Resolution invariant wavelet features of melanoma studied by SVM classifiers.

This article refers to the Computer Aided Diagnosis of the melanoma skin cancer. We derive wavelet-based features of melanoma from the dermoscopic images of pigmental skin lesions and apply binary C-SVM classifiers to discriminate malignant melanoma from dysplastic nevus. The aim of this research is to select the most efficient model of the SVM classifier for various image resolutions and to search for the best resolution-invariant wavelet bases. We show AUC as a function of the wavelet number and SVM kernels optimized by the Bayesian search for two independent data sets. Our results are compatible with the previous experiments to discriminate melanoma in dermoscopy images with ensembling and feed-forward neural networks.

Symbolic learning supporting early diagnosis of melanoma.

We present a classification analysis of the pigmented skin lesion images taken in white light based on the inductive learning methods by Michalski (AQ). Those methods are developed for a computer system supporting the decision making process for early diagnosis of melanoma. Symbolic (machine) learning methods used in our study are tested on two types of features extracted from pigmented lesion images: coloristic/geometric features, and wavelet-based features. Classification performance with the wavelet features, although achieved with simple rules, is very high. Symbolic learning applied to our skin lesion data seems to outperform other classical machine learning methods, and is more comprehensive both in understanding, and in application of further improvements.

Cyclin D1 and D3 expression in melanocytic skin lesions.

Cyclins, cyclin-dependent kinases, as well as proteins cooperating with them are responsible for cell cycle regulation which is crucial for normal development, injury repair, and tumorigenesis. D-type cyclins regulate G1 cell cycle progression by enhancing the activities of cyclin-dependent kinases, and their expression is frequently altered in tumors. Disturbances in cyclin expression were also reported in melanocytic skin lesions. The objective of the study was to evaluate the expression of cyclins D1 and D3 in common, dysplastic, and malignant melanocytic skin lesions. Forty-eight melanocytic skin lesions including common nevi (10), dysplastic nevi (24), and melanomas (14) were diagnosed by dermoscopy and excised. Expression of cyclin D1 and D3 was detected by immunohistochemistry and quantified as percentage of immunostained cell nuclei in each sample. In normal skin, expression of cyclins D1 and D3 was not detected. The mean percentage of cyclin D1-positive nuclei was 7.75% for melanoma samples, 5% for dysplastic nevi samples, and 0.34% for common nevi samples. For cyclin D3, the respective values were 17.8, 6.4, and 1.8%. Statistically significant differences in cyclin D1 expression were observed between melanomas and common nevi as well as between dysplastic and common nevi (p = 0.0001), but not between melanomas and dysplastic nevi. Cyclin D3 expression revealed significant differences between all investigated lesion types (p = 0.0000). The mean cyclin D1 and D3 scores of melanomas with Breslow thickness <1 mm and >1 mm were not significantly different. G1/S abnormalities are crucial for the progression of malignant melanoma, and enhanced cyclin D1 and D3 expression leading to increased melanocyte proliferation is observed in both melanoma and dysplastic nevi. In histopathologically ambiguous cases, lower cyclin D3 expression in dysplastic nevi can be a diagnostic marker for that lesion type.

Different learning paradigms for the classification of melanoid skin lesions using wavelets.

We use the wavelet-based decomposition to generate the multiresolution representation of dermatoscopic images of potentially malignant pigmented lesions. Three different machine learning methods are experimentally applied, namely neural networks, support vector machines, and Attributional Calculus. The obtained results confirm that neighborhood properties of pixels in dermatoscopic images are a sensitive probe of the melanoma progression and together with the selected machine learning methods may be an important diagnostic tool.

A tabular approach to the sequence-to-structure relation in proteins (tetrapeptide representation) for de novo protein design.

BACKGROUND: Experimental observations classify the protein-folding process as a multi-step event. The backbone conformation has been experimentally recognized as responsible for the early-stage structural forms of a polypeptide. The sequence-to-structure and structure-to-sequence relation is critical for predicting protein structure. A contingency table representing this relation for tetrapeptides in their early-stage is presented. Their correlation seems to be essential in protein-folding simulation. MATERIAL/METHODS: The polypeptide chains of all the proteins in the Protein Data Bank were transformed into their early-stage structural forms. The tetrapeptide was selected as the structural unit. Tetrapetide sequences and structures were expressed by letter codes. The transformation of a contingency table of any size (here: 160,000x2401) to a 2x2 table performed for each non-zero cell of the original table allowed calculation of the rho-coefficient measuring the strength of the relation. RESULTS: High values of the rho-coefficient extracted sequences of strong structural determinability and structures of high sequence selectivity. The web-site program to calculate the rho-coefficient ranking list was constructed to enable applying this method to any problem of contingency table analysis. CONCLUSIONS: The results revealed sequence-to-structure (and vice versa) correlation in early-stage folding. Surprisingly, the irregular structural forms of loops and bends appeared to be highly determined. Comparison of these results with another method based on information entropy revealed high accordance. The method oriented on interpretation of a large contingency table seems very useful especially for large-scale microarray analysis, a very popular technique in the post-genomic era.