A Morphological Post-Processing Approach for Overlapped Segmentation of Bacterial Cell Images

Abstract

Scanning electron microscopy (SEM) techniques have been extensively performed to image and study bacterial cells with high-resolution images. Bacterial image segmentation in SEM images is an essential task to distinguish an object of interest and its specific region. These segmentation results can then be used to retrieve quantitative measures (e.g., cell length, area, cell density) for the accurate decision-making process of obtaining cellular objects. However, the complexity of the bacterial segmentation task is a barrier, as the intensity and texture of foreground and background are similar, and also, most clustered bacterial cells in images are partially overlapping with each other. The traditional approaches for identifying cell regions in microscopy images are labor intensive and heavily dependent on the professional knowledge of researchers. To mitigate the aforementioned challenges, in this study, we tested a U-Net-based semantic segmentation architecture followed by a post-processing step of morphological over-segmentation resolution to achieve accurate cell segmentation of SEM-acquired images of bacterial cells grown in a rotary culture system. The approach showed an 89.52% Dice similarity score on bacterial cell segmentation with lower segmentation error rates, validated over several cell overlapping object segmentation approaches with significant performance improvement.

Super resolution-based methodology for self-supervised segmentation of microscopy images

“…Segmentation, also known as pixel-level classification, is employed to identify objects in microscopy images. Different architectural techniques such as single-staged (Long et al, 2015;Ronneberger et al, 2015) and two-staged (He et al, 2017) architectures can effectively perform microscopic image segmentation by training models using SL techniques (Vuola et al, 2019;Abeyrathna et al, 2021Abeyrathna et al, , 2022bKromp et al, 2021). Typically, training high-performance DNN using SL requires large amounts of manually annotated data (Huang et al, 2019).…”

Accurate and fast extraction of adhesive cells based on concave points detection and matching

“…Due to the lack of core algorithm programming and specified dataset support, we directly cite the accuracy results from References [19,20] for comparison. The Reference [19] uses the weak labeled data set for pre training, and introduced the transfer learning into the cell segmentation model.…”

“…The accuracy is 93.5%, which is similar to the proposed method. Reference [20] is based on a semantic segmentation architecture of U-Net, which uses the morphological super-segmentation resolution to achieve precise segmentation for bacterial cells. The morphological postprocessing part of this method is relatively complex, and it has poor segmentation performance for severely adhesive cells.…”

“…As for References [6,8,13,22], we programmed the algorithm and tested it using the same datasets constructed in this article. For References [19,20], due to the lack of core algorithm programming and specified dataset support, we directly cited the detection accuracy results for comparison.…”

“…Tong et al 19 constructed a transfer learning‐based cell segmentation model and used smooth constraint loss to obtain the segmented cell contour. Abeyrathna et al 20 used the semantic segmentation architecture of U‐Net and morphological supersegmentation resolution method to realize the precise segmentation of bacterial cells. However, the segmentation method based on deep learning has high training time cost, which is mainly suitable for cell classification and recognition, showing limited segmentation effect for severely adhesive images.…”