We demonstrate a Bayesian statistics-based outlier separation algorithm, which clearly distinguishes microscope captured images of unstained human cervical tissue sections of normal and different grades of precancerous tissues. The semi-automated global and adaptive method implements outlier separation based on the statistical characterization of the image histogram distribution. This multi-level thresholding achieves an effective image quantization of the high cell density domain, most affected in the progression of the disease, which yields a precise visualization of the lesions in the epithelium cellular structures, revealing their temporal changes with the progression of the disease. The pixel count ratio of the quantized high cell density region, below a statistically well-defined threshold, quantitatively discriminates different grades of precancer tissues through Receiver Operating Characteristics.
Cervical cancer ranks as the fourth most prevalent cancer globally, emphasizing the critical need for early detection, which is vital for effective treatment. Traditional diagnostic methods have shown limitations in detecting the progression of the disease. Optical techniques, known for their high sensitivity and specificity, are emerging as reliable tools, especially in cancer-related applications. Among these techniques, fluorescence spectroscopy is one of the highly sensitive approaches for identifying biochemical changes that occur during the advancement of cancer. In our study, fluorescence spectral data was collected from human cervix from a diverse group of individuals using a portable smartphone-based fluorescence spectroscopy device. The spectral signals were processed by initially breaking them down into Fourier Bessel series (FBS) coefficients. Subsequently, the Hessian locally linear embedding (HLLE) based dimensionality reduction method was applied to the FBS coefficients, followed by the implementation of a 1D convolutional neural network classifier. The combination of polarized fluorescence spectra acquired from the device and the proposed classification approach has shown promising results, thus it is proven to be a minimally invasive method with the capability to provide real-time diagnoses for patients
Cervical cancer is of significant health concern globally, particularly in developing countries where access to advanced healthcare facilities and medical resources is limited, leading to increased mortality rates. The gold standard for diagnosing CIN (cervical intraepithelial neoplasia) and invasive cervical cancer involves performing a colposcopy-guided biopsy followed by a pathological diagnosis. However, its effectiveness is challenged by limited sensitivity in distinguishing between various stages of cervical cancer, especially in regions where there is a shortage of skilled colposcopists and insufficient access to medical resources. This study presents a method for categorizing infectious, pre-cancerous, and cancerous conditions through the application of multifractal analysis, specifically two dimensional multifractal detrended fluctuation analysis (2D MFDFA), using images obtained through colposcopy. The utilization of multifractal parameters, namely the generalized Hurst exponent and the width of the singularity spectrum, in the analysis distinctly demonstrated variations among the infectious, precancerous, and cancerous conditions. Therefore, it offers valuable insights to healthcare professionals, assisting in the accurate classification and effective management of cervical cancer using Hurst exponent and multifractal spectrum width.
In this paper a novel two-stage adaptive framework for denoising of differential interference contrast (DIC) images followed by Gabor based gray-level co-occurrence matrix (GLCM) feature extraction methodology is proposed. The first stage consists of a hybrid cascade of anisotropic diffusion denoising (Perona–Malik diffusion) and unsharp masking (USM) based detail enhancement filter to remove noise from DIC images without losing significant morphological features of healthy and precancerous tissues while enlarging the image sharpness. The hybrid filter parameters are obtained by joint stochastic optimization of the image quality metrics. The estimated denoised image with the highest signal to noise ratio (SNR) from Stage I, is used for subsequent textural feature extraction. GLCM window considers neighborhood blocks with similar local statistics with well-preserved local structures between a pixel texture and its nearest neighbors. The efficacy of our denoised DIC imaging with Gabor based GLCM feature descriptors in analysis of healthy and precancerous tissues is experimentally validated for its competitive denoising performance and detail structure preservation of DIC images. The relative change in magnitude and phase information as manifested from Gabor filter coupled with GLCM based spatial statistical measures of tissues as cancer progress validates the adequacy of the proposed scheme for its early stage cancer detection ability in cervical tissues.
In this contribution, we have done exploratory experiments using deep learning framework to classify elastic scattering spectra of biological tissues into normal and cancerous ones. An analytical assessment highlighting the superiority of convolutional neural network (CNN) extracted deep features over classical hand crafted biomarkers is discussed. The proposed method employs elastic scattering spectra of the tissues as input to CNN and thereby, averting the requirement of domain experts for extraction of diagnostic feature descriptors. Experimental results are discussed in detail.
KEYWORDS: Tissue optics, Tissues, In vivo imaging, Statistical analysis, Data analysis, Eye, Refractive index, Scientific research, Analytical research, Biological research
Diabetic Retinopathy (DR) is one of the most dominant diseases across the globe which causes blindness. In this manuscript, we have probed tissue multifractality in order to identify the submicron level changes in medium refractive indices due to progress of diabetic retinopathy from mild to severe stages. Hence the quantification of multifractal parameters like Hurst exponent (measurement of correlation) and width of singularity (measurement of heterogeneity) have been executed. As we proceed from healthy to different stages (mild, moderate and severe) of diabetic retinopathy, there are decrement of Hurst exponent value, whereas, width of singularity spectrum increases. In general, the use of multifractal analysis on in vivo diabetic retinopathy images lead to a diagnostic modality as a potential statistical biomarker.
Diabetic retinopathy damages retina due to diabetes mellitus which leads to blindness. Here, we have applied local binary pattern (LBP) in order to capture the spatial variations of the refractive indices due to progress of diabetic retinopathy among retinal tissues. After extraction of discriminative textures as binary numbers, state of art machine learning algorithms like decision tree and K-NN have been applied to get the optimum detection accuracy in multiclass classifications of in vivo diabetic retinopathy images. Here it is quite apparent that K-NN provides better accuracy and specificity than decision tree.
In this contribution, combined variational mode decomposition (VMD) aided non-linear feature descriptors & artificial neural network (ANN) for identification of different healthy and precancerous cervical tissues. Owing to the inherent problems of background laser system noise interferences in elastic scattering spectroscopic data, VMD method being noise robust is of paramount interest. VMD is used to decompose the normalized spectral data into 2 modes for analysis and attributes extraction. For each of these VMD separated modes, non-linear entropy and multifractal features, namely Shannon entropy (SE), Renyi entropy (RE), Tsallis entropy (TE) and Singularity spectrum width (SSW) are extracted to form the feature set. The extracted features are subjected to analysis of variance (ANOVA) test for subsequent feature ranking & selection of the statistically most significant features. The designated features are trained with ANN to classify the backscattered tissue spectra into healthy and cancerous ones.
This contribution proposes a novel extension of the existing ‘Hyper Kurtosis based Modified Duo-Histogram Equalization’ (HKMDHE) algorithm, for multi-objective contrast enhancement of biomedical images. A novel modified objective function has been formulated by joint optimization of the individual histogram equalization objectives. The optimal adequacy of the proposed methodology with respect to image quality metrics such as brightness preserving abilities, peak signal-to-noise ratio (PSNR), Structural Similarity Index (SSIM) and universal image quality metric has been experimentally validated. The performance analysis of the proposed Stochastic HKMDHE with existing histogram equalization methodologies like Global Histogram Equalization (GHE) and Contrast Limited Adaptive Histogram Equalization (CLAHE) has been given for comparative evaluation.
Background subtraction of knee MRI images has been performed, followed by edge detection through canny edge detector. In order to avoid the discontinuities among edges, Daubechies-4 (Db-4) discrete wavelet transform (DWT) methodology is applied for the smoothening of edges identified through canny edge detector. The approximation coefficients of Db-4, having highest energy is selected to get rid of discontinuities in edges. Hough transform is then applied to find imperfect knee locations, as a function of distance (r) and angle (θ). The final outcome of the linear Hough transform is a two-dimensional array i.e., the accumulator space (r, θ) where one dimension of this matrix is the quantized angle θ and the other dimension is the quantized distance r. A novel algorithm has been suggested such that any deviation from the healthy knee bone structure for diseases like osteoarthritis can clearly be depicted on the accumulator space.
In this paper, a comparative study between SVM and HMM has been carried out for multiclass classification of cervical healthy and cancerous tissues. In our study, the HMM methodology is more promising to produce higher accuracy in classification.
We report the application of a hidden Markov model (HMM) on multifractal tissue optical properties derived via the Born approximation-based inverse light scattering method for effective discrimination of precancerous human cervical tissue sites from the normal ones. Two global fractal parameters, generalized Hurst exponent and the corresponding singularity spectrum width, computed by multifractal detrended fluctuation analysis (MFDFA), are used here as potential biomarkers. We develop a methodology that makes use of these multifractal parameters by integrating with different statistical classifiers like the HMM and support vector machine (SVM). It is shown that the MFDFA-HMM integrated model achieves significantly better discrimination between normal and different grades of cancer as compared to the MFDFA-SVM integrated model.
We report fabrication of optical volume grating inside the quartz glass and PDMS sample using femtosecond laser based micromachining system with 10μm grating period and areaof 2mm x 6mm x 2mm thickness. The wavelength of femtosecond (fs) laser system is 775nm, repetition rate and pulse width of the system is 1 KHz and 100 fs respectively. Both embedded micro gratings are successfully demonstrated under same fabrication parameters. The comparison of diffraction efficiency and angle of diffraction up to the third order of diffraction for grating inside the quartz glass and PDMS has been examined at 632.8 nm, 532 nm and 447 nm respectively. The change in R.I of laser modified region was obtained ~10-3.
An augmented set of multifractal parameters with physical interpretations have been proposed to quantify the varying distribution and shape of the multifractal spectrum. The statistical classifier with accuracy of 84.17% validates the adequacy of multi-feature MFDFA characterization of elastic scattering spectroscopy for optical diagnosis of cancer.
In this contribution, we report the application of higher order statistical moments using decision tree and ensemble based learning methodology for the development of diagnostic algorithms for optical diagnosis of cancer. The classification results were compared to those obtained with an independent feature extractors like linear discriminant analysis (LDA). The performance and efficacy of these methodology using higher order statistics as a classifier using boosting has higher specificity and sensitivity while being much faster as compared to other time-frequency domain based methods.
In this paper, we make use of the empirical mode decomposition (EMD) to discriminate the cervical cancer tissues from normal ones based on elastic scattering spectroscopy. The phase space has been reconstructed through decomposing the optical signal into a finite set of bandlimited signals known as intrinsic mode functions (IMFs). It has been shown that the area measure of the analytic IMFs provides a good discrimination performance. Simulation results validate the efficacy of the IMFs followed by SVM based classification.
In this paper, the spectroscopy signals have been analyzed in recurrence plots (RP), and extract recurrence quantification analysis (RQA) parameters from the RP in order to classify the tissues into normal and different precancerous grades. Three RQA parameters have been quantified in order to extract the important features in the spectroscopy data. These features have been fed to different classifiers for classification. Simulation results validate the efficacy of the recurrence quantification as potential bio-markers for diagnosis of pre-cancer.
A probabilistic robust diagnostic algorithm is very much essential for successful cancer diagnosis by optical spectroscopy. We report here support vector machine (SVM) classification to better discriminate the colon and cervical cancer tissues from normal tissues based on elastic scattering spectroscopy. The efficacy of SVM based classification with different kernel has been tested on multifractal parameters like Hurst exponent, singularity spectrum width in order to classify the cancer tissues.
We demonstrate that the localized nonlinear gain induces stable chirped dissipative double-kink, fractional-transform,
bell, and kink type solitons in optical mediawith quadratic nonlinearity. To compensate spatially uniform loss in both
fundamental-frequency (FF) and second-harmonic (SH) component of the system, a strongly localized hot-spot, carrying
the nonlinear gain, is added, acting either on FF component or on the SH one.
We explore the exact optical similaritons of a generalized nonlinear Schrodinger equation(GNLSE) with space-time modulated
dispersion, nonlinearity, external potential and inhomogeneous source. As an application, we study the dynamics of
these similaritons for a spatial Bessel modulated nonlinearity.
Intrinsic fluorescence spectra of the human normal, cervical intraepithelial neoplasia 1 (CIN1), CIN2, and cervical cancer tissue have been extracted by effectively combining the measured polarized fluorescence and polarized elastic scattering spectra. The efficacy of principal component analysis (PCA) to disentangle the collective behavior from smaller correlated clusters in a dimensionally reduced space in conjunction with the intrinsic fluorescence is examined. This combination unambiguously reveals the biochemical changes occurring with the progression of the disease. The differing activities of the dominant fluorophores, collagen, nicotinamide adenine dinucleotide, flavins, and porphyrin of different grades of precancers are clearly identified through a careful examination of the sectorial behavior of the dominant eigenvectors of PCA. To further classify the different grades, the Mahalanobis distance has been calculated using the scores of selected principal components.
Differential interference contrast images (DIC) are the direct representation of the refractive index fluctuations of human cervical tissues. These refractive index fluctuations are known to follow self-similar behaviour and in general are multifractal in nature. In this present study, multifractal detrended fluctuation analysis (MFDFA) on refractive index fluctuations from DIC images has been performed by unfolding the tissue-images horizontally and vertically. Our analysis clearly shows that refractive index fluctuations of human cervical tissues are anisotropic-fractal in nature and anisotropy reduced as cancer progress.
DIC (Differential Interference Contrast Image) images of cervical pre-cancer tissues are taken from epithelium region, on which wavelet transform and multi-fractal analysis are applied. Discrete wavelet transform (DWT) through Daubechies basis are done for identifying fluctuations over polynomial trends for clear characterization and differentiation of tissues. A systematic investigation of denoised images is carried out through the continuous Morlet wavelet. The scalogram reveals the changes in coefficient peak values from grade-I to grade-III. Wavelet normalized energy plots are computed in order to show the difference of periodicity among different grades of cancerous tissues. Using the multi-fractal de-trended fluctuation analysis (MFDFA), it is observed that the values of Hurst exponent and width of singularity spectrum decrease as cancer progresses from grade-I to grade-III tissue.
The objective of the present work is to diagnose pre-cancer by wavelet transform and multi-fractal de-trended fluctuation analysis of DIC images of normal and different grades of cancer tissues. Our DIC imaging and fluctuation analysis methods (Discrete and continuous wavelet transform, MFDFA) confirm the ability to diagnose and detect the early stage of cancer in cervical tissue.
The two highest principal components of fluorescence spectra in visible region obtained, using Xenon lamp as an
excitation source of normal and dysplastic human cervical tissues are analyzed using scatter plots and probability
density functions. These yield significant differences between the tissue types.
Multi-resolution analysis on the spatial refractive index inhomogeneities in the connective tissue regions of human
cervix reveals clear signature of multifractality. We have thus developed an inverse analysis strategy for extraction and
quantification of the multifractality of spatial refractive index fluctuations from the recorded light scattering signal. The method is based on Fourier domain pre-processing of light scattering data using Born approximation, and its subsequent analysis through Multifractal Detrended Fluctuation Analysis model. The method has been validated on several mono- and multi-fractal scattering objects whose self-similar properties are user controlled and known a-priori. Following successful validation, this approach has initially been explored for differentiating between different grades of precancerous human cervical tissues.
In this work, we report a wavelet based multi-fractal study of images of dysplastic and neoplastic HE- stained
human cervical tissues captured in the transmission mode when illuminated by a laser light (He-Ne 632.8nm
laser). It is well known that the morphological changes occurring during the progression of diseases like cancer
manifest in their optical properties which can be probed for differentiating the various stages of cancer. Here,
we use the multi-resolution properties of the wavelet transform to analyze the optical changes. For this, we have
used a novel laser imagery technique which provides us with a composite image of the absorption by the different
cellular organelles. As the disease progresses, due to the growth of new cells, the ratio of the organelle to cellular
volume changes manifesting in the laser imagery of such tissues. In order to develop a metric that can quantify
the changes in such systems, we make use of the wavelet-based fluctuation analysis. The changing self- similarity
during disease progression can be well characterized by the Hurst exponent and the scaling exponent. Due to the
use of the Daubechies' family of wavelet kernels, we can extract polynomial trends of different orders, which help
us characterize the underlying processes effectively. In this study, we observe that the Hurst exponent decreases
as the cancer progresses. This measure could be relatively used to differentiate between different stages of cancer
which could lead to the development of a novel non-invasive method for cancer detection and characterization.
Using the multiresolution ability of wavelets and effectiveness of singular value decomposition (SVD) to identify statistically robust parameters, we find a number of local and global features, capturing spectral correlations in the co- and cross-polarized channels, at different scales (of human breast tissues). The copolarized component, being sensitive to intrinsic fluorescence, shows different behavior for normal, benign, and cancerous tissues, in the emission domain of known fluorophores, whereas the perpendicular component, being more prone to the diffusive effect of scattering, points out differences in the Kernel-Smoother density estimate employed to the principal components, between malignant, normal, and benign tissues. The eigenvectors, corresponding to the dominant eigenvalues of the correlation matrix in SVD, also exhibit significant differences between the three tissue types, which clearly reflects the differences in the spectral correlation behavior. Interestingly, the most significant distinguishing feature manifests in the perpendicular component, corresponding to porphyrin emission range in the cancerous tissue. The fact that perpendicular component is strongly influenced by depolarization, and porphyrin emissions in cancerous tissue has been found to be strongly depolarized, may be the possible cause of the above observation.
Wavelet Transform based multi-resolution analysis has been used to characterize the intrinsic fluorescence of both
dysplastic and normal human cervical tissues. The fluorescence spectra corresponding to 325nm and 370nm excitation
from cervical dysplastic tissues of 48 patients from diverse age groups are studied in detail using Morlet wavelet. The
wavelet modulus maxima lines for 325nm excitation indicated a distinct shift for dysplastic tissues towards the lower
wavelengths with respect to normal ones. For 370nm excitation however, the shift for dysplastic tissues were towards the
higher wavelengths. Sensitivities of 72% and 81% for spectral shifts of 325 and 370nm excitations were observed in the
wavelength band of 425-500nm of the intensity spectra.
A systematic investigation of the polarization characteristics of the auto-fluorescence of normal and benign human
breast tissues is carried out complementing our earlier studies on normal and cancer tissues. Co- and
cross-polarized auto-fluorescence are collected in the 500 to 700nm range through excitation at 488nm using
laser as excitation source. A number of parameters, capturing spectral variations are extracted in the co- and
cross-polarized channels through singular value decomposition and wavelet decomposition, which differentiate
normal and benign tissues. The correlation matrix differs significantly in normal and benign tissues reflecting
the presence of different fluorophores. The eigenvectors corresponding to the dominant eigenvalues reveal
differences between tissue types. The co-polarized component being sensitive to intrinsic fluorescence shows
different behavior for normal and benign tissues in the emission domain of known fluorophores. Interestingly, the
benign tissue samples show correlation properties intermediate to malignant and normal cases. In the wavelet
domain the standard deviation of percentage fluctuation reveal differences between tissues type. The correlation
characteristics manifest prominently in the wavelet low pass (average) domain.
Properties of spectral fluctuations and prominent spectral features of fluorescence spectra in visible region using
laser as an excitation source of normal, benign and cancer human breast tissues are studied through wavelet
transform and principal component analysis.
A systematic investigation of the fluorescence characteristics of normal and cancerous human breast tissues is
carried out, using laser and lamp as excitation sources. It is found that earlier observed subtle differences between
these two tissue types in the wavelet domain are absent, when lamp is used as excitation source. However, singular
value decomposition of the average spectral profile in the wavelet domain yields strong correlation for the cancer
tissues in the 580-750 nm regimes indicating weak fluorophore activity in this wavelength range.
We study the spectral correlation properties of the polarized fluorescence spectra of normal and cancerous human breast tissues, corresponding to patients belonging to diverse age groups and socioeconomic backgrounds. The emission range in the visible wavelength regime of 500 to 700 nm is analyzed, with the excitation wavelength at 488 nm, where flavin is one of the active fluorophores. The correlation matrices for parallel and perpendicularly polarized fluorescence spectra reveal correlated domains, differing significantly in normal and cancerous tissues. These domains can be ascribed to different fluorophores and absorbers in the tissue medium. The spectral fluctuations in the perpendicular component of the cancerous tissue clearly reveal randomization not present in the normal channel. Random matrix-based predictions for the spectral correlations match quite well with the observed behavior. The eigenvectors of the correlation matrices corresponding to large eigenvalues clearly separate out different tissue types and identify the dominant wavelengths, which are active in cancerous tissues.
The statistical and characteristic features of the polarized fluorescence spectra from cancer, normal and benign
human breast tissues are studied through wavelet transform and singular value decomposition. The discrete
wavelets enabled one to isolate high and low frequency spectral fluctuations, which revealed substantial randomization
in the cancerous tissues, not present in the normal cases. In particular, the fluctuations fitted well
with a Gaussian distribution for the cancerous tissues in the perpendicular component. One finds non-Gaussian
behavior for normal and benign tissues' spectral variations. The study of the difference of intensities in parallel
and perpendicular channels, which is free from the diffusive component, revealed weak fluorescence activity in
the 630nm domain, for the cancerous tissues. This may be ascribable to porphyrin emission. The role of both
scatterers and fluorophores in the observed minor intensity peak for the cancer case is experimentally confirmed
through tissue-phantom experiments. Continuous Morlet wavelet also highlighted this domain for the cancerous
tissue fluorescence spectra. Correlation in the spectral fluctuation is further studied in different tissue types
through singular value decomposition. Apart from identifying different domains of spectral activity for diseased
and non-diseased tissues, we found random matrix support for the spectral fluctuations. The small eigenvalues of
the perpendicular polarized fluorescence spectra of cancerous tissues fitted remarkably well with random matrix
prediction for Gaussian random variables, confirming our observations about spectral fluctuations in the wavelet
domain.
Fluorescence intensity fluctuations in the visible wavelength regime in normal, benign, and cancerous human breast tissue samples are studied through wavelet transform. The analyses have been carried out in unpolarized, parallel and perpendicularly polarized channels, for optimal tissue characterization. It has been observed that polarized fluorescence data, particularly the perpendicular components, differentiate various tissue types quite well. Wavelet transform, because of its ability for multiresolution analysis, provides the ideal tool to separate and characterize fluctuations in the fluorescence spectra at different scales. We quantify these differences and find that the fluctuations in the perpendicular channel of the cancerous tissues are more randomized as compared to their normal counterparts. Furthermore, for cancerous tissues, the same is very well described by the normal distribution, which is not the case for normal and benign samples. It has also been observed that, up to a certain point, fluctuations at larger scales are more sensitive to tissue types. The differences in the average, low-pass wavelet coefficients of normal, cancerous, pericanalicular, and intracanalicular benign tissues are also pointed out.
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.