A New Image Analysis Based Method for Measuring Electrospun Nanofiber Diameter
© to the authors 2007
Received: 7 August 2007
Accepted: 27 August 2007
Published: 14 November 2007
In this paper, a new image analysis based method for electrospun nanofiber diameter measurement has been presented. The method was tested by a simulated image with known characteristics and a real web. Mean (M) and standard deviation (STD) of fiber diameter obtained using this method for the simulated image were 15.02 and 4.80 pixels respectively, compared to the true values of 15.35 and 4.47 pixels. For the real web, applying the method resulted in M and STD of 324 and 50.4 nm which are extremely close to the values of 319 and 42 nm obtained using manual method. The results show that this approach is successful in making fast, accurate automated measurements of electrospun fiber diameters.
KeywordsElectrospinning Nanofiber diameter Image analysis Distance transform Local thresholding μ-ramdomness
Nanofibers are considered as fibers with diameters in the range of one or two orders of magnitude smaller than those of conventional textile fibers. Electrospinning is a straightforward method for manufacturing nanofibers that combines the benefits of a controlled fiber diameter with the possibility for large scale automated production. In this process, a high electric field is applied between a target plate and nozzle of a cylinder containing a polymer/solvent dispersion. As the dispersion is pushed through the nozzle, an electrically charged jet of polymer is formed, which, on drying by means of evaporation of solvent, results in randomly oriented nanofibers collected on the target [1–4]. The fiber diameter is one of the most important structural characteristics in the electrospun nanofiber webs and it may be altered by varying electrospinning parameters such as the dispersion viscosity. However, the extremely small fiber size and random orientation of the nanofibers make the measurement of their diameter very difficult. In addition, large scale production of nanofibers requires unique on-line quality control.
To date there has been no successful method developed for the automated measuring of electrospun fiber diameter which has hindered the uptake of this process for large-scale production. The objective of this investigation is to develop a new image analysis based algorithm for measuring electrospun nanofiber diameter.
Fiber diameter is usually determined from Scanning Electron Microscopy (SEM) images obtained of the electrospun webs. Due to the small fiber dimensions, high-quality images with appropriate magnifications are required.
Routine measurement of fiber diameter and its distribution are carried out by manual method. First the scale is set. Then, pixels between two edges of a fiber perpendicular to the fiber axis are counted. The number of the pixels is then converted to nm using the scale and the resulting diameter is recorded. On a typical image, the diameter of 100 fibers are measured using this method and the histogram of fiber diameter distribution is plotted. This process is very time-consuming and operator consistency and fatigue may reduce the accuracy. Identifying the edges of the fibers needs attention and the measurements are not exactly made perpendicular to the fiber axis. It cannot be used as on-line method for quality control, due to the need of an operator for measurement. Automating the fiber diameter measurement and eliminating the use of the human operator is a natural solution to this problem.
An image analysis based method was proposed by Pourdeyhimi et al.  for measuring fiber diameter in nonwoven textiles. In this method, a binary image of the textile is used to create a distance map and skeleton. The fiber diameter may be determined from the values of the distance map at any pixel location on the skeleton. However, the occurrence of a broken skeleton at intersection points is a main challenging area within the use of this method. Since two or more fibers cross each other at these intersections, the value of the center of the object in the distance map doesn’t coincide with the fiber diameter at these points because it isn’t associated with a single fiber. The problem becomes more serious as fibers get thicker and for points where more fibers cross each other. Hence, the method fails in measuring fiber diameter at intersections.
New Distance Transform Algorithm
In the next step, the obtained skeleton after deleting the intersections is used as a guide for tracking the distance transformed image and the diameters are computed from the intensities of this image at all points along the skeleton. The data in pixels may then be converted to nm and the histogram of fiber diameter distribution is plotted.
In order to validate the method, test samples with known structural characteristics are required. Algorithms for simulation of nonwoven mats have been proposed by Abdel-Ghani et al.  and Pourdeyhimi et al. . One of the most important features of simulation is that it allows several structural characteristics such as fiber diameter and orientation distribution and web density to be taken into consideration. Since these characteristics of simulated images are known exactly, one can employ them to test the usefulness of algorithm used in characterizing diameter and other structural features. Lately it has been discovered that the best way to simulate nonwoven mats of continuous fibers is through μ-randomness procedure . In this scheme, a line with a specified thickness is drawn at a distance away from a reference point and perpendicular to an angular position [7, 8]. In this study μ-randomness procedure was used for generating a simulated image with known characteristics.
Real Web Treatment
Fiber diameter determination by the use of image analysis requires the initial segmentation of the micrographs in order to produce binary images. The typical way of producing a binary image from a grayscale image is by global thresholding [6, 9] where a single constant threshold, usually selected by trial and error, is applied to segment the image. Global thresholding is very sensitive to any inhomogeneities in the gray-level distributions of the object and background pixels. This effect can be eliminated through the use of a local thresholding [6, 9] scheme. Automatic selection of the appropriate thresholds can be carried out based on, Otsu’s method . Note that, since the process is extremely sensitive to noise, before the segmentation, a procedure to clean the noise and enhance the contrast of the image is necessary.
Results and Discussion
Figure 4b shows the diameter distribution for the real web. The respective M and STD of the fiber diameter obtained by this new method were 24.74 and 3.85 in terms of pixel and 323.7 and 50.4 in term of nm which are in good agreement with the values 24.36 and 3.19 pixels and 318.7 and 41.8 nm obtained from manual methods. The differences here can also be attributed to the different number of measurements taken between the methods used (over 2,000 for our method versus 100 for the manual method). Nevertheless, in each case presented, the difference observed was within 1-pixel measurement error suggesting the main limitation with the process is with the resolution of the taken image.
In this study, a new image analysis based method for assessing nanofibers diameters was successfully developed. The validity of the method was tested using a simulated image as well as an image of a real electrospun nanofiber web. In the case of the real web, local thresholding was applied on the micrograph of the web taken from SEM to attain the necessary binary image. The M and STD of fiber diameter which were obtained using this new method were extremely close to true values on the simulated image. For the real web, M and STD of fiber diameter measured by the method were also in good agreement with those obtained from the manual method. The results show the effectiveness of the method for diameter measurement. The method is automated, accurate, and much faster than manual method and has the capability of being used as an on-line technique for quality control.
- Haghi AK, Akbari M: Phys. Stat. Sol. A. 2007, 204: 1830. COI number [1:CAS:528:DC%2BD2sXntl2ms7c%3D] 10.1002/pssa.200675301View ArticleGoogle Scholar
- Reneker DH, Chun I: Nanotechnology. 1996, 7: 216. COI number [1:CAS:528:DyaK28XmvFOjtrs%3D] 10.1088/0957-4484/7/3/009View ArticleGoogle Scholar
- H. Fong, D.H. Reneker, Electrospinning and the Formation of Nanofibers, in Structure Formation in Polymeric Fibers, chap. 6, ed. by D.R. Salem (Hanser, Cincinnati, 2001), pp. 225–246Google Scholar
- Subbiah Th, Bhat GS, Tock RW, Parameswaran S, Ramkumar SS: J. Appl. Polym. Sci.. 2005, 96: 557. COI number [1:CAS:528:DC%2BD2MXisFGntLk%3D] 10.1002/app.21481View ArticleGoogle Scholar
- Pourdeyhimi B, Dent R: Text. Res. J.. 1999, 69: 233. COI number [1:CAS:528:DyaK1MXisFyrtb0%3D]View ArticleGoogle Scholar
- Gonzalez RC, Woods RE: Digital Image Processing. Prentice Hall, New Jersey; 2001.Google Scholar
- Abdel-Ghani MS, Davis GA: Chem. Eng. Sci.. 1985, 40: 117. COI number [1:CAS:528:DyaL2MXhtFGlsr4%3D] 10.1016/0009-2509(85)85052-1View ArticleGoogle Scholar
- Pourdeyhimi B, Ramanathan R, Dent R: Text. Res. J.. 1996, 66: 713. COI number [1:CAS:528:DyaK28XmvVKlt74%3D] 10.1177/004051759606601107View ArticleGoogle Scholar
- Petrou M, Bosdogianni P: Image Processing the Fundamentals. John Wiley and Sons, England; 1999.View ArticleGoogle Scholar