Interpretable deep learning model for major depressive disorder assessment based on functional near-infrared spectroscopy.

BACKGROUND: Major depressive disorder (MDD) affects a substantial number of individuals worldwide. New approaches are required to improve the diagnosis of MDD, which relies heavily on subjective reports of depression-related symptoms. AIM: Establish an objective measurement and evaluation of MDD. METHODS: Functional near-infrared spectroscopy (fNIRS) was used to investigate the brain activity of MDD patients and healthy controls (HCs). Leveraging a sizeable fNIRS dataset of 263 HCs and 251 patients with MDD, including mild to moderate MDD (mMDD; n = 139) and severe MDD (sMDD; n = 77), we developed an interpretable deep learning model for screening MDD and staging its severity. RESULTS: The proposed deep learning model achieved an accuracy of 80.9% in diagnostic classification and 78.6% in severity staging for MDD. We discerned five channels with the most significant contribution to MDD identification through Shapley additive explanations (SHAP), located in the right medial prefrontal cortex, right dorsolateral prefrontal cortex, right superior temporal gyrus, and left posterior superior frontal cortex. The findings corresponded closely to the features of haemoglobin responses between HCs and individuals with MDD, as we obtained a good discriminative ability for MDD using cortical channels that are related to the disorder, namely the frontal and temporal cortical channels with areas under the curve of 0.78 and 0.81, respectively. CONCLUSION: Our study demonstrated the potential of integrating the fNIRS system with artificial intelligence algorithms to classify and stage MDD in clinical settings using a large dataset. This approach can potentially enhance MDD assessment and provide insights for clinical diagnosis and intervention.

Multi-level and joint attention networks on brain functional connectivity for cross-cognitive prediction.

Deep learning on resting-state functional MRI (rs-fMRI) has shown great success in predicting a single cognition or mental disease. Nevertheless, cognitive functions or mental diseases may share neural mechanisms that can benefit their prediction/classification. We propose a multi-level and joint attention (ML-Joint-Att) network to learn high-order representations of brain functional connectivities that are specific and shared across multiple tasks. We design the ML-Joint-Att network with edge and node convolutional operators, an adaptive inception module, and three attention modules, including network-wise, region-wise, and region-wise joint attention modules. The adaptive inception learns brain functional connectivity at multiple spatial scales. The network-wise and region-wise attention modules take the multi-scale functional connectivities as input and learn features at the network and regional levels for individual tasks. Moreover, the joint attention module is designed as region-wise joint attention to learn shared brain features that contribute to and compensate for the prediction of multiple tasks. We employed the Adolescent Brain Cognitive Development (ABCD) dataset (n =9092) to evaluate the ML-Joint-Att network for the prediction of cognitive flexibility and inhibition. Our experiments demonstrated the usefulness of the three attention modules and identified brain functional connectivities and regions specific and common between cognitive flexibility and inhibition. In particular, the joint attention module can significantly improve the prediction of both cognitive functions. Moreover, leave-one-site cross-validation showed that the ML-Joint-Att network is robust to independent samples obtained from different sites of the ABCD study. Our network outperformed existing machine learning techniques, including Brain Bias Set (BBS), spatio-temporal graph convolution network (ST-GCN), and BrainNetCNN. We demonstrated the generalization of our method to other applications, such as the prediction of fluid intelligence and crystallized intelligence, which also outperformed the ST-GCN and BrainNetCNN.

Light-sheet laser speckle imaging for cilia motility assessment.

Mucociliary clearance is an important innate defense mechanism predominantly mediated by ciliated cells in the upper respiratory tract. Ciliary motility on the respiratory epithelium surface and mucus pathogen trapping assist in maintaining healthy airways. Optical imaging methods have been used to obtain several indicators for assessing ciliary movement. Light-sheet laser speckle imaging (LSH-LSI) is a label-free and non-invasive optical technique for three-dimensional and quantitative mapping of velocities of microscopic scatterers. Here, we propose to use an inverted LSH-LSI platform to study cilia motility. We have experimentally confirmed that LSH-LSI can reliably measure the ciliary beating frequency and has the potential to provide many additional quantitative indicators for characterizing the ciliary beating pattern without labeling. For example, the asymmetry between the power stroke and the recovery stroke is apparent in the local velocity waveform. PIV (particle imaging velocimetry) analysis of laser speckle data could determine the cilia motion directions in different phases.

Identifying neuroimaging biomarkers of major depressive disorder from cortical hemodynamic responses using machine learning approaches.

BACKGROUND: Early diagnosis of major depressive disorder (MDD) could enable timely interventions and effective management which subsequently improve clinical outcomes. However, quantitative and objective assessment tools for the suspected cases who present with depressive symptoms have not been fully established. METHODS: Based on a large-scale dataset (n = 363 subjects) collected with functional near-infrared spectroscopy (fNIRS) measurements during the verbal fluency task (VFT), this study proposed a data representation method for extracting spatiotemporal characteristics of NIRS signals, which emerged as candidate predictors in a two-phase machine learning framework to detect distinctive biomarkers for MDD. Supervised classifiers (e.g., support vector machine (SVM), k-nearest neighbors (KNN)) cooperated with cross-validation were implemented to evaluate the predictive capability of selected features in a training set. Another test set that was not involved in developing the algorithms enabled the independent assessment of the model's generalization. FINDINGS: For the classification with the optimal fusion features, the SVM classifier achieved the highest accuracy of 75.6% ± 4.7% in the nested cross-validation, and the correct prediction rate of 78.0% with a sensitivity of 75.0% and a specificity of 81.4% in the test set. Moreover, the multiway ANOVA test on clinical and demographic factors confirmed that twenty out of 39 optimal features were significantly correlated with the MDD-distinctive consequence. INTERPRETATION: The abnormal prefrontal activity of MDD may be quantified as diminished relative intensity and inappropriate activation timing of hemodynamic response, resulting in an objectively measurable biomarker for assessing cognitive deficits and screening MDD at the early stage. FUNDING: This study was funded by NUS iHeathtech Other Operating Expenses (R-722-000-004-731).

Optical breast atlas as a testbed for image reconstruction in optical mammography.

We present two optical breast atlases for optical mammography, aiming to advance the image reconstruction research by providing a common platform to test advanced image reconstruction algorithms. Each atlas consists of five individual breast models. The first atlas provides breast vasculature surface models, which are derived from human breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data using image segmentation. A finite element-based method is used to deform the breast vasculature models from their natural shapes to generate the second atlas, compressed breast models. Breast compression is typically done in X-ray mammography but also necessary for some optical mammography systems. Technical validation is presented to demonstrate how the atlases can be used to study the image reconstruction algorithms. Optical measurements are generated numerically with compressed breast models and a predefined configuration of light sources and photodetectors. The simulated data is fed into three standard image reconstruction algorithms to reconstruct optical images of the vasculature, which can then be compared with the ground truth to evaluate their performance.

Effects of extended pharmacological disruption of zebrafish embryonic heart biomechanical environment on cardiac function, morphology, and gene expression.

BACKGROUND: Biomechanical stimuli are known to be important to cardiac development, but the mechanisms are not fully understood. Here, we pharmacologically disrupted the biomechanical environment of wild-type zebrafish embryonic hearts for an extended duration and investigated the consequent effects on cardiac function, morphological development, and gene expression. RESULTS: Myocardial contractility was significantly diminished or abolished in zebrafish embryonic hearts treated for 72 hours from 2 dpf with 2,3-butanedione monoxime (BDM). Image-based flow simulations showed that flow wall shear stresses were abolished or significantly reduced with high oscillatory shear indices. At 5 dpf, after removal of BDM, treated embryonic hearts were maldeveloped, having disrupted cardiac looping, smaller ventricles, and poor cardiac function (lower ejected flow, bulboventricular regurgitation, lower contractility, and slower heart rate). RNA sequencing of cardiomyocytes of treated hearts revealed 922 significantly up-regulated genes and 1,698 significantly down-regulated genes. RNA analysis and subsequent qPCR and histology validation suggested that biomechanical disruption led to an up-regulation of inflammatory and apoptotic genes and down-regulation of ECM remodeling and ECM-receptor interaction genes. Biomechanics disruption also prevented the formation of ventricular trabeculation along with notch1 and erbb4a down-regulation. CONCLUSIONS: Extended disruption of biomechanical stimuli caused maldevelopment, and potential genes responsible for this are identified.

Multifunctional laser speckle imaging.

We have developed a multi-functional laser speckle imaging system, which can be operated in both the surface illumination laser speckle contrast imaging (SI-LSCI) mode and the line scan laser speckle contrast imaging (LS-LSCI) mode. The system has been applied to imaging the chicken embryos to visualize both the blood flow and morphological details of the vasculature. The experimental results demonstrated that LS-LSCI is capable of detecting and quantifying blood flow in blood vessels smaller and deeper than those detectable by conventional SI-LSCI. Furthermore, the line scan mode is also capable of producing depth-resolved absorption-based morphological images of tissue, augmenting flow-based functional images.

4D modelling of fluid mechanics in the zebrafish embryonic heart.

Abnormal blood flow mechanics can result in pathological heart malformation, underlining the importance of understanding embryonic cardiac fluid mechanics. In the current study, we performed image-based computational fluid dynamics simulation of the zebrafish embryonic heart ventricles and characterized flow mechanics, organ dynamics, and energy dynamics in detail. 4D scans of 5 days post-fertilization embryonic hearts with GFP-labelled myocardium were acquired using line-scan focal modulation microscopy. This revealed that the zebrafish hearts exhibited a wave-like contractile/relaxation motion from the inlet to the outlet during both systole and diastole, which we showed to be an energy efficient configuration. No impedance pumping effects of pressure and velocity waves were observed. Due to its tube-like configuration, inflow velocities were higher near the inlet and smaller at the outlet and vice versa for outflow velocities. This resulted in an interesting spatial wall shear stress (WSS) pattern where WSS waveforms near the inlet and those near the outlet were out of phase. There was large spatial variability in WSS magnitudes. Peak WSS was in the range of 47.5-130 dyne/cm2 at the inflow and outflow tracts, but were much smaller, in the range of 4-11 dyne/cm2, in the mid-ventricular segment. Due to very low Reynolds number and the highly viscous environment, intraventricular pressure gradients were high, suggesting substantial energy losses of flow through the heart.

Quadrature demodulation in line-scan focal modulation microscopy for imaging three-dimensional zebrafish neural structure.

Visualizing biological processes in neuroscience requires in vivo functional imaging at single-neuron resolution, high image acquisition speed and strong optical sectioning ability. However, due to light scattering of in tissue, very often conventional wide-field fluorescence microscopes are unable to resolve cells in the presence of a strong out-of-focus background. Line-scan focal modulation microscopy enables high temporal resolution and good optical sectioning ability at the same time. Here we demonstrate a quadrature demodulation method to extract the focal information with an extended frequency bandwidth and therefore higher spatial resolution. The performance of the demodulation scheme in line-scan focal modulation microscope has been evaluated by performing imaging experiments with fluorescence beads and zebrafish neural structure. Reduced background, reduced artifacts and more detailed morphological information are evident in the obtained images.

Three-dimensional cellular imaging in thick biological tissue with confocal detection of one-photon fluorescence in the near-infrared II window.

Fluorescence imaging in the second near-infrared optical window (NIR-II, 900-1700 nm) has become a technique of choice for noninvasive in vivo imaging in recent years. Greater penetration depths with high spatial resolution and low background can be achieved with this NIR-II window, owing to low autofluorescence within this optical range and reduced scattering of long wavelength photons. Here, we present a novel design of confocal laser scanning microscope tailored for imaging in the NIR-II window. We showcase the outstanding penetration depth of our confocal setup with a series of imaging experiments. HeLa cells labeled with PbS quantum dots with a peak emission wavelength of 1276 nm can be visualized through a 3.5-mm-thick layer of scattering medium, which is a 0.8% Lipofundin solution. A commercially available organic dye IR-1061 (emission peak at 1132 nm), in its native form, is used for the first time, as a NIR-II fluorescence label in cellular imaging. Our confocal setup is capable of capturing optically sectioned images of IR-1061 labeled chondrocytes in fixed animal cartilage at a depth up to 800 µm, with a superb spatial resolution of around 2 µm.

Laplace-domain diffuse optical measurement.

Time-domain diffuse optical measurement systems determine depth-resolved absorption changes by using the time of flight distribution of the detected photons. It is well known that certain feature data, such as the Laplace transform of the temporal point spread function, is sufficient for image reconstruction and diffuse optical sensing. Conventional time-domain systems require the acquisition of full temporal profiles of diffusive photons and then numerically compute the feature dataset, for example, Laplace transformed intensities for imaging applications. We have proposed a novel method for directly obtaining the Laplace transform data. Our approach can significantly improve the data acquisition speed for time-domain diffuse optical imaging. We also demonstrated that the use of negative Laplace parameters can provide enhanced sensitivity to perturbations located in deep regions.

Numerical modeling of two-photon focal modulation microscopy with a sinusoidal phase filter.

A spatiotemporal phase modulator (STPM) is theoretically investigated using the vectorial diffraction theory. The STPM is equivalent to a time-dependent phase-only pupil filter that alternates between a homogeneous filter and a stripe-shaped filter with a sinusoidal phase distribution. It is found that two-photon focal modulation microscopy (TPFMM) using this STPM can significantly suppress the background contribution from out-of-focus ballistic excitation and achieve almost the same resolution as two-photon microscopy. The modulation depth is also evaluated and a compromise exists between the signal-to-background ratio and signal-to-noise ratio. The theoretical investigations provide important insights into future implementations of TPFMM and its potential to further extend the penetration depth of nonlinear microscopy in imaging multiple-scattering biological tissues.

Ultrahigh-speed line-scan SD-OCT for four-dimensional in vivo imaging of small animal models.

We report an ultrahigh-speed and high-resolution line-scan spectral-domain optical coherence tomography (SD-OCT) system that integrates a number of mechanisms for improving image quality. The illumination uniformity is significantly improved by the use of a Powell lens; Phase stepping and differential reconstruction are combined to suppress autocorrelation artifacts; Nonlocal means (NLM) is employed to enhance the signal to noise ratio while minimizing motion artifacts. The system is capable of acquiring cross-sectional images at more than 3,500 B-scans per second with sensitivities between 70dB and 90dB. The high B-scan rate enables image post-processing with nonlocal means, an advanced noise reduction algorithm that affords enhanced morphological details and reduced motion artifacts. The achieved axial and lateral resolutions are 2.0 and 6.2 microns, respectively. We have used this system to acquire four-dimensional (three-dimensional space and one-dimensional time) imaging data from live chicken embryos at up to 40 volumes per second. Dynamic cardiac tissue deformation and blood flow could be clearly visualized at high temporal and spatial resolutions, providing valuable information for understanding the mechanical and fluid dynamic properties of the developing cardiac system.

Effects of pupil filter patterns in line-scan focal modulation microscopy.

Line-scan focal modulation microscopy (LSFMM) is an emerging imaging technique that affords high imaging speed and good optical sectioning at the same time. We present a systematic investigation into optimal design of the pupil filter for LSFMM in an attempt to achieve the best performance in terms of spatial resolutions, optical sectioning, and modulation depth. Scalar diffraction theory was used to compute light propagation and distribution in the system and theoretical predictions on system performance, which were then compared with experimental results.

Line-scan focal modulation microscopy.

We report the development of a line-scan focal modulation microscope (LSFMM) that is capable of high-speed image acquisition ( > 40 ?? fps ) with uncompromised optical sectioning capability. The improved background rejection and axial resolution of this imaging modality, enabled by focal modulation, are quantified with three-dimensional imaging data obtained from fluorescent beads. The signal-to-background ratio for the LSFMM system is one- to two-orders of magnitude higher than that for line-scanning confocal systems when imaging deep (up to 100 ?m) into a turbid medium of optical properties similar to biological tissues. The imaging performance of LSFMM, in terms of both spatial and temporal resolutions, is further demonstrated with in vivo imaging experiments with live zebrafish larvae.

Spread spectrum time-resolved diffuse optical measurement system for enhanced sensitivity in detecting human brain activity.

Diffuse optical spectroscopy (DOS) and imaging methods have been widely applied to noninvasive detection of brain activity. We have designed and implemented a low cost, portable, real-time one-channel time-resolved DOS system for neuroscience studies. Phantom experiments were carried out to test the performance of the system. We further conducted preliminary human experiments and demonstrated that enhanced sensitivity in detecting neural activity in the cortex could be achieved by the use of late arriving photons.

Augmented line-scan focal modulation microscopy for multi-dimensional imaging of zebrafish heart in vivo.

Multi-dimensional fluorescence imaging of live animal models demands strong optical sectioning, high spatial resolution, fast image acquisition, and minimal photobleaching. While conventional laser scanning microscopes are capable of deep penetration and sub-cellular resolution, they are generally too slow and causing excessive photobleaching for volumetric or time-lapse imaging. We demonstrate the performance of an augmented line-scan focal modulation microscope (aLSFMM), a high-speed imaging platform that affords above video-rate imaging speed by the use of line scanning. Exceptional background rejection is accomplished by combining a confocal slit with focal modulation. The image quality is further improved by merging the information from simultaneously acquired focal modulation and confocal images. Such a hybrid imaging scheme makes it possible to use very low power excitation light in high-speed imaging, and therefore leads to reduced photobleaching that is desirable for three-dimensional (3D) and four-dimensional (4D) in vivo image acquisition.

A method to study the hemodynamics of chicken embryo's aortic arches using optical coherence tomography.

Congenital cardiovascular defects are the leading cause of birth defect related death. It has been hypothesized that fluid mechanical forces of embryonic blood flow affect cardiovascular development and play a role in congenital malformations. Studies in small animal embryos can improve our understanding of congenital malformations and can lead to better treatment. We present a feasibility study in which high-resolution optical coherence tomography (OCT) and computational fluid dynamics (CFD) are combined to provide quantitative analysis of the embryonic flow mechanics and the associated anatomy in a small animal model.

In vivo simultaneous multispectral fluorescence imaging with spectral multiplexed volume holographic imaging system.

A simultaneous multispectral fluorescence imaging system incorporating multiplexed volume holographic grating (VHG) is developed to acquire multispectral images of an object in one shot. With the multiplexed VHG, the imaging system can provide the distribution and spectral characteristics of multiple fluorophores in the scene. The implementation and performance of the simultaneous multispectral imaging system are presented. Further, the system's capability in simultaneously obtaining multispectral fluorescence measurements is demonstrated with in vivo experiments on a mouse. The demonstrated imaging system has the potential to obtain multispectral images fluorescence simultaneously.

Hybrid wide-field and scanning microscopy for high-speed 3D imaging.

Wide-field optical microscopy is efficient and robust in biological imaging, but it lacks depth sectioning. In contrast, scanning microscopic techniques, such as confocal microscopy and multiphoton microscopy, have been successfully used for three-dimensional (3D) imaging with optical sectioning capability. However, these microscopic techniques are not very suitable for dynamic real-time imaging because they usually take a long time for temporal and spatial scanning. Here, a hybrid imaging technique combining wide-field microscopy and scanning microscopy is proposed to accelerate the image acquisition process while maintaining the 3D optical sectioning capability. The performance was demonstrated by proof-of-concept imaging experiments with fluorescent beads and zebrafish liver.