Segmentation of Drosophila heart in optical coherence microscopy images using convolutional neural networks

Abstract

Convolutional neural networks (CNNs) are powerful tools for image segmentation and classification. Here, we use this method to identify and mark the heart region of Drosophila at different developmental stages in the cross-sectional images acquired by a custom optical coherence microscopy (OCM) system. With our well-trained CNN model, the heart regions through multiple heartbeat cycles can be marked with an intersection over union of ~86%. Various morphological and dynamical cardiac parameters can be quantified accurately with automatically segmented heart regions. This study demonstrates an efficient heart segmentation method to analyze OCM images of the beating heart in Drosophila.

LeukocyteMask: An automated localization and segmentation method for leukocyte in blood smear images using deep neural networks

“…Different from traditional data augment techniques as used in Refs. , in which only affine transformation was utilized which functions weakly in this case, in this paper, by observing the samples in the collected datasets, we postulate that the distribution of data has some invariance with respect to not only affine transformations, but also elastic deformations caused by the adhesion between cells and the growth of cell itself. So, in the experiment, not only affine transformations such as flip from horizontal (FlipH), flip vertical (FlipV), Rotation (range from −180° to 180°), are employed, but also a kind of elastic transformation (ET) used in Refs.…”

“…For evaluation metrics, 6 measures are used in the experiment, which consist of 3 commonly used measure scores in deep learning‐based methods , namely Precision, Dice coefficient (Dice) and mean Intersection over Union (mIoU), respectively, and other 3 commonly used metrics for traditional segmentation methods , namely false positive rate (FPR), false negative rate (FNR) and misclassification error (ME), respectively. All 6 metrics are defined as in Eqs.…”

“…In medical image segmentation, lots of CNN‐based methods are also widely used benefitting from its powerful ability of feature learning and representation. Among these methods, the fully convolutional network (FCN) has shown the state‐of‐the‐art performance in cell and organ segmentation problems . In Ref.…”

“…In Ref. , U‐Net‐based network is employed to identify and segment the heart region of Drosophila at different developmental stages, which achieves a high intersection over union (IoU) rate on custom optical coherence microscopy images. In Ref.…”

Deep learning a boon for biophotonics?

“…Similar to Figure , the nonlinear multimodal image is composed of CARS, TPEF and SHG signal. The images are reprinted from references with permissions. AE, auto‐encoder; CARS, coherent anti‐stokes Raman scattering; SHG, second‐harmonic generation; TPEF, two‐photon excited fluorescence microscopy…”