Correcting anisotropic intensity in light sheet images using dehazing and image morphology

Abstract

Light-sheet fluorescence microscopy (LSFM) provides access to multi-dimensional and multi-scale in vivo imaging of animal models with highly coherent volumetric reconstruction of the tissue morphology, via a focused laser light sheet. The orthogonal illumination and detection LSFM pathways account for minimal photobleaching and deep tissue optical sectioning through different perspective views. Although rotation of the sample and deep tissue scanning constitutes major advantages of LSFM, images may suffer from intrinsic problems within the modality, such as light mismatch of refractive indices between the sample and mounting media and varying quantum efficiency across different depths. To overcome these challenges, we hereby introduce an illumination correction technique integrated with depth detail amelioration to achieve symmetric contrast in large field-of-view images acquired using a low power objective lens. Due to an increase in angular dispersion of emitted light flux with the depth, we combined the dehazing algorithm with morphological operations to enhance poorly separated overlapping structures with subdued intensity. The proposed method was tested on different LSFM modalities to illustrate its applicability on correcting anisotropic illumination affecting the volumetric reconstruction of the fluorescently tagged region of interest.

Automatic segmentation and cardiac mechanics analysis of evolving zebrafish using deep-learning

“…In our case, using the fluorescent label zebrafish, tg(cmlc2:gfp), yielded overall strong results from the ventricle, but on occasion could have issues with the atria due to lower cardiomyocyte density and its location deeper in the chest, leading to more imaging issues. Circumventing this issue was not a trivial action, as exploring extensive pre-processing methods was required to be able to find the boundary of the inner volume of the heart [26]. Other limitations which could be found in the experimentation process could be issues within the imaging process, in which user and systematic errors could hinder image fidelity.…”

Feature Detection to Segment Cardiomyocyte Nuclei for Investigating Cardiac Contractility

“…In this regard, feature detection methods help refine data dimensionality by discarding dispensable image attributes, thereby producing a more compact feature space 1,2 . In biomedical research, endogenous fluorescent signals are commonly used to study in vivo spatiotemporal organogenesis. However, optical microscopy inherently suffers from anisotropic illumination in the optical image plane 4 . This anisotropy is due to aberrations caused by photon propagation through heterogeneous tissue morphology or significantly varied sample size; hence, optimization of image acquisition parameters is a challenging task 4 .…”

“…However, optical microscopy inherently suffers from anisotropic illumination in the optical image plane 4 . This anisotropy is due to aberrations caused by photon propagation through heterogeneous tissue morphology or significantly varied sample size; hence, optimization of image acquisition parameters is a challenging task 4 . Precise orchestration of any volumetric reconstruction analysis requires high feature detection sensitivity with respect to tissue protrusions and changes in illumination, rotation, local scaling changes, and motion translation 5 .…”

“…These optical aberrations are prone to induce redundancy in image features 4 , rendering the task of manual feature extraction intrinsically subjective and laborious 6,7 . Hence, feature detection has gained popularity for augmenting performance of downstream classifiers or regression estimators to improve their interpretability and accuracy 8 .…”

“…The feature detection framework can be efficiently applied for cell segmentation in developmental biology 3,23 with post-processing strategies based on mathematical morphology such as opening, closing, top-hat and bottom-hat transform for pruning redundant binary features 4,28 . Contemporary sophistication for dealing with heterogeneity in nuclear volumes involves a predetermined template based approach 29 .…”