Deep learning-based aberration compensation improves contrast and resolution in fluorescence microscopy.

Optical aberrations hinder fluorescence microscopy of thick samples, reducing image signal, contrast, and resolution. Here we introduce a deep learning-based strategy for aberration compensation, improving image quality without slowing image acquisition, applying additional dose, or introducing more optics into the imaging path. Our method (i) introduces synthetic aberrations to images acquired on the shallow side of image stacks, making them resemble those acquired deeper into the volume and (ii) trains neural networks to reverse the effect of these aberrations. We use simulations to show that applying the trained 'de-aberration' networks outperforms alternative methods, and subsequently apply the networks to diverse datasets captured with confocal, light-sheet, multi-photon, and super-resolution microscopy. In all cases, the improved quality of the restored data facilitates qualitative image inspection and improves downstream image quantitation, including orientational analysis of blood vessels in mouse tissue and improved membrane and nuclear segmentation in C. elegans embryos.

Mapping port wine stain in vivo by optical coherence tomography angiography and multi-metric characterization.

Port wine stain (PWS) is a congenital cutaneous capillary malformation composed of ecstatic vessels, while the microstructure of these vessels remains largely unknown. Optical coherence tomography angiography (OCTA) serves as a non-invasive, label-free and high-resolution tool to visualize the 3D tissue microvasculature. However, even as the 3D vessel images of PWS become readily accessible, quantitative analysis algorithms for their organization have mainly remained limited to analysis of 2D images. Especially, 3D orientations of vasculature in PWS have not yet been resolved at a voxel-wise basis. In this study, we employed the inverse signal-to-noise ratio (iSNR)-decorrelation (D) OCTA (ID-OCTA) to acquire 3D blood vessel images in vivo from PWS patients, and used the mean-subtraction method for de-shadowing to correct the tail artifacts. We developed algorithms which mapped blood vessels in spatial-angular hyperspace in a 3D context, and obtained orientation-derived metrics including directional variance and waviness for the characterization of vessel alignment and crimping level, respectively. Combining with thickness and local density measures, our method served as a multi-parametric analysis platform which covered a variety of morphological and organizational characteristics at a voxel-wise basis. We found that blood vessels were thicker, denser and less aligned in lesion skin in contrast to normal skin (symmetrical parts of skin lesions on the cheek), and complementary insights from these metrics led to a classification accuracy of ∼90% in identifying PWS. An improvement in sensitivity of 3D analysis was validated over 2D analysis. Our imaging and analysis system provides a clear picture of the microstructure of blood vessels within PWS tissues, which leads to a better understanding of this capillary malformation disease and facilitates improvements in diagnosis and treatment of PWS.

Mapping variation of extracellular matrix in human keloid scar by label-free multiphoton imaging and machine learning.

Significance: Rapid diagnosis and analysis of human keloid scar tissues in an automated manner are essential for understanding pathogenesis and formulating treatment solutions. Aim: Our aim is to resolve the features of the extracellular matrix in human keloid scar tissues automatically for accurate diagnosis with the aid of machine learning. Approach: Multiphoton microscopy was utilized to acquire images of collagen and elastin fibers. Morphological features, histogram, and gray-level co-occurrence matrix-based texture features were obtained to produce a total of 28 features. The minimum redundancy maximum relevancy feature selection approach was implemented to rank these features and establish feature subsets, each of which was employed to build a machine learning model through the tree-based pipeline optimization tool (TPOT). Results: The feature importance ranking was obtained, and 28 feature subsets were acquired by incremental feature selection. The subset with the top 23 features was identified as the most accurate. Then stochastic gradient descent classifier optimized by the TPOT was generated with an accuracy of 96.15% in classifying normal, scar, and adjacent tissues. The area under curve of the classification results (scar versus normal and adjacent, normal versus scar and adjacent, and adjacent versus normal and scar) was 1.0, 1.0, and 0.99, respectively. Conclusions: The proposed approach has great potential for future dermatological clinical diagnosis and analysis and holds promise for the development of computer-aided systems to assist dermatologists in diagnosis and treatment.

Mapping the 3D remodeling of the extracellular matrix in human hypertrophic scar by multi-parametric multiphoton imaging using endogenous contrast.

The hypertrophic scar is an aberrant form of wound healing process, whose clinical efficacy is limited by a lack of understanding of its pathophysiology. Remodeling of collagen and elastin fibers in the extracellular matrix (ECM) is closely associated with scar progression. Herein, we perform label-free multiphoton microscopy (MPM) of both fiber components from human skin specimens and propose a multi-fiber metrics (MFM) analysis model for mapping the structural remodeling of the ECM in hypertrophic scars in a highly-sensitive, three-dimensional (3D) manner. We find that both fiber components become wavier and more disorganized in scar tissues, while content accumulation is observed from elastin fibers only. The 3D MFM analysis can effectively distinguish normal and scar tissues with better than 95% in accuracy and 0.999 in the area under the curve value of the receiver operating characteristic curve. Further, unique organizational features with orderly alignment of both fibers are observed in scar-normal adjacent regions, and an optimized combination of features from 3D MFM analysis enables successful identification of all the boundaries. This imaging and analysis system uncovers the 3D architecture of the ECM in hypertrophic scars and exhibits great translational potential for evaluating scars in vivo and identifying individualized treatment targets.

Identification of human ovarian cancer relying on collagen fiber coverage features by quantitative second harmonic generation imaging.

Ovarian cancer has the highest mortality rate among all gynecological cancers, containing complicated heterogeneous histotypes, each with different treatment plans and prognoses. The lack of screening test makes new perspectives for the biomarker of ovarian cancer of great significance. As the main component of extracellular matrix, collagen fibers undergo dynamic remodeling caused by neoplastic activity. Second harmonic generation (SHG) enables label-free, non-destructive imaging of collagen fibers with submicron resolution and deep sectioning. In this study, we developed a new metric named local coverage to quantify morphologically localized distribution of collagen fibers and combined it with overall density to characterize 3D SHG images of collagen fibers from normal, benign and malignant human ovarian biopsies. An overall diagnosis accuracy of 96.3% in distinguishing these tissue types made local and overall density signatures a sensitive biomarker of tumor progression. Quantitative, multi-parametric SHG imaging might serve as a potential screening test tool for ovarian cancer.

Highly accurate, automated quantification of 2D/3D orientation for cerebrovasculature using window optimizing method.

Significance: Deep-imaging of cerebral vessels and accurate organizational characterization are vital to understanding the relationship between tissue structure and function. Aim: We aim at large-depth imaging of the mouse brain vessels based on aggregation-induced emission luminogens (AIEgens), and we create a new algorithm to characterize the spatial orientation adaptively with superior accuracy. Approach: Assisted by AIEgens with near-infrared-II excitation, three-photon fluorescence (3PF) images of large-depth cerebral blood vessels are captured. A window optimizing (WO) method is developed for highly accurate, automated 2D/3D orientation determination. The application of this system is demonstrated by establishing the orientational architecture of mouse cerebrovasculature down to the millimeter-level depth. Results: The WO method is proved to have significantly higher accuracy in both 2D and 3D cases than the method with a fixed window size. Depth- and diameter-dependent orientation information is acquired based on in vivo 3PF imaging and the WO analysis of cerebral vessel images with a penetration depth of 800 µm in mice. Conclusions: We built an imaging and analysis system for cerebrovasculature that is conducive to applications in neuroscience and clinical fields.

Mapping physiological and pathological functions of cortical vasculature through aggregation-induced emission nanoprobes assisted quantitative, in vivo NIR-II imaging.

Cerebrovascular disease includes all disorders that affect cerebrovascular and cerebral circulation. Unfortunately, there is currently a lack of a systematic method to image blood vessels directly and achieve accurate quantification. Herein, we build a non-invasive, quantitative imaging and characterization system applicable to mapping physiological and pathological functions of cortical vasculature. Assisted by aggregation-induced emission (AIE) luminogens with either excitation or emission at near-infrared-II (NIR-II) region, large-depth and/or high signal-to-background ratio images of cerebral blood vessels from mice and marmosets are captured, based on which we develop an optical metric of vessel thickness in an automated, pixel-wise manner and both two-dimensional (2D) and three-dimensional (3D) contexts. By monitoring time-dependent cerebrovascular images in marmosets, periodic changes in the diameter of vibrating cerebral blood vessels are found to be regulated mainly by heartbeat. In mice photothrombosis model, vessel alterations throughout the whole process of thrombotic stroke are found to be stage-dependent. From a large field of view, the distance-dependent vessel thickness variation before and right after stroke is obtained away from the thrombus site. Importantly, a buffer zone exists right surrounding the lesion, indicating the inhomogeneity of vascular morphological changes. Biologically excretable AIE nanoparticles are used for assessing physiological and pathological functions, offering great potential for clinical translation.

Optimized number of the primary singular values for image reconstruction in reflection matrix based optical coherence tomography.

A reflection matrix based optical coherence tomography (OCT) is recently proposed and expected to extend the imaging-depth limit twice. However, the imaging depth and hence the image quality heavily depend on the number of primary singular values considered for image reconstruction. To this regard, we propose a method based on correlation between image pairs reconstructed from different number of singular values and corresponding remainders. The obtained correlation curve and another feature curve fetched from the former are then fed to a long short-term memory (LSTM) network classifier to identify the optimized number of primary singular values for image reconstruction. Simulated targets with different combinations of filling fraction and signal-to-noise ratio (SNR) are reconstructed by the developed method as well as two current adopted methods for comparison. The results demonstrate that the proposed method is robust to recover the image with satisfactory similarity close to the reference one. To our knowledge, this is the first comprehensive study on the optimized number of the primary singular values considered for image reconstruction in reflection matrix based OCT.

Identification of endoplasmic reticulum formation mechanism by multi-parametric, quantitative super-resolution imaging.

The endoplasmic reticulum (ER) is a highly dynamic membrane-bound organelle in eukaryotic cells which spreads throughout the whole cell and contacts and interacts with almost all organelles, yet quantitative approaches to assess ER reorganization are lacking. Herein we propose a multi-parametric, quantitative method combining pixel-wise orientation and waviness features and apply it to the time-dependent images of co-labeled ER and microtubule (MT) from U2OS cells acquired from two-dimensional structured illumination microscopy (2D SIM). Analysis results demonstrate that these morphological features are sensitive to ER reshaping and a combined use of them is a potential biomarker for ER formation. A new, to the best of our knowledge, mechanism of MT-associated ER formation, termed hooking, is identified based on distinct organizational alterations caused by interaction between ER and MT which are different from those of the other three mechanisms already known, validated by 100% discrimination accuracy in classifying four MT-associated ER formation mechanisms.