Data Mining Approach to Characterization of Bioactive Inorganic Scaffold Properties Using Synthetic Images

Abstract

The fundamental scientific problem in the paper is the patterns discovery using artificial intelligence to search for and determine the properties of bioactive inorganic scaffold nanomaterials from datasets with scanning electron microscopy (SEM) images. Bioactive inorganic scaffold nanomaterials are bioactive structures that are designed to temporarily provide bone functions. Due to the complexity of biomechanical processes of bone tissue regeneration, there is a set of requirements for scaffolds that exists. For example, porosity, diffusivity, permeability, and tortuosity are included in that set as main mechanical properties. Experimental chemical synthesis of scaffolds with predefined properties requires routine and manual cyclic laboratory operations with consumption of chemical substances until now. In this cycle, operators control desired scaffold properties by visually looking at SEM images of synthesized nanomaterial. Usually, SEM image datasets are not publicity distributed, have a limited number of samples and properties of material shown in SEM images can vary and out of interest depending on research conditions. As a novel approach, we have proposed an intelligent technology that applies the synthetic generation of SEM images and involves properties detected from images to the control process. Intelligent technology requires the minimum number of iterations in experimental synthesis of a material with the predefined biomechanical and functional characteristics. The main result of this paper is the intelligent convergence the chemistry exploration with SEM image synthesis that improves the characterization of scaffold morphology and determination of its mechanical properties.