SEM Image Analysis for Quality Control of Nanoparticles

In nano-medicine, mesoporous silicon particles provide efficient vehicles for the dissemination and delivery of key proteins at the micron scale. We propose a new quality-control method for the nanopore structure of these particles, based on image analysis software developed to automatically inspect scanning electronic microscopy (SEM) images of nanoparticles in a fully automated fashion. Our algorithm first identifies the precise position and shape of each nanopore, then generates a graphic display of these nanopores and of their boundaries. This is essentially a texture segmentation task, and a key quality-control requirement is fast computing speed. Our software then computes key shape characteristics of individual nanopores, such as area, outer diameter, eccentricity, etc., and then generates means, standard deviations, and histograms of each pore-shape feature. Thus, the image analysis algorithms automatically produce a vector from each image which contains relevant nanoparticle quality control characteristics, either for comparison to pre-established acceptability thresholds, or for the analysis of homogeneity and the detection of outliers among families of nanoparticles. 1 SEM Image Data and Quality Control Targets Quality control in the production of nanostructures poses a challenge for image processing, because it requires interpreting high-resolution images which may be plagued by substantial amounts of noise of various characteristics, and because the material is by design heterogeneous and thus requires flexible analysis algorithms. We present new algorithms for SEM images analysis focused on quality control of porous silicon (pSi) and associated microparticles (PSMs). Visual inspection of nanoparticles is performed on 24bit SEM images typically of size 1024×768. Each nanoparticle occupies an image surface of approximately 500 × 500 pixels, has dimensions of the order of 3× 3 microns and gathers between 500 and 1000 nanopores having similar shapes. The main goal of our algorithmic image analysis is first to identify the precise positions and shapes of each nanopore and its boundary in a fully automated fashion, and generate a graphic display of these nanopores and boundaries. This is essentially a texture segmentation task, and a key quality control requirement is fast computing X. Jiang and N. Petkov (Eds.): CAIP 2009, LNCS 5702, pp. 590–597, 2009. c © Springer-Verlag Berlin Heidelberg 2009 SEM Image Analysis for Quality Control of Nanoparticles 591 (a) SEM Image (b) Region of Interest (c) Segmented Pores Fig. 1. Example nanoparticle AN24-GP-12 speed. Then a second algorithm automatically analyzes the shapes of the detected nanopores in order to compute key shape characteristics of nanopores, such as area, outer diameter, eccentricity, boundary thickness, etc.); we thus generate for each nanoparticle a database of roughly 500 to 1000 vectors of individual pore features. At the level of each nanoparticle, we then launch automatic extraction of statistical characteristics of this database, to compute the means, standard deviations, and histograms of each type of shape feature. This defines a vector of nanoparticle characteristics, which thus provides very natural quality control features, either for comparison to pre-established acceptability thresholds, or to analyze homogeneity and outliers among families of nanoparticles. 2 Porous Silicon Microparticles and Nanomedicine Applications Since the initial proof of its biocompatibility [1] porous silicon has been actively researched as a biomaterial[2,3,4,5,6]. Porous silicon microparticles (PSMs) have demonstrated their efficacy as delivery vectors for therapeutics. PSMs obtained by sonication or ball milling of pSi layers successfully acted as loading and release agents for different drug molecules, encompassing a wide spectrum of solubility and acid/base characteristics[7]. Proteins were also successfully loaded and released from PSM[8]. Oral delivery of pSi has been proven safe[9] and paracellular drug delivery by means of PSMs has been demonstrated in vitro [10]. However, the size and shape polydispersion of PSMs obtained by sonication or ball milling forbids their use as vascular delivery systems. Our group has successfully developed a strategy based on mathematical models [11,12,13], to produce monodisperse porous silicon microparticles of tailored pore size, shape, and porosity (porous silicon elements, PSEs)[14]. We have proven short term safety of PSEs upon injection [15], and demonstrated their suitability as primary vectors in a multi-stage delivery system [16]. The porous silicon elements are produced in a silicon fabrication environment using a top-down approach resulting in selective porosification of bulk silicon wafers. The fabrication process involves multiple 592 S.K. Alexander et al. steps: thin film deposition, photolithography, selective dry etch, electrochemical etch, etc. The process is subject to batch-to-batch variations that may influence the final product. Variations in lithographic steps may lead to a different PSE size or shape, and hence to variation of pore size and porosity. But to guarantee the PSEs efficacy as primary delivery vectors, their size, shape, and pore sizes must be reproduced within stringent limits to avoid modifying their flow and margination characteristics , altering the payload biodistribution and release profile due to different diffusion characteristics [17] and pSi degradation kinetics [16]. Currently, quality assessment for PSEs is a two step process. Initially a statistically relevant sample of particles from a single production lot is analyzed by expert interactive measurements on SEM images to assess size and shape uniformity. Secondly ten or more production lots are joined to obtain the minimum 10mg sample size necessary for nitrogen absorption/desorption analysis of pore size and porosity. This latter step risks rejection of good quality lots (representing significant time and resources spent) due to necessarily mixing with other lots. The alternative software based algorithmic image analysis we propose here for quality control of PSEs is much faster and generates robust quantitative evaluations of pore sizes and shapes.