Microscopy cell nuclei segmentation with enhanced U-Net

Abstract

Background: Cell nuclei segmentation is a fundamental task in microscopy image analysis, based on which multiple biological related analysis can be performed. Although deep learning (DL) based techniques have achieved state-of-the-art performances in image segmentation tasks, these methods are usually complex and require support of powerful computing resources. In addition, it is impractical to allocate advanced computing resources to each dark-or bright-field microscopy, which is widely employed in vast clinical institutions, considering the cost of medical exams. Thus, it is essential to develop accurate DL based segmentation algorithms working with resources-constraint computing. Results: An enhanced, light-weighted U-Net (called U-Net+) with modified encoded branch is proposed to potentially work with low-resources computing. Through strictly controlled experiments, the average IOU and precision of U-Net+ predictions are confirmed to outperform other prevalent competing methods with 1.0% to 3.0% gain on the first stage test set of 2018 Kaggle Data Science Bowl cell nuclei segmentation contest with shorter inference time. Conclusions: Our results preliminarily demonstrate the potential of proposed U-Net+ in correctly spotting microscopy cell nuclei with resources-constraint computing.

3D Automated Segmentation of Lower Leg Muscles Using Machine Learning on a Heterogeneous Dataset

“…U-Net architectures are often used as a comparative baseline for other network architectures [11]. They have been widely used and adapted to clinical applications from detecting skin lesions [12], parotid glands [13], pulmonary nodules [14], segmented infant-brain MR-images [15], cardiac segmentation [16], as well as cell structures in light microscopy images [17]. ResNet architectures use residual connection and allow blocks to learn residual functions.…”

Deep learning for bioimage analysis in developmental biology

“…DL approaches originating from computer vision have greatly enhanced the speed and accuracy of both object detection and segmentation in biological images. Since U-net, countless customized DL models have adapted to bioimage-specific object detection ( Waithe et al, 2020 ; Wollmann and Rohr, 2021 ) and segmentation problems have been proposed ( Long, 2020 ; Chidester et al, 2019 ; Tokuoka et al, 2020 ). A strong link to computer vision remains, as many of these methods draw from partitioning tasks in natural images.…”

PyConvU-Net: a lightweight and multiscale network for biomedical image segmentation

“…Based on the accurate segmentation, multiple biological or medical analyses [ 2 ] can be performed subsequently, including cell counting [ 3 ], quantitative measurement of anatomical structure [ 4 ], cell phenotype analysis [ 5 ], subcellular localization [ 6 ], etc., providing valuable diagnostic information for doctors and researchers [ 7 ]. Although conventional image processing techniques are still employed for this time and labor-consuming task, they often cannot achieve the optimized performance due to different reasons, such as the limited capability of dealing with diverse images [ 8 ], lack of computing source, and so on.…”

“…With Eqs. (7) and (8), regardless of the number of levels of PyConv and the increasing kernel size, the computational cost (in terms of FLOPs) and the number of parameters are the same as the standard convolution with a single kernel size.…”

“…Comparing with the U-Net, PyConvU-Net has a similar number of parameters and requirements in computational resources, as shown in Eqs. 7and (8). Meanwhile, PyConvU-Net offers a high degree of parallelism due to the fact that the pyramid levels can be independently computed in parallel.…”