RESUMO
Morphology of erythrocyte membrane has been recognized as an alternative biomarker of several patho-physiological states. Numerous attempts have been made to upgrade the existing method of primitive manual counting, particularly exploring the light scattering properties of erythrocyte. All the techniques are at best semi-empirical and heavily rely on the effectiveness of the statistical correlations. Precisely, this is due to the lack of a non-empirical scale of the so-called "morphological scores". In this article, fractal dimension of erythrocyte membrane has been used to formulate a suitable scoring scale. Subsequently, the rapid experimental output of flow-cytometry has been functionally related to the mean morphological quantifier of the whole cell population via an optimum neural network model (R2 = 0.98). Moreover, the fractal dimension has been further demonstrated to be an important parameter in early detection of an abnormal patho-physiological state, even without any noticeable poikilocytic transformation in micrometric domain.
Assuntos
Membrana Eritrocítica , Citometria de Fluxo , Fractais , Humanos , Microscopia de Força Atômica , Microscopia Confocal , Redes Neurais de Computação , Propriedades de SuperfícieRESUMO
Erythrocyte morphology is gaining importance as a powerful pathological index in identifying the severity of any blood related disease. However, the existing technique of quantitative microscopy is highly time consuming and prone to personalized bias. On the other hand, relatively unexplored, complementary technique based on flow cytometry has not been standardized till date, particularly due to the lack of a proper morphological scoring scale. In this article, we have presented a new approach to formulate a non-empirical scoring scale based on membrane roughness (R(rms)) data obtained from atomic force microscopy. Subsequently, the respective morphological quantifier of the whole erythrocyte population, commonly known as morphological index, was expressed as a function of highest correlated statistical parameters of scattered signal profiles generated by flow cytometry. Feed forward artificial neural network model with multilayer perceptron architecture was used to develop the intended functional form. High correlation coefficient (R(2) = 0.95), even for model-formulation exclusive samples, clearly indicates the universal validity of the proposed model. Moreover, a direct pathological application of the proposed model has been illustrated in relation to patients, diagnosed to be suffering from a wide variety of cancer.