Hugo Guterman

Diagnostic potential of Fourier-transform infrared microspectroscopy and advanced computational methods in colon cancer patients

Colon cancer is the third leading class of cancer causing increased mortality in developed countries. A polyp is one type of lesion observed in a majority of colon cancer patients. Here, we report a microscopic Fourier transform infrared (FTIR) study of normal, adenomatous polyp and malignant cells from biopsies of 24 patients. The goal of our study was to differentiate an adenomatous polyp from a malignant cell using FTIR microspectroscopy and artificial neural network (ANN) analysis. FTIR spectra and biological markers such as phosphate, RNA/DNA derived from spectra, were useful in identifying normal cells from abnormal ones that consisted of adenomatous polyp and malignant cells. However, the biological markers failed to differentiate between adenomatous polyp and malignant cases. By employing a combination of wavelet features and an ANN based classifier, we were able to classify the different cells as normal, adenomatous polyp and cancerous in a given tissue sample. The percentage of success of classification was 89%, 81%, and 83% for normal, adenomatous polyp, and malignant cells, respectively. A comparison of the method proposed with the pathological method is also discussed.

Possible common biomarkers from FTIR microspectroscopy of cervical cancer and melanoma

Detection of malignancy at early stages is crucial in cancer prevention and management. Fourier transform infrared (FTIR) spectroscopy has shown promise as a non‐invasive method with diagnostic potential in cancer detection. Studies were conducted with formalin‐fixed biopsies of melanoma and cervical cancer by FTIR microspectroscopy (FTIR‐MSP) to detect common biomarkers, which occurred in both types of cancer distinguishing them from the respective non‐malignant tissues. Both types of cancer are diagnosed on skin surfaces. The spectra were analysed for changes in levels of biomolecules such as RNA, DNA, phosphates and carbohydrate (glycogen). Whereas carbohydrate levels showed a good diagnostic potential for detection of cervical cancer, this was not the case for melanoma. However, variation of the RNA/DNA ratio as measured at I(1121)/I(1020) showed similar trends between non‐malignant and malignant tissues in both types of cancer. The ratio was higher for malignant tissues in both types of cancer.

PHYSIOLOGICAL CHARACTERISTICS OF SPIRULINA PLATENSIS (CYANOBACTERIA) CULTURED AT ULTRAHIGH CELL DENSITIES1

Photosynthetic activity and growth physiology of Spirulina platensis (Nordstedt) Geitler cultures maintained at ultrahigh cell densities (i.e. above 100 mg chlorophyll‐L−1) in a newly designed photobioreactor were investigated. Nitrogen (NaNO3) in standard Zarouk medium was characterized as a major nutrient‐limiting factor in such cultures. The effect of ultrahigh cell density on photoinhibition of photosynthesis, as reflected by chlorophyll fluorescence and photosynthetic oxygen evolution, was studied: elevating the population density may arrest photoinhibition induced by high photon flux density, as well as low temperature. The relationship between incident irradiance and oxygen production rate was linear in situ for cultures at the optimal cell density, indicating that light limitation rather than light saturation or photoinhibition is the dominant condition outdoors in cultures of ultrahigh cell densities.

Unsupervised speaker recognition based on competition between self-organizing maps

We present a method for clustering the speakers from unlabeled and unsegmented conversation (with known number of speakers), when no a priori knowledge about the identity of the participants is given. Each speaker was modeled by a self-organizing map (SOM). The SOMs were randomly initiated. An iterative algorithm allows the data move from one model to another and adjust the SOMs. The restriction that the data can move only in small groups but not by moving each and every feature vector separately force the SOMs to adjust to speakers (instead of phonemes or other vocal events). This method was applied to high-quality conversations with two to five participants and to two-speaker telephone-quality conversations. The results for two (both high- and telephone-quality) and three speakers were over 80% correct segmentation. The problem becomes even harder when the number of participants is also unknown. Based on the iterative clustering algorithm a validity criterion was also developed to estimate the number of speakers. In 16 out of 17 conversations of high-quality conversations between two and three participants, the estimation of the number of the participants was correct. In telephone-quality the results were poorer.

A Probabilistic Approach to Spectral Graph Matching

Spectral Matching (SM) is a computationally efficient approach to approximate the solution of pairwise matching problems that are np-hard. In this paper, we present a probabilistic interpretation of spectral matching schemes and derive a novel Probabilistic Matching (PM) scheme that is shown to outperform previous approaches. We show that spectral matching can be interpreted as a Maximum Likelihood (ML) estimate of the assignment probabilities and that the Graduated Assignment (GA) algorithm can be cast as a Maximum a Posteriori (MAP) estimator. Based on this analysis, we derive a ranking scheme for spectral matchings based on their reliability, and propose a novel iterative probabilistic matching algorithm that relaxes some of the implicit assumptions used in prior works. We experimentally show our approaches to outperform previous schemes when applied to exhaustive synthetic tests as well as the analysis of real image sequences.

Distinction of cervical cancer biopsies by use of infrared microspectroscopy and probabilistic neural networks

Fourier-transform infrared spectroscopy has shown alterations of spectral characteristics of cells and tissues as a result of carcinogenesis. The research reported here focuses on the diagnosis of cancer in formalin-fixed biopsied tissue for which immunochemistry is not possible and when PAP-smear results are to be confirmed. The data from two groups of patients (a control group and a group of patients diagnosed with cervical cancer) were analyzed. It was found that the glucose/phosphate ratio decreases (by 23-49%) and the RNA/DNA ratio increases (by 38-150%) in carcinogenic compared with normal tissue. Fourier-transform microspectroscopy was used to examine these tissues. This type of study in larger populations may help to set standards or classes with which to use treated biopsied tissue to predict the possibility of cancer. Probabilistic neural networks and statistical tests as parts of these biopsies predict the possibility of cancer with a high degree of accuracy (> 95%).

Optimization of feeding profile for a fed-batch bioreactor by an evolutionary algorithm.

The optimal feeding profile of a fed batch process was designed by means of an evolutionary algorithm. The algorithm chromosomes include the real-valued parameters of a profile function, defined by previous knowledge. Each chromosome is composed of the parameters that define the feeding profile: the feed rates, the singular arc parameters and the switching times between the profile states. The feed profile design was tested on a fed-batch process simulation. The accepted profiles were smooth and similar to those derived analytically in other studies. Two selection functions, roulette wheel and geometric ranking, were compared. In order to overcome the problem of model mismatches, a novel optimization scheme was carried out. During its operation the process was sampled, the model was updated and the optimization procedure was applied. The on-line optimization showed improvement in the objective function for relatively low sample times. Choosing the sampling frequencies depends on the process dynamics and the time required for the measurements and optimization. Further study on experiments of fed-batch process demonstrated the use of complex, non-differentiable model and produced improved process performances using the optimal feeding profile.

Improving Shape Retrieval by Spectral Matching and Meta Similarity

We propose two computational approaches for improving the retrieval of planar shapes. First, we suggest a geometrically motivated quadratic similarity measure, that is optimized by way of spectral relaxation of a quadratic assignment. By utilizing state-of-the-art shape descriptors and a pairwise serialization constraint, we derive a formulation that is resilient to boundary noise, articulations and nonrigid deformations. This allows both shape matching and retrieval. We also introduce a shape meta-similarity measure that agglomerates pairwise shape similarities and improves the retrieval accuracy. When applied to the MPEG-7 shape dataset in conjunction with the proposed geometric matching scheme, we obtained a retrieval rate of 92.5%.

On pattern classification with Sammon's nonlinear mapping an experimental study

Abstract Sammons mapping is conventionally used for exploratory data projection, and as such is usually inapplicable for classification. In this paper we apply a neural network (NN) implementation of Sammons mapping to classification by extracting an arbitrary number of projections. The projection map and classification accuracy of the mapping are compared with those of the auto-associative NN (AANN), multilayer perceptron (MLP) and principal component (PC) feature extractor for chromosome data. We demonstrate that chromosome classification based on Sammons (unsupervised) mapping is superior to the classification based on the AANN and PC feature extractor and highly comparable with that based on the (supervised) MLP. c 1998 Pattern Recognition Society.

Medial axis transform-based features and a neural network for human chromosome classification

Abstract Medial axis transform (MAT) based features and a multilayer perceptron (MLP) neural network (NN) were used for human chromosome classification. Two approaches to the MAT, one based on skeletonization and the other based on a piecewise linear (PWL) approximation, were examined. The former yielded a finer medial axis, as well as better chromosome classification performances. Geometrical along with intensity-based features were extracted and tested. The probability of correct training set classification of five chromosome types was 99.3–99.6%. The probability of correct test set classification was greater than 98% and greater than 97% using features extracted by the first and second approaches, respectively. It was found that only 5–10, out of all the considered features, were required to correctly classify the chromosomes with almost no performance degradation.