards ground truth simulated spectra. Using this sequential approach, the values of a 4-parameters vector were estimated with a relative errors of a few percent only and three times faster compared to current optimization method. Moreover, it brings morphological and physiological dimension to the inverse problem solving.
KEYWORDS: Spectroscopes, In vivo imaging, Skin, Tissues, Tissue optics, Scattering, Photodetectors, Monte Carlo methods, Inverse optics, Diffuse reflectance spectroscopy
In the context of cutaneous carcinoma in vivo diagnosis, Diffuse Relectance (DR) acquired using Spatially Resolved (SR) optical biopsy, can be analysed to discard healthy from pathological areas. Indeed, carcinogenesis induces local morphological and metabolic changes affecting the skin optical answer to white light excitation. The present contribution aims at studying the epidermis thickness impact on the path and propagation depth distribution of DR photons in skin in the perspective of analyzing how these photons contribute to the corresponding acquired spectra carrying local physiological information from the visited layers. Modified CudaMCML-based simulations were performed on a five-layer human skin optical model using (i) wavelength-resolved scattering and absorption properties and (ii) the geometrical configuration of a multi-optical fiber probe implemented on a SR-DR spectroscopic device currently used in clinics. Through maps of scattering events and histograms of maximum probed depth, we provide numerical evidences linking the characteristic penetration depth of the detected photons to their wavelengths and four source-sensor distances for thin, intermediate and wide skin thicknesses model. The study provides qualitative and quantitative tools that can be useful during the design of an optical SR-DR spectroscopy biopsy device.
In the context of cutaneous carcinoma in vivo diagnosis, Diffuse Relectance (DR) and skin endogenous fluores- cence (AF) spectra, acquired using Spatially Resolved (SR) multimodal optical biopsy, can be analysed to discard healthy from pathological areas. Indeed, carcinogenesis induces morphological and metabolic changes affecting endogenous fluorophores such as for instance elastosis and enzymatic degradation of collagen fluorescence in the dermis or decreased NADH fluorescence in the epidermis. The present contribution aims at studying the path and propagation depth distribution of DR and AF photons in skin in the perspective of analyzing how these photons contribute to the corresponding acquired spectra carrying local physiological information. Modified CudaMCML-based simulations were performed on a five-layer human skin optical model with (i) wavelength resolved scattering, absorption and endogenous fluorescence properties and (ii) multiple fiber optic probe ge- ometrical configuration of a SR-DR and -AF spectroscopic device. The simulation results provided numerical evidences of the behaviour of detected photons in the tissue. In particular, we succeeded in linking the character- istic penetration depth of the detected photons to their wavelengths and the source-sensor distance. In addition, we managed to identify the region where the fluorescence events associated with the AF spectrum photon take place. The study provides qualitative and quantitative tools that can be useful during the design of an optical multimodal biopsy device.
A 5-ALA-induced fluorescence-based imaging device for guidance during surgery of malignant and non-malignant preliminary photosensitized tumors is presented. The setup fits existing clinical optical rigid and flexible endoscopes and operation microscopes. It consists of three light sources including white light, red light fluorescence excitation and blue light fluorescence excitation sources. The light from any combination of the latter sources is delivered to tissue using specially designed fiber optic light guide. Two cameras are used to acquire fluorescence and back reflected white light images: a gray-level camera for fluorescence in the far red range and a color camera for white light images. A dichroic mirror is implemented to spectrally split the light coming from tissue. Images from both cameras are processed into a computer with specially developed software where it can be displayed in different modes including overlaying or been used for image mosaicing which allows for increasing the intrinsic reduced field of view of endoscopes by providing highly resolved extended cartography. Experiments were carried out on phantoms and on patients in clinical conditions during surgery of brain and other tissues. Blue light excitation was more sensitive for thin tumors but red light excitation was more beneficial for solid tumors and for navigation in presence of slight bleeding.
This contribution presents a fast global adjustment scheme exploiting SURF descriptor locations for constructing large skin mosaics. Precision in pairwise image registration is well-preserved while significantly reducing the global mosaicing error.
Cystoscopy is the standard procedure for clinical diagnosis of bladder cancer diagnosis. Bladder carcinoma in situ are often multifocal and spread over large areas. In vivo, localization and follow-up of these tumors and their nearby sites is necessary. But, due to the small field of view (FOV) of the cystoscopic video images, urologists cannot easily interpret the scene. Bladder mosaicing using image registration facilitates this interpretation through the visualization of entire lesions with respect to anatomical landmarks. The reference white light (WL) modality is affected by a strong variability in terms of texture, illumination conditions and motion blur. Moreover, in the complementary fluorescence light (FL) modality, the texture is visually different from that of the WL. Existing algorithms were developed for a particular modality and scene conditions. This paper proposes a more general on fly image registration approach for dealing with these variability issues in cystoscopy. To do so, we present a novel, robust and accurate image registration scheme by redefining the data-term of the classical total variational (TV) approach. Quantitative results on realistic bladder phantom images are used for verifying accuracy and robustness of the proposed model. This method is also qualitatively assessed with patient data mosaicing for both WL and FL modalities.
Determining the optical properties of biological tissues in vivo from spectral intensity measurements performed at their surface is still a challenge. Based on spectroscopic data acquired, the aim is to solve an inverse problem, where the optical parameter values of a forward model are to be estimated through optimization procedure of some cost function. In many cases it is an ill-posed problem because of small numbers of measures, errors on experimental data, nature of a forward model output data, which may be affected by statistical noise in the case of Monte Carlo (MC) simulation or approximated values for short inter-fibre distances (for Diffusion Equation Approximation (DEA)). In case of optical biopsy, spatially resolved diffuse reflectance spectroscopy is one simple technique that uses various excitation-toemission fibre distances to probe tissue in depths. The aim of the present contribution is to study the characteristics of some classically used cost function, optimization methods (Levenberg-Marquardt algorithm) and how it is reaching global minimum when using MC and/or DEA approaches. Several methods of smoothing filters and fitting were tested on the reflectance curves, I(r), gathered from MC simulations. It was obtained that smoothing the initial data with local regression weighted second degree polynomial and then fitting the data with double exponential decay function decreases the probability of the inverse algorithm to converge to local minima close to the initial point of first guess.
Bladder cancer is widely spread in the world. Many adequate diagnosis techniques exist. Video-endoscopy
remains the standard clinical procedure for visual exploration of the bladder internal surface. However, video-endoscopy
presents the limit that the imaged area for each image is about nearly 1 cm2. And, lesions are,
typically, spread over several images. The aim of this contribution is to assess the performance of two mosaicing
algorithms leading to the construction of panoramic maps (one unique image) of bladder walls. The quantitative
comparison study is performed on a set of real endoscopic exam data and on simulated data relative to bladder
phantom.
Bladder cancer is widely spread. Moreover, carcinoma in situ can be difficult to diagnose as it may be difficult to see,
and become invasive in 50 % of case. Non invasive diagnosis methods like photodynamic or autofluorescence
endoscopy allow enhancing sensitivity and specificity. Besides, bladder tumors can be multifocal. Multifocality
increases the probability of recurrence and infiltration into bladder muscle. Analysis of spatial distribution of tumors
could be used to improve diagnosis. We explore the feasibility to combine fluorescence and spatial information on
phantoms. We developed a system allowing the acquisition of consecutive images under white light or UV excitation
alternatively and automatically along the video sequence. We also developed an automatic image processing algorithm
to build a partial panoramic image from a cystoscopic sequence of images. Fluorescence information is extracted from
wavelength bandpass filtered images and superimposed over the cartography. Then, spatial distribution measures of
fluorescent spots can be computed. This cartography can be positioned on a 3D generic shape of bladder by selecting
some reference points. Our first results on phantoms show that it is possible to obtain cartography with fluorescent spots
and extract quantitative information of their spatial distribution on a "wide" field of view basis.
The contribution aims at describing a computer-based structured light imaging system to be applied to automated recovery of quantitative 3D information on sculptured surfaces, in order to take in charge (industrial) inspection/3D reconstruction tasks. Recovery is based on evaluation of images of the light pattern induced by projection into the scene of a specifically deviced parallel grid. The system has been designed for direct use in industrial environments, e.g. for integration into on-line quality control systems. Consequently, particular emphasis has been put on efforts for fulfilling requirements usually implied by this type of application, such as simplicity of set-up, application real-time, high accuracy, and low cost. This paper gives a discription of the system realized, including the algorithms specifically designed and implemented for calibration, nonambiguous labeling of the imaged fringes, and subpixel evaluation of their locations. The integration of the system into an on-line inspection system for 100% control of manufactured parts illustrates its application. Inspection is based on comparison of extracted features gained from a CAD model of the part and including tolerance information. Currently, a measurement accuracy of the order of 25 micrometers can be routinely achieved.
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.