Ensembles of radial basis function networks for spectroscopic detection of cervical precancer

Cervical Cancer Prediction by Merging Features of Different Colposcopic Images and Using Ensemble Classifier

Background

Cervical cancer is a significant cause of cancer mortality in women, particularly in low-income countries. In regular cervical screening methods, such as colposcopy, an image is taken from the cervix of a patient. The particular image can be used by computer-aided diagnosis (CAD) systems that are trained using artificial intelligence algorithms to predict the possibility of cervical cancer. Artificial intelligence models had been highlighted in a number of cervical cancer studies. However, there are a limited number of studies that investigate the simultaneous use of three colposcopic screening modalities including Greenlight, Hinselmann, and Schiller.

Methods

We propose a cervical cancer predictor model which incorporates the result of different classification algorithms and ensemble classifiers. Our approach merges features of different colposcopic images of a patient. The feature vector of each image includes semantic medical features, subjective judgments, and a consensus. The class label of each sample is calculated using an aggregation function on expert judgments and consensuses.

Results

We investigated different aggregation strategies to find the best formula for aggregation function and then we evaluated our method using the quality assessment of digital colposcopies dataset, and our approach performance with 96% of sensitivity and 94% of specificity values yields a significant improvement in the field.

Conclusion

Our model can be used as a supportive clinical decision-making strategy by giving more reliable information to the clinical decision makers. Our proposed model also is more applicable in cervical cancer CAD systems compared to the available methods.

Automated cervical precancerous cells screening system based on Fourier transform infrared spectroscopy features

Fourier transform infrared (FTIR) spectroscopy technique can detect the abnormality of a cervical cell that occurs before the morphological change could be observed under the light microscope as employed in conventional techniques. This paper presents developed features extraction for an automated screening system for cervical precancerous cell based on the FTIR spectroscopy as a second opinion to pathologists. The automated system generally consists of the developed features extraction and classification stages. Signal processing techniques are used in the features extraction stage. Then, discriminant analysis and principal component analysis are employed to select dominant features for the classification process. The datasets of the cervical precancerous cells obtained from the feature selection process are classified using a hybrid multilayered perceptron network. The proposed system achieved 92% accuracy.

Novel advancements in colposcopy: historical perspectives and a systematic review of future developments

OBJECTIVE

To describe novel innovations and techniques for the detection of high-grade dysplasia.

MATERIALS AND METHODS

Studies were identified through the PubMed database, spanning the last 10 years. The key words (["computerized colposcopy" or "digital colposcopy" or "spectroscopy" or "multispectral digital colposcopy" or "dynamic spectral imaging", or "electrical impedance spectroscopy" or "confocal endomicroscopy" or "confocal microscopy"or "optical coherence tomography"] and ["cervical dysplasia" or cervical precancer" or "cervix" or "cervical"]) were used. The inclusion criteria were published articles of original research referring to noncolposcopic evaluation of the cervix for the detection of cervical dysplasia. Only English-language articles from the past 10 years were included, in which the technologies were used in vivo, and sensitivities and specificities could be calculated.

RESULTS

The single author reviewed the articles for inclusion. Primary search of the database yielded 59 articles, and secondary cross-reference yielded 12 articles. Thirty-two articles met the inclusion criteria.

CONCLUSIONS

An instrument that globally assesses the cervix, such as computer-assisted colposcopy, optical spectroscopy, and dynamic spectral imaging, would provided the most comprehensive estimate of disease and is therefore best suited when treatment is preferred. Electrical impedance spectroscopy, confocal microscopy, and optical coherence tomography provide information at the cellular level to estimate histology and are therefore best suited when deferment of treatment is preferred. If a device is to eventually replace the colposcope, it will likely combine technologies to best meet the needs of the target population, and as such, no single instrument may prove to be universally appropriate. Analyses of false-positive rates, additional colposcopies and biopsies, cost, and absolute life-savings will be important when considering these technologies and are limited thus far.

Intelligent screening systems for cervical cancer

Advent of medical image digitalization leads to image processing and computer-aided diagnosis systems in numerous clinical applications. These technologies could be used to automatically diagnose patient or serve as second opinion to pathologists. This paper briefly reviews cervical screening techniques, advantages, and disadvantages. The digital data of the screening techniques are used as data for the computer screening system as replaced in the expert analysis. Four stages of the computer system are enhancement, features extraction, feature selection, and classification reviewed in detail. The computer system based on cytology data and electromagnetic spectra data achieved better accuracy than other data.

Multiple adaptive neuro-fuzzy inference system with automatic features extraction algorithm for cervical cancer recognition

To date, cancer of uterine cervix is still a leading cause of cancer-related deaths in women worldwide. The current methods (i.e., Pap smear and liquid-based cytology (LBC)) to screen for cervical cancer are time-consuming and dependent on the skill of the cytopathologist and thus are rather subjective. Therefore, this paper presents an intelligent computer vision system to assist pathologists in overcoming these problems and, consequently, produce more accurate results. The developed system consists of two stages. In the first stage, the automatic features extraction (AFE) algorithm is performed. In the second stage, a neuro-fuzzy model called multiple adaptive neuro-fuzzy inference system (MANFIS) is proposed for recognition process. The MANFIS contains a set of ANFIS models which are arranged in parallel combination to produce a model with multi-input-multioutput structure. The system is capable of classifying cervical cell image into three groups, namely, normal, low-grade squamous intraepithelial lesion (LSIL) and high-grade squamous intraepithelial lesion (HSIL). The experimental results prove the capability of the AFE algorithm to be as effective as the manual extraction by human experts, while the proposed MANFIS produces a good classification performance with 94.2% accuracy.

Optimized endoscopic autofluorescence spectroscopy for the identification of premalignant lesions in Barrett's oesophagus

OBJECTIVE

Fluorescence spectroscopy has the potential to detect early cellular changes in Barrett's oesophagus before these become visible. As the technique is based on varying concentrations of intrinsic fluorophores, each with its own optimal excitation wavelength, it is important to assess the optimal excitation wavelength(s) for identification of premalignant lesions in patients with Barrett's oesophagus.

METHODS

The endoscopic spectroscopy system used contained five (ultra)violet light sources (λexc=369-416 nm) to generate autofluorescence during routine endoscopic surveillance. Autofluorescence spectroscopy was followed by a biopsy for histological assessment and spectra correlation. Three intensity ratios (r1, r2, r3) were calculated by dividing the area, A, under the spectral curve of selected emission wavelength ranges for each spectrum generated by each excitation wavelength λexc as follows (Equation is included in full-text article.). Double intensity ratios were calculated using two excitation wavelengths.

RESULTS

Fifty-eight tissue areas from 22 patients were used for autofluorescence spectra analysis. Excitation with 395, 405 or 410 nm showed a significant (P≤0.0006) differentiation between intestinal metaplasia and grouped high-grade dysplasia/early carcinoma for intensity ratios r2 and r3. A sensitivity of 80.0% and specificity of 89.5% with an area under the ROC curve of 0.85 was achieved using 395 nm excitation and intensity ratio r3.

CONCLUSION

Double excitation showed no additional value over single excitation. The combination of 395 nm excitation and intensity ratio r3 showed optimal conditions to discriminate nondysplastic from early neoplasia in Barrett's oesophagus.

The use of optical spectroscopy for in vivo detection of cervical pre-cancer

In order to investigate the effectiveness of optical spectroscopy for in vivo diagnosis of cervical pre-cancerous conditions, a series of published studies are surveyed. The six optical technologies investigated include fluorescence spectroscopy, reflectance spectroscopy, and their combination using point probe or multispectral imaging approaches. Searching in the well-known databases, the most recent published works were sought out. Various aspects of the studies were evaluated including the details of the technology used, the pathologic threshold for tissue classification and the gold standard, the study population and prevalence of disease in this population, the method of measurement, the number of clinicians involved in the study, the classification and validation algorithms, and the performance in terms of sensitivity, specificity and, when available, the area under the receiver operating characteristic curve. Forty-four studies conducted from 1994 to 2012 were evaluated. The data are gathered in two comprehensive tables, and five illustrations are provided to simplify a comparison between studies from different points of view. There is a broad band of studies from small pilot studies through phase III clinical trials. Among the reviewed articles, only three factors were found to influence the performance of the optical spectroscopy studies. Multispectral approaches show higher specificity than the point probe approaches (p = 0.001). The use of acetic acid before measurement and prevalence of disease among the studied population, also, have an impact on the sensitivity and specificity of the studies (p < 0.05), respectively.

SSC: a classifier combination method based on signal strength

We propose a new classifier combination method, the signal strength-based combining (SSC) approach, to combine the outputs of multiple classifiers to support the decision-making process in classification tasks. As ensemble learning methods have attracted growing attention from both academia and industry recently, it is critical to understand the fundamental issues of the combining rule. Motivated by the signal strength concept, our proposed SSC algorithm can effectively integrate the individual vote from different classifiers in an ensemble learning system. Comparative studies of our method with nine major existing combining rules, namely, geometric average rule, arithmetic average rule, median value rule, majority voting rule, Borda count, max and min rule, weighted average, and weighted majority voting rules, is presented. Furthermore, we also discuss the relationship of the proposed method with respect to margin-based classifiers, including the boosting method (AdaBoost.M1 and AdaBoost.M2) and support vector machines by margin analysis. Detailed analyses of margin distribution graphs are presented to discuss the characteristics of the proposed method. Simulation results for various real-world datasets illustrate the effectiveness of the proposed method.

Spectral classifier design with ensemble classifiers and misclassification-rejection: application to elastic-scattering spectroscopy for detection of colonic neoplasia

Optical spectroscopy has shown potential as a real-time, in vivo, diagnostic tool for identifying neoplasia during endoscopy. We present the development of a diagnostic algorithm to classify elastic-scattering spectroscopy (ESS) spectra as either neoplastic or non-neoplastic. The algorithm is based on pattern recognition methods, including ensemble classifiers, in which members of the ensemble are trained on different regions of the ESS spectrum, and misclassification-rejection, where the algorithm identifies and refrains from classifying samples that are at higher risk of being misclassified. These "rejected" samples can be reexamined by simply repositioning the probe to obtain additional optical readings or ultimately by sending the polyp for histopathological assessment, as per standard practice. Prospective validation using separate training and testing sets result in a baseline performance of sensitivity = .83, specificity = .79, using the standard framework of feature extraction (principal component analysis) followed by classification (with linear support vector machines). With the developed algorithm, performance improves to Se ∼ 0.90, Sp ∼ 0.90, at a cost of rejecting 20-33% of the samples. These results are on par with a panel of expert pathologists. For colonoscopic prevention of colorectal cancer, our system could reduce biopsy risk and cost, obviate retrieval of non-neoplastic polyps, decrease procedure time, and improve assessment of cancer risk.

Synchronous fluorescence spectroscopy for the detection and characterization of cervical cancers in vitro

The objective of this study was to assess the diagnostic potential of synchronous fluorescence (SF) spectroscopy (SFS) technique for the detection and characterization of normal and different malignancy stages of moderately differentiated squamous cell carcinoma (MDSCC), poorly differentiated squamous cell carcinoma (PDSCC) cervical tissues. SF spectra were measured from 45 biopsies from 30 patients in vitro. Characteristic, highly resolved peaks and significant spectral differences between normal and MDSCC, PDSCC cervical tissues were obtained. Nine potential ratios were calculated and used as input variables for a discriminant analysis across different groups. The potentiality of the SFS technique was estimated by two discriminant analyses. Discriminant analysis I performed across normal and abnormal (including MDSCC and PDSCC) cervical tissues classified as 100% both original and the cross-validated grouped cases. In discriminant analysis II performed across the three groups, normal, MDSCC and PDSCC, 100% of both original and the cross-validated grouped cases were correctly classified. Using the SFS technique, one can obtain all the key biochemical markers such as tryptophan, collagen, hemoglobin, reduced form of nicotinamide adenine dinucleotide and flavin adenine dinucleotide in a single scan and hence they can be targeted as tumor markers in the detection of normal from abnormal cervical tissues.

Comparative study between Pap smear cytology and FTIR spectroscopy: a new tool for screening for cervical cancer

AIMS

To evaluate Fourier transform infrared (FTIR) spectroscopy as new tool for screening of cervical cancer in comparison with cervical cytology.

METHODS

A total of 800 cervical scrapings were taken by cytobrush and placed in ThinPrep medium. The samples were dried over infrared transparent matrix. Beams of infrared light were directed at the dried samples at frequency of 4000 to 400 cm(-1). The absorption data were produced using a Spectrum BX II FTIR spectrometer. Data were compared with the reference absorption data of known samples using FTIR spectroscopy software. FTIR spectroscopy was compared with cytology (gold standard).

RESULTS

FTIR spectroscopy could differentiate normal from abnormal cervical cells in the samples examined. The sensitivity was 85%, specificity 91%, positive predictive value 19.5% and negative predictive value of 99.5%.

CONCLUSION

This study suggests that FTIR spectroscopy could be used as an alternative method for screening for cervical cancer.

An automated cervical pre-cancerous diagnostic system

OBJECTIVE

This paper proposes to develop an automated diagnostic system for cervical pre-cancerous. METHODS AND DATA SAMPLES: The proposed automated diagnostic system consists of two parts; an automatic feature extraction and an intelligent diagnostic. In the automatic feature extraction, the system automatically extracts four cervical cells features (i.e. nucleus size, nucleus grey level, cytoplasm size and cytoplasm grey level). A new features extraction algorithm called region-growing-based features extraction (RGBFE) is proposed to extract the cervical cells features. The extracted features will then be fed as input data to the intelligent diagnostic part. A new artificial neural network (ANN) architecture called hierarchical hybrid multilayered perceptron (H(2)MLP) network is proposed to predict the cervical pre-cancerous stage into three classes, namely normal, low grade intra-epithelial squamous lesion (LSIL) and high grade intra-epithelial squamous lesion (HSIL). We empirically assess the capability of the proposed diagnostic system using 550 reported cases (211 normal cases, 143 LSIL cases and 196 HSIL cases).

RESULTS

For evaluation of the automatic feature extraction performance, correlation test approach was used to determine the capability of the RGBFE algorithm as compared to manual extraction by cytotechnologist. The manual extraction of size was recorded in micrometer while the automatic extraction of size was recorded in number of pixels. Region color was recorded in mean of grey level value for both manual and automatic extraction. The results show that the estimated size and mean of grey level have strong linear relationship (correlation test more than 0.8) with those extracted manually by cytotechnologist. Hence, the size of nucleus, size of cytoplasm and grey level of cytoplasm created very strong linear relationship with correlation test more than 0.95 (approaching one). For the intelligent diagnostic, the performance of the H(2)MLP network was compared with three standard ANNs (i.e. multilayered perceptron (MLP), radial basis function (RBF) and hybrid multilayered perceptron (HMLP)). The performance was done based on accuracy, sensitivity, specificity, false negative and false positive. The H(2)MLP network performed the best diagnostic performance as compared to other ANNs. It was able to achieve 97.50% accuracy, 100% specificity and 96.67% sensitivity. The false negative and false positive were 1.33% and 3.00%, respectively.

CONCLUSIONS

This project has successfully developed an automatic diagnostic system for cervical pre-cancerous. This study has also successfully proposed one image processing technique namely the RGBFE algorithm for automatic feature extraction process and a new ANN architecture namely the H(2)MLP network for better diagnostic performance.

A probability-based spectroscopic diagnostic algorithm for simultaneous discrimination of brain tumor and tumor margins from normal brain tissue

This paper reports the development of a probability-based spectroscopic diagnostic algorithm capable of simultaneously discriminating tumor core and tumor margins from normal human brain tissues. The algorithm uses a nonlinear method for feature extraction based on maximum representation and discrimination feature (MRDF) and a Bayesian method for classification based on sparse multinomial logistic regression (SMLR). Both the autofluorescence and the diffuse-reflectance spectra acquired in vivo from patients undergoing craniotomy or temporal lobectomy at the Vanderbilt University Medical Center were used to train and validate the algorithm. The classification accuracy was observed to be approximately 96%, 80%, and 97% for the tumor, tumor margin, and normal brain tissues, respectively, for the training data set and approximately 96%, 94%, and 100%, respectively, for the corresponding tissue types in an independent validation data set. The inherently multi-class nature of the algorithm facilitates a rapid and simultaneous classification of tissue spectra into various tissue categories without the need for a hierarchical multi-step binary classification scheme. Further, the probabilistic nature of the algorithm makes it possible to quantitatively assess the certainty of the classification and recheck the samples that are classified with higher relative uncertainty.

Comparison of the Performance of Linear Multivariate Analysis Methods for Normal and Dyplasia Tissues Differentiation Using Autofluorescence Spectroscopy

We compared the performance of three widely used linear multivariate methods for autofluorescence spectroscopic tissues differentiation. Principal component analysis (PCA), partial least squares (PLS), and multivariate linear regression (MVLR) were compared for differentiating at normal, tubular adenoma/epithelial dysplasia and cancer in colorectal and oral tissues. The methods' performances were evaluated by cross-validation analysis. The group-averaged predictive diagnostic accuracies were 85% (PCA), 90% (PLS), and 89% (MVLR) for colorectal tissues; 89% (PCA), 90% (PLS), and 90% (MVLR) for oral tissues. This study found that both PLS and MVLR achieved higher diagnostic results than did PCA.

Multi-class classification algorithm for optical diagnosis of oral cancer

We report development of a direct multi-class spectroscopic diagnostic algorithm for discrimination of high-grade cancerous tissue sites from low-grade as well as precancerous and normal squamous tissue sites of human oral cavity. The algorithm was developed making use of the recently formulated theory of total principal component regression (TPCR). The in vivo autofluorescence spectral data acquired from patients screened for neoplasm of oral cavity at the Government Cancer Hospital, Indore, was used to train and validate the algorithm. The diagnostic algorithm based on TPCR was found to provide satisfactory performance in classifying the tissue sites in four different classes - high-grade squamous cell carcinoma, low-grade squamous cell carcinoma, leukoplakia, and normal squamous tissue. The classification accuracy for these four classes was observed to be approximately 94%, 100%, 100% and 91% for the training data set (based on leave-one-out cross-validation), and was approximately 90%, 90%, 85% and 88%, respectively for the corresponding classes for the independent validation data set.

Laguerre-based method for analysis of time-resolved fluorescence data: application to in-vivo characterization and diagnosis of atherosclerotic lesions

We report the application of the Laguerre deconvolution technique (LDT) to the analysis of in-vivo time-resolved laser-induced fluorescence spectroscopy (TR-LIFS) data and the diagnosis of atherosclerotic plaques. TR-LIFS measurements were obtained in vivo from normal and atherosclerotic aortas (eight rabbits, 73 areas), and subsequently analyzed using LDT. Spectral and time-resolved features were used to develop four classification algorithms: linear discriminant analysis (LDA), stepwise LDA (SLDA), principal component analysis (PCA), and artificial neural network (ANN). Accurate deconvolution of TR-LIFS in-vivo measurements from normal and atherosclerotic arteries was provided by LDT. The derived Laguerre expansion coefficients reflected changes in the arterial biochemical composition, and provided a means to discriminate lesions rich in macrophages with high sensitivity (>85%) and specificity (>95%). Classification algorithms (SLDA and PCA) using a selected number of features with maximum discriminating power provided the best performance. This study demonstrates the potential of the LDT for in-vivo tissue diagnosis, and specifically for the detection of macrophages infiltration in atherosclerotic lesions, a key marker of plaque vulnerability.

N2 laser excited autofluorescence spectroscopy of formalin-fixed human breast tissue

The paper reports results of an in vitro study on autofluorescence spectroscopy of fresh and formalin-fixed human breast tissue samples to investigate the effect of formalin fixation on the measured autofluorescence spectra. It also explores the applicability of the approach in discriminating cancerous from the uninvolved sites of the formalin-fixed breast tissues based on their autofluorescence spectra. A probability-based diagnostic algorithm, making use of the theory of relevance vector machine (RVM), a powerful recent approach for statistical pattern recognition, was developed for that purpose. The algorithm provided sensitivity values of up to 97% and specificity values of up to 100% towards cancer for both the independent validation data set as well as for the training data set based on leave-one-out cross-validation. These results suggest that autofluorescence spectroscopy may prove to be a valuable additional in vitro diagnostic modality in clinical pathology setting for discriminating cancerous tissue sites from normal sites.

OPTICAL SPECTROSCOPY CHARACTERISTICS CAN DIFFERENTIATE BENIGN AND MALIGNANT RENAL TISSUES: A POTENTIALLY USEFUL MODALITY

PURPOSE

Promising results of optical signals have been reported in the literature for the diagnosis of Barrett's esophagus, oral cavity lesions, brain tumor margins, cervical intraepithelial neoplasia, skin cancer and bladder cancer. The potential usefulness of these techniques in renal tissues and neoplasms has not been described to date. This initial study examined the feasibility of using fluorescence and diffuse reflectance spectroscopy to differentiate between malignant and benign renal tissues.

MATERIALS AND METHODS

An ex vivo study was conducted to identify optical characteristics of various renal tissue types. Pathologically confirmed benign and malignant renal samples were obtained from nephrectomy specimens from patients undergoing radical nephrectomy. Fluorescence and diffuse reflectance spectra were measured from benign and malignant renal tissues.

RESULTS

All renal tissues, malignant or benign, contain 2 primary emission peaks-a strong one at approximately 285 nm excitation, approximately 340 nm emission (Peak A), and a weak one at approximately 340 nm excitation, approximately 460 nm emission (Peak B). Peak A of normal renal tissue typically locates at the shorter excitation wavelength region than that of malignant tissue. The intensity of Peak B from benign tissues tends to be greater than that from malignant renal tissues. Diffuse reflectance intensities from malignant renal tissues between 600 and 800 nm are markedly greater than those from normal renal tissue. Empirical discrimination algorithms developed based on selected fluorescence and diffuse reflectance spectral characteristics yields accurate differentiation between benign and malignant renal tissues.

CONCLUSIONS

Highly accurate differentiation between normal human renal tissues and renal cell cancers is feasible using combined fluorescence and diffuse reflectance spectroscopy in an ex vivo setting. If successful in future clinical studies, optical spectroscopy could aid in margin detection and tissue discrimination while performing nephron sparing surgery.

Relevance vector machine for optical diagnosis of cancer

BACKGROUND AND OBJECTIVES

A probability-based, robust diagnostic algorithm is an essential requirement for successful clinical use of optical spectroscopy for cancer diagnosis. This study reports the use of the theory of relevance vector machine (RVM), a recent Bayesian machine-learning framework of statistical pattern recognition, for development of a fully probabilistic algorithm for autofluorescence diagnosis of early stage cancer of human oral cavity. It also presents a comparative evaluation of the diagnostic efficacy of the RVM algorithm with that based on support vector machine (SVM) that has recently received considerable attention for this purpose.

STUDY DESIGN/MATERIALS AND METHODS

The diagnostic algorithms were developed using in vivo autofluorescence spectral data acquired from human oral cavity with a N(2) laser-based portable fluorimeter. The spectral data of both patients as well as normal volunteers, enrolled at Out Patient department of the Govt. Cancer Hospital, Indore for screening of oral cavity, were used for this purpose. The patients selected had no prior confirmed malignancy and were diagnosed of squamous cell carcinoma (SCC), Grade-I on the basis of histopathology of biopsy taken from abnormal site subsequent to acquisition of spectra. Autofluorescence spectra were recorded from a total of 171 tissue sites from 16 patients and 154 healthy squamous tissue sites from 13 normal volunteers. Of 171 tissues sites from patients, 83 were SCC and the rest were contralateral uninvolved squamous tissue. Each site was treated separately and classified via the diagnostic algorithm developed. Instead of the spectral data from uninvolved sites of patients, the data from normal volunteers were used as the normal database for the development of diagnostic algorithms.

RESULTS

The diagnostic algorithms based on RVM were found to provide classification performance comparable to the state-of-the-art SVMs, while at the same time explicitly predicting the probability of class membership. The sensitivity and specificity towards cancer were up to 88% and 95% for the training set data based on leave- one-out cross validation and up to 91% and 96% for the validation set data. When implemented on the spectral data of the uninvolved oral cavity sites from the patients, it yielded a specificity of up to 91%.

CONCLUSIONS

The Bayesian framework of RVM formulation makes it possible to predict the posterior probability of class membership in discriminating early SCC from the normal squamous tissue sites of the oral cavity in contrast to dichotomous classification provided by the non-Bayesian SVM. Such classification is very helpful in handling asymmetric misclassification costs like assigning different weights for having a false negative result for identifying cancer compared to false positive. The results further demonstrate that for comparable diagnostic performances, the RVM-based algorithms use significantly fewer kernel functions and do not need to estimate any hoc parameters associated with the learning or the optimization technique to be used. This implies a considerable saving in memory and computation in a practical implementation.

Support vector machine for optical diagnosis of cancer

We report the application of a support vector machine (SVM) for the development of diagnostic algorithms for optical diagnosis of cancer. Both linear and nonlinear SVMs have been investigated for this purpose. We develop a methodology that makes use of SVM for both feature extraction and classification jointly by integrating the newly developed recursive feature elimination (RFE) in the framework of SVM. This leads to significantly improved classification results compared to those obtained when an independent feature extractor such as principal component analysis (PCA) is used. The integrated SVM-RFE approach is also found to outperform the classification results yielded by traditional Fisher's linear discriminant (FLD)-based algorithms. All the algorithms are developed using spectral data acquired in a clinical in vivo laser-induced fluorescence (LIF) spectroscopic study conducted on patients being screened for cancer of the oral cavity and normal volunteers. The best sensitivity and specificity values provided by the nonlinear SVM-RFE algorithm over the data sets investigated are 95 and 96% toward cancer for the training set data based on leave-one-out cross validation and 93 and 97% toward cancer for the independent validation set data. When tested on the spectral data of the uninvolved oral cavity sites from the patients it yielded a specificity of 85%.

Optical detection of high-grade cervical intraepithelial neoplasia in vivo: results of a 604-patient study

OBJECTIVE

The purpose of this study was to assess the in vivo optical detection of high-grade cervical intraepithelial neoplasia (2/3+) on the whole cervix with a noncontact, spectroscopic device.

STUDY DESIGN

Cervical scanning devices collected intrinsic fluorescence and broadband white light spectra and video images from 604 women during routine colposcopy examinations at 6 clinical centers. A statistically significant dataset was developed of intrinsic fluorescence and white light-induced cervical tissue spectra that was correlated to expert histopathologic determination. On the basis of a retrospective analysis of the acquired data, a classification algorithm was developed, validated, and optimized.

RESULTS

Intrinsic fluorescence, backscattered white light, and video imaging each contribute complementary information to diagnostic algorithms for high-grade cervical neoplasia. More than 10000 measurements that were made on colposcopically identified tissue from >500 subjects were the basis for algorithm training and testing. Algorithm performance demonstrated a sensitivity of approximately 90%. This performance was confirmed by various training methods. With the use of a multivariate classification algorithm, optical detection is predicted to detect 33% more high-grade cervical intraepithelial neoplasia (2/3+) than colposcopy alone.

CONCLUSION

Full cervix optical interrogation for the detection of high-grade cervical intraepithelial neoplasia is feasible and appears capable of detecting more high-grade cervical intraepithelial neoplasia than colposcopy alone. With the use of this classification algorithm, a multisite, randomized controlled trial is underway that compares the combination of optical detection and colposcopy versus colposcopy alone.

Classification of in vivo autofluorescence spectra using support vector machines

An algorithm based on support vector machines (SVM), the most recent advance in pattern recognition, is presented for use in classifying light-induced autofluorescence collected from cancerous and normal tissues. The in vivo autofluorescence spectra used for development and evaluation of SVM diagnostic algorithms were measured from 85 nasopharyngeal carcinoma (NPC) lesions and 131 normal tissue sites from 59 subjects during routine nasal endoscopy. Leave-one-out cross-validation was used to evaluate the performance of the algorithms. An overall diagnostic accuracy of 96%, a sensitivity of 94%, and a specificity of 97% for discriminating nasopharyngeal carcinomas from normal tissues were achieved using a linear SVM algorithm. A diagnostic accuracy of 98%, a sensitivity of 95%, and a specificity of 99% for detecting NPC were achieved with a nonlinear SVM algorithm. In a comparison with previously developed algorithms using the same dataset and the principal component analysis (PCA) technique, the SVM algorithms produced better diagnostic accuracy in all instances. In addition, we investigated a method combining PCA and SVM techniques for reducing the complexity of the SVM algorithms.

Nonlinear pattern recognition for laser-induced fluorescence diagnosis of cancer

BACKGROUND AND OBJECTIVES

Use of laser induced fluorescence (LIF) spectroscopy for the diagnosis of cancer requires an appropriate diagnostic algorithm for spectral pattern recognition. While most of the diagnostic algorithms reported in the literature use standard linear feature extraction techniques like principal component analysis (PCA), partial least square (PLS) analysis etc., use of nonlinear techniques is expected to provide improved discrimination. We report here the performance of an algorithm based on nonlinear Maximum Representation and Discrimination Feature (MRDF) method for diagnosis of early stage cancer of human oral cavity. The diagnostic efficacy of the algorithm has been compared with a linear PCA based algorithm.

STUDY DESIGN/MATERIALS AND METHODS

The diagnostic algorithms were developed based on spectral data acquired in an in-vivo LIF study, at the outpatient department (OPD) of the Government Cancer Hospital, Indore, involving 16 patients with cancer of oral cavity and 13 normal volunteers with healthy oral cavity. In-vivo autofluorescence spectra were recorded using a N(2) laser based portable fluorimeter. The patients had no prior confirmed malignancy, were suspected on visual examination of having early cancer of the oral cavity and were diagnosed of squamous cell carcinoma (SCC) on the basis of histopathology of biopsy taken from abnormal site subsequent to acquisition of spectra. The spectra were acquired from a total of 171 tissue sites from patients, of which 83 were from SCC and 88 were from uninvolved squamous tissue, and 154 sites from healthy squamous tissue from normal volunteers. In each patient, the normal tissue sites interrogated were from the adjacent apparently uninvolved region of the oral cavity. Each site was treated separately and classified via the diagnostic algorithm developed. Instead of the spectral data from uninvolved sites of patients, the data from normal volunteers were used as the normal database for the development of diagnostic algorithms.

RESULTS

The nonlinear diagnostic algorithm based on MRDF provided a sensitivity of 93% and a specificity of 96% towards cancer for the training set data and a sensitivity of 95% and a specificity of 96% towards cancer for the validation set data. When implemented on the spectral data of the uninvolved oral cavity sites from the patients it yielded a specificity of 96%. On the other hand, the linear PCA based algorithm provided a sensitivity of 83% and a specificity of 66% towards cancer for the training set data and a sensitivity of 80% and a specificity of 58% towards cancer for the validation set data. When spectral data of the uninvolved oral cavity sites from the patients were considered as the unknown data set, it resulted in a specificity value of 56%.

CONCLUSIONS

The nonlinear MRDF algorithm provided significantly improved diagnostic performance as compared to the linear PCA based algorithm in discriminating the cancerous tissue sites of the oral cancer patients from the healthy squamous tissue sites of normal volunteers as well as the uninvolved tissue sites of the oral cavity of the patients with cancer.

Neural network-based assessment of prognostic markers and outcome prediction in bilharziasis-associated bladder cancer

In this paper the potential value of two prognostic factors, namely, bilharziasis status and tumor histological type, is investigated in relation to their abilities to predict disease progression and outcome of patients with bladder cancer, using radial basis function (RBF) neural networks. The bladder cancer data set is described by eight clinical and pathological markers. Two outcomes are of interest: either a patient is alive and free of disease or the patient is dead within five years of diagnosis. Three hundred and twenty-one (321) patients are involved in this retrospective study, 83.5% of whom had been confirmed with bilharziasis history. Selected marker subsets are examined to improve the outcome predictive accuracy and to evaluate the effects of the assessed prognostic factors on such outcome. The highest predictive accuracy for patients with bladder adenocarcinoma, as obtained from the RBF network, is found to be 85% with one subset of markers. The predictive analysis shows that bilharziasis history and patients' histology type are both important prognostic factors in prediction and, for each histology type, different marker combinations with significant characteristics have been observed.

Fiber optic probes for biomedical optical spectroscopy

Fiber optic probes are a key element for biomedical spectroscopic sensing. We review the use of fiber optic probes for optical spectroscopy, focusing on applications in turbid media, such as tissue. The design of probes for reflectance, polarized reflectance, fluorescence, and Raman spectroscopy is illustrated. We cover universal design principles as well as technologies for beam deflecting and reshaping.

Detecting the signal of the menstrual cycle in fluorescence spectroscopy of the cervix

Fluorescence spectroscopy of the cervix has been shown to be an effective noninvasive diagnostic tool for cervical intraepithelial neoplasia (precancer). To assess the effect of the menstrual cycle on fluorescence spectroscopy, daily measurements were made on ten subjects for the length of their cycle. These measurements were analyzed to determine if there was a statistically significant signal associated with the menstrual cycle. A signal was found for emission wavelengths between 425 and 445 nm inclusive--near the main hemoglobin absorption band, the Soret band, at 420 nm. We suspect that the slight displacement of the Soret band is due to the nearby dominant NAD(P)H peak, which increases the signal-to-noise ratio and affects statistical significance. The signal consists of a reduction in fluorescence intensity for the first few days of the cycle. This analysis indicates that hemoglobin absorption is the main menstrual-cycle effect on the use of fluorescence spectroscopy on the cervix. The effect is confined to a small set of excitation/emission wavelengths and to approximately the first 8 days of the cycle. This suggests that any problems from the menstrual cycle can be avoided with a simple requirement that the device not be used during the period of menstrual bleeding.

Organotypic Raft Cultures as an Effective In Vitro Tool for Understanding Raman Spectral Analysis of Tissue¶

There is a growing body of evidence showing that optical spectroscopy has the potential to be a useful in vivo diagnostic tool. Yet, so far there is no definitive cellular and biochemical understanding for the differences seen in the spectra from different tissue categories and disease states. In this study, we examine the use of organotypic raft cultures as an in vitro model of in vivo tissue conditions in an attempt to overcome some of the limitations of previously used methods. Organotypic raft cultures resembling normal and dysplastic epithelial cervical tissue were constructed and grown at an air-liquid interface for 2 weeks. Raman spectra of normal as well as dysplastic raft cultures were measured and compared with in vivo spectra from the corresponding tissue type. Histologic comparisons ensured that the raft cultures had similar structure and morphology to the corresponding intact tissue types. Raman spectra were also acquired from different layers of tissue. Spectral comparisons show that the Raman spectra of the raft cultures are similar to the spectra acquired from the cervix in vivo for both normal and dysplastic tissues. These results show that organotypic raft cultures are an effective and useful tool for the cellular and biochemical analysis of tissue spectroscopy.

Fluorescence spectroscopy for cervical precancer detection: Is there variance across the menstrual cycle?

This study assesses one possible cause of inter-patient variation in fluorescence spectroscopy of the cervix: the menstrual cycle. Ten patients with no history of an abnormal Pap smear were seen daily throughout 30 consecutive days of their cycle. Fluorescence excitation-emission matrices were measured from three cervical sites on each patient. Principal component analysis was used to determine which spectral regions varied with the day of the cycle. Classification was performed to assess the influence of menstrual cycle on precancer diagnosis. Variations in the principal component scores and the redox ratio values show that the fluorescence emission spectra at 340-380 nm excitation appear to correlate with the cell metabolism of the cervical epithelium throughout the menstrual cycle; these changes do not affect diagnostic classification. The menstrual cycle affects intra-patient variation but does not appear to cause a significant level of inter-patient variation. It does not need to be controlled for in optical detection strategies based on fluorescence spectroscopy.

Reflectance spectroscopy for in vivo detection of cervical precancer

Optical technologies, in particular fluorescence spectroscopy, have shown the potential to provide improved detection methods for cervical neoplasia that are sensitive and cost effective through accurate, objective, instantaneous point-of-care diagnostic tools. The specific goals of this study were to analyze reflectance spectra of normal and neoplastic cervical tissue in vivo and to evaluate the data for use in diagnostic algorithm development. Spectroscopic measurements were obtained at four distinct source-detector separations from 324 sites in 161 patients. As the source-detector separation increases, greater tissue depth is probed. The average spectra of each diagnostic class differed at all source-detector separations, with the greatest differences occurring at the smallest source-detector separations. Algorithms, based on principal-component analysis and Mahalanobis distance classification, were developed and evaluated for all combinations of source-detector separations relative to the gold standard of colposcopically directed biopsy. The diagnostic combination of squamous normal versus high-grade squamous intraepithelial lesions gave good discrimination with a sensitivity of 72% and a specificity of 81%; discrimination of columnar normal versus high-grade squamous intraepithelial lesions also was good, with sensitivity of 72% and specificity of 83%. Thus, reflectance spectroscopy appears promising for in vivo detection of cervical precancer. Strategies that combine fluorescence and reflectance spectroscopy may enhance the discrimination capabilities.

Skin lesion classification using oblique-incidence diffuse reflectance spectroscopic imaging

We discuss the use of a noninvasive in vivo optical technique, diffuse reflectance spectroscopic imaging with oblique incidence, to distinguish between benign and cancer-prone skin lesions. Various image features were examined to classify the images from lesions into benign and cancerous categories. Two groups of lesions were processed separately: Group 1 includes keratoses, warts versus carcinomas; and group 2 includes common nevi versus dysplastic nevi. A region search algorithm was developed to extract both one- and two-dimensional spectral information. A bootstrap-based Bayes classifier was used for classification. A computer-assisted tool was then devised to act as an electronic second opinion to the dermatologist. Our approach generated only one false-positive misclassification out of 23 cases collected for group 1 and two misclassifications out of 34 cases collected for group 2 under the worst estimation condition.

Preliminary study of in vivo autofluorescence of nasopharyngeal carcinoma and normal tissue

BACKGROUND AND OBJECTIVE

In nasopharyngeal cancer, conventional white light endoscopy does not provide adequate information to detect the flat/small lesion and identify the margin of observable tumor. In the present study, we evaluate the potential of light-induced fluorescence spectroscopic imaging for the localization of cancerous nasopharyngeal tissue.

STUDY DESIGN/MATERIALS AND METHODS

We built a multiple channel spectrometer specifically for the investigation of fluorescence collected by a conventional endoscopic system. Nasopharyngeal fluorescence were measured in vivo from 27 subjects during the routine endoscopy. The biopsy specimens for histologic analysis were taken from the tissue sites where the fluorescence were measured.

RESULTS

Two algorithms to discriminate the nasopharyngeal carcinoma from normal tissue were created based on the good correlation between the tissue autofluorescence and histologic diagnosis. For the two-wavelength algorithm, carcinoma can be differentiated from normal tissue with a sensitivity and specificity of 93% and 92%, respectively. For the three-wavelength algorithm with compensation of variation of blood content in tissue, a sensitivity of 98% and specificity of 95% were achieved.

CONCLUSION

Fluorescence endoscopic imaging used with the algorithms developed in this report is an efficient method for detecting the nasopharyngeal carcinoma.

Excitation-and-collection geometry insensitive fluorescence imaging of tissue-simulating turbid media

We present an imaging technique for the correction of geometrical effects in fluorescence measurement of optically thick, turbid media such as human tissue. Specifically, we use the cross-polarization method to reject specular reflection and enhance the diffusive backscattering of polarized fluorescence excitation light from the turbid media. We correct the nonuniformity of the image field caused by the excitation-and-collection geometry of a fluorescence imaging system by normalizing the fluorescence image to the cross-polarized reflection image. The ratio image provides a map of relative fluorescence yield, defined as the ratio of emerging fluorescence power to incident excitation, over the surface of an imaged homogeneous turbid medium when fluorescence excitation-and-collection geometries vary in a wide range. We investigate the mechanism of ratio imaging by using Monte Carlo modeling. Our findings show that this technique could have a potential use in the detection of early cancer, which usually starts from a superficial layer of tissue, based on the contrast in the tissue fluorescence of an early lesion and of the surrounding normal tissue.

Staging of cervical cancer with soft computing

This paper describes a way of designing a hybrid decision support system in soft computing paradigm for detecting the different stages of cervical cancer. Hybridization includes the evolution of knowledge-based subnetwork modules with genetic algorithms (GA's) using rough set theory and the Interactive Dichotomizer 3 (ID3) algorithm. Crude subnetworks obtained via rough set theory and the ID3 algorithm are evolved using GA's. The evolution uses a restricted mutation operator which utilizes the knowledge of the modular structure, already generated, for faster convergence. The GA tunes the network weights and structure simultaneously. The aforesaid integration enhances the performance in terms of classification score, network size and training time, as compared to the conventional multilayer perceptron. This methodology also helps in imposing a structure on the weights, which results in a network more suitable for extraction of logical rules and human interpretation of the inferencing procedure.

Fluorescence spectroscopy of neoplastic and non-neoplastic tissues

Fast and non-invasive, diagnostic techniques based on fluorescence spectroscopy have the potential to link the biochemical and morphologic properties of tissues to individual patient care. One of the most widely explored applications of fluorescence spectroscopy is the detection of endoscopically invisible, early neoplastic growth in epithelial tissue sites. Currently, there are no effective diagnostic techniques for these early tissue transformations. If fluorescence spectroscopy can be applied successfully as a diagnostic technique in this clinical context, it may increase the potential for curative treatment, and thus, reduce complications and health care costs. Steady-state, fluorescence measurements from small tissue regions as well as relatively large tissue fields have been performed. To a much lesser extent, time-resolved, fluorescence measurements have also been explored for tissue characterization. Furthermore, sources of both intrinsic (endogenous fluorophores) and extrinsic fluorescence (exogenous fluorophores) have been considered. The goal of the current report is to provide a comprehensive review on steady-state and time-resolved, fluorescence measurements of neoplastic and non-neoplastic, biologic systems of varying degrees of complexity. First, the principles and methodology of fluorescence spectroscopy are discussed. Next, the endogenous fluorescence properties of cells, frozen tissue sections and excised and intact bulk tissues are presented; fluorescence measurements from both animal and human tissue models are discussed. This is concluded with future perspectives.

Ensembles of radial basis function networks for spectroscopic detection of cervical precancer

The mortality related to cervical cancer can be substantially reduced through early detection and treatment. However, current detection techniques, such as Pap smear and colposcopy, fail to achieve a concurrently high sensitivity and specificity. In vivo fluorescence spectroscopy is a technique which quickly, noninvasively and quantitatively probes the biochemical and morphological changes that occur in precancerous tissue. A multivariate statistical algorithm was used to extract clinically useful information from tissue spectra acquired from 361 cervical sites from 95 patients at 337-, 380-, and 460-nm excitation wavelengths. The multivariate statistical analysis was also employed to reduce the number of fluorescence excitation-emission wavelength pairs required to discriminate healthy tissue samples from precancerous tissue samples. The use of connectionist methods such as multilayered perceptrons, radial basis function (RBF) networks, and ensembles of such networks was investigated. RBF ensemble algorithms based on fluorescence spectra potentially provide automated and near real-time implementation of precancer detection in the hands of nonexperts. The results are more reliable, direct, and accurate than those achieved by either human experts or multivariate statistical algorithms.