Abstract:
The analysis of nanopore signal depends on the exact extraction of short and multi-level blockade events. In this paper, a new method based on grayscale is proposed to detect short and multi-level events from signal automatically. In order to reduce the interferences of noise and avoid the distortion of events via the low-pass filter, the wavelet decomposition is used to achieve the denoising of signal. And then, the denoised signal is used to generate the grayscale in which the color of blockage events differs from that of baseline. The blockade events can be extract from the grayscale by thresholding the grayscale. Short-events extracted by threshold often have shorter length. To deal with this problem, this paper utilizes fuzzy C-means(FCM) to recover the extracted short events. Compared with second-order differential method (DBC) and full width half maximum (FWHM) , the FCM can recover the short event more precisely. Multi-level events are processed by converted to grayscale without thresholding. The pixels from different levels are different and these pixels with the same level are similar. These similar pixels can be regarded as one superpixel. By counting the number of different superpixels, we can obtain the levels of the multi-level events. To get the different superpixel from the grayscale automatically, this paper introduces simple linear iterative clustering (SLIC) to split the graysacle into several superpixels. Meanwhile, we can also obtain the starting point, ending point, and the average pixel value from split superpixels. Since SLIC is sensitive to the number of cluster centers, the number of cluster center may be different for different multi-level events. To avoid this shortcoming, we set the number of cluster center as 10 to cover the most cases of multi-level events. After getting the over-splite superpixels, we utilize superpixel fusion to connect two adjacent superpixels belonging to the same level by comparing the average pixel value of each superpixel. It is shown from the simulation that the SLIC superpixel fusion can obtain the accurate number of superpixels that is the same as the one of multi-level events.