Sunanda Mitra

Adaptive fuzzy leader clustering of complex data sets in pattern recognition

A modular, unsupervised neural network architecture that can be used for clustering and classification of complex data sets is presented. The adaptive fuzzy leader clustering (AFLC) architecture is a hybrid neural-fuzzy system that learns online in a stable and efficient manner. The system used a control structure similar to that found in the adaptive resonance theory (ART-1) network to identify the cluster centers initially. The initial classification of an input takes place in a two-stage process: a simple competitive stage and a distance metric comparison stage. The cluster prototypes are then incrementally updated by relocating the centroid position from fuzzy C-means (FCM) system equations for the centroids and the membership values. The operational characteristics of AFLC and the critical parameters involved in its operation are discussed. The AFLC algorithm is applied to the Anderson iris data and laser-luminescent finger image data. The AFLC algorithm successfully classifies features extracted from real data, discrete or continuous, indicating the potential strength of this new clustering algorithm in analyzing complex data sets.

Digital stereo image analyzer for generating automated 3-D measures of optic disc deformation in glaucoma

The major limitations of precise evaluation of retinal structures in present clinical situations are the lack of standardization, the inherent subjectivity involved in the interpretation of retinal images, and intra- as well as interobserver variability. While evaluating optic disc deformation in glaucoma, these limitations could be overcome by using advanced digital image analysis techniques to generate precise metrics; from stereo optic disc image pairs. A digital stereovision system for visualizing the topography of the optic nerve head from stereo optic disc images is presented. We have developed an algorithm, combining power cepstrum and zero-mean-normalized cross correlation techniques, which extracts depth information using coarse-to-fine disparity between corresponding windows in a stereo pair. The gray level encoded sparse disparity matrix is subjected to a cubic B-spline operation to generate smooth representations of the optic cup/disc surfaces and new three-dimensional (3-D) metrics from isodisparity contours. Despite the challenges involved in 3-D surface recovery, the robustness of our algorithm in finding disparities within the constraints used has been validated using stereo pairs with known disparities. In a preliminary longitudinal study of glaucoma patients, a strong correlation is found between the computer-generated quantitative cup/disc volume metrics and manual metrics commonly used in a clinic. The computer generated new metrics, however, eliminate the subjective variability and greatly reduce the time and cost involved in manual metric generation in follow-up studies of glaucoma.

An adaptive integrated fuzzy clustering model for pattern recognition

Abstract The extension of neural net based crisp clustering algorithms to fuzzy clustering algorithms has been addressed by many researchers in recent years. However, such neuro-fuzzy clustering algorithms developed so far suffer from restrictions in identifying the actual decision boundaries among clusters with overlapping regions. These restrictions are induced by the choice of the similarity measure and representation of the clusters. An integrated adaptive fuzzy clustering (IAFC) algorithm is presented to generate improved decision boundaries by introducing a new similarity measure and by integrating the advantages of the fuzzy optimization constraint of fuzzy c-means (FCM), the control structure of adaptive resonance theory (ART-1), and a fuzzified Kohonen-type learning rule. The effect of the new similarity measure in finding nonlinear decision boundaries among closely located cluster centroids is demonstrated with computer generated data. We use the IRIS data set and a subset of the tethered satellite system simulation data set to compare the convergence rate and misclassifications resulting from IAFC algorithm with other clustering algorithms.

Power cepstrum and spectrum techniques applied to image registration

The use of power cepstrum analysis in image registration is explored. Rotational shifts and translational shifts are corrected separately. The technique involves two main ideas. First, after preprocessing to remove extraneous information and information which could result in false registration parameters, a rotational shift is changed into a translational shift by using the shift-invariant property of the power spectrum. Second, power cepstrum analysis is used to correct the translational shift. Because of the introduction of these ideas, this new algorithm can work very fast and accurately compared to conventional correlation techniques. This registration technique is applied to sequential fundus images with potential application in detecting changes in fundus anomalies.

Customized Hough transform for robust segmentation of cervical vertebrae from X-ray images

This paper addresses the issues involved in developing a robust segmentation technique capable of finding the location and orientation of the cervical vertebrae in X-ray images. This technique should be invariant to rotation, scale, noise, occlusions and shape variability. A customized approach, based on the generalized Hough transform (GHT), that captures shape variability and exploits shape information embedded in the accumulator structure to overcome noise and occlusions is proposed. This approach effectively finds estimates of the location and orientation of the cervical vertebrae boundaries in digitized X-ray images.

Fuzzy and possibilistic clustering methods for computer vision

Uncertainty abounds in all aspects of computer vision. As a result, methods which explicitly model and manage that uncertainty have a better chance in producing meaningful results in such complex situations. Fuzzy set theory provides a framework to initially model the uncertainty in a vision problem, and many methods to exploit it in producing realistic results. One very important tool which has been used extensively in pattern recognition and computer vision is that of objective function based clustering. In this chapter, we will review classical and novel clustering methods as they apply to computer vision and show examples of their utility. In particular, we will focus on the use of clustering to detect and recognize regular boundaries of objects from images of edge magnitudes. The problem of fitting an unknown number of boundary curves to edge magnitude data is one of the major challenges of computer vision. We will show that fuzzy clustering can be readily adapted to the problem of curve detection, and that a new possibilistic clustering method introduced by the authors can produce significantly more accurate results than either crisp or fuzzy clustering in noisy situations.

High fidelity adaptive vector quantization at very low bit rates for progressive transmission of radiographic images

An adaptive vector quantizer (VQ) using a clustering technique known as adaptive fuzzy leader clustering (AFLC) that is similar in concept to deterministic annealing (DA) for VQ codebook design has been developed. This vector quantizer, AFLC-VQ, has been designed to vector quantize wavelet decomposed sub images with optimal bit allocation. The high-resolution sub images at each level have been statistically analyzed to conform to generalized Gaussian probability distributions by selecting the optimal number of filter taps. The adaptive characteristics of AFLC-VQ result from AFLC, an algorithm that uses self-organizing neural networks with fuzzy membership values of the input samples for upgrading the cluster centroids based on well known optimization criteria. By gen- erating codebooks containing codewords of varying bits, AFLC-VQ is capable of compressing large color/monochrome medical images at extremely low bit rates (0.1 bpp and less) and yet yielding high fidelity reconstructed images. The quality of the reconstructed im- ages formed by AFLC-VQ has been compared with JPEG and EZW, the standard and the well known wavelet based compression tech- nique (using scalar quantization), respectively, in terms of statistical performance criteria as well as visual perception. AFLC-VQ exhibits much better performance than the above techniques. JPEG and EZW were chosen as comparative benchmarks since these have been used in radiographic image compression. The superior perfor- mance of AFLC-VQ over LBG-VQ has been reported in earlier pa- pers.

A Probabilistic Approach to Segmentation and Classification of Neoplasia in Uterine Cervix Images Using Color and Geometric Features

Automated segmentation and classification of diagnostic markers in medical imagery are challenging tasks. Numerous algorithms for segmentation and classification based on statistical approaches of varying complexity are found in the literature. However, the design of an efficient and automated algorithm for precise classification of desired diagnostic markers is extremely image-specific. The National Library of Medicine (NLM), in collaboration with the National Cancer Institute (NCI), is creating an archive of 60,000 digitized color images of the uterine cervix. NLM is developing tools for the analysis and dissemination of these images over the Web for the study of visual features correlated with precancerous neoplasia and cancer. To enable indexing of images of the cervix, it is essential to develop algorithms for the segmentation of regions of interest, such as acetowhitened regions, and automatic identification and classification of regions exhibiting mosaicism and punctation. Success of such algorithms depends, primarily, on the selection of relevant features representing the region of interest. We present color and geometric features based statistical classification and segmentation algorithms yielding excellent identification of the regions of interest. The distinct classification of the mosaic regions from the non-mosaic ones has been obtained by clustering multiple geometric and color features of the segmented sections using various morphological and statistical approaches. Such automated classification methodologies will facilitate content-based image retrieval from the digital archive of uterine cervix and have the potential of developing an image based screening tool for cervical cancer.

A multispectral digital Cervigram analyzer in the wavelet domain for early detection of cervical cancer

The significance and need for expert interpretation of cervigrams (images of the cervix) in the study of the uterine cervix changes and pre-neoplasic lesions preceding cervical cancer are being investigated. The National Cancer Institute has collected a unique dataset taken from patients with normal cervixes and at various stages of cervical pre-cancer and cancer. This dataset allows us the opportunity for studying the uterine cervix changes for validating the potential of automated classification and recognition algorithms in discriminating cervical neoplasia and normal tissue. Pilot studies have been designed (1) to evaluate the effect of image transformation and optimal color mapping on the accepted levels of compression needed for effective dissemination of cervical image data over a network and (2) for automated detection of lesions from feature extraction, registration, and segmentation of lesions in cervix image sequences. In this paper, we present the results of the effectiveness of a novel, wavelet based, multi-spectral analyzer in retaining diagnostic features in encoded cervical images, thus allowing investigation on the potential of automated detection of lesions in cervix image sequences using automated registration, color transformation and bit-rate control, and a statistical segmentation approach.

Toward the Improvement of Image-Guided Interventions for Minimally Invasive Surgery: Three Factors That Affect Performance

Objectives: The objectives were to measure the impact of specific features of imaging devices on tasks relevant to minimally invasive surgery (MIS) and to investigate cognitive and perceptual factors in such tasks. Background: Although image-guided interventions used in MIS provide benefits for patients, they pose drawbacks for surgeons, including degraded depth perception and reduced field of view (FOV). It is important to identify design factors that affect performance. Method: In two navigation experiments, observers fed a borescope through an object until it reached a target. Task completion time and object shape judgments were measured. In a motion perception experiment, observers reported the direction of a line that moved behind an aperture. A motion illusion associated with reduced FOV was measured. Results: Navigation through an object was faster when a preview of the objects exterior was provided. Judgments about the objects shape were more accurate with a preview (compared with none) and with active viewing (compared with passive viewing). The motion illusion decreased with a rectangular or rotating octagonal viewing aperture (compared with circular). Conclusions: Navigation performance may be enhanced when surgeons develop a mental model of the surgical environment, when surgeons (rather than assistants) control the camera, and when the shape of the image is designed to reduce visual illusions. Application: Unintentional contact between surgical tools and healthy tissues may be reduced during MIS when (a) visual aids permit surgeons to maintain a mental model of the surgical environment, (b) images are bound by noncircular apertures, and (c) surgeons manually control the camera.