A dataset of rodent cerebrovasculature from in vivo multiphoton fluorescence microscopy imaging.

We present MiniVess, the first annotated dataset of rodent cerebrovasculature, acquired using two-photon fluorescence microscopy. MiniVess consists of 70 3D image volumes with segmented ground truths. Segmentations were created using traditional image processing operations, a U-Net, and manual proofreading. Code for image preprocessing steps and the U-Net are provided. Supervised machine learning methods have been widely used for automated image processing of biomedical images. While much emphasis has been placed on the development of new network architectures and loss functions, there has been an increased emphasis on the need for publicly available annotated, or segmented, datasets. Annotated datasets are necessary during model training and validation. In particular, datasets that are collected from different labs are necessary to test the generalizability of models. We hope this dataset will be helpful in testing the reliability of machine learning tools for analyzing biomedical images.

Latent Transformer Models for out-of-distribution detection.

Any clinically-deployed image-processing pipeline must be robust to the full range of inputs it may be presented with. One popular approach to this challenge is to develop predictive models that can provide a measure of their uncertainty. Another approach is to use generative modelling to quantify the likelihood of inputs. Inputs with a low enough likelihood are deemed to be out-of-distribution and are not presented to the downstream predictive model. In this work, we evaluate several approaches to segmentation with uncertainty for the task of segmenting bleeds in 3D CT of the head. We show that these models can fail catastrophically when operating in the far out-of-distribution domain, often providing predictions that are both highly confident and wrong. We propose to instead perform out-of-distribution detection using the Latent Transformer Model: a VQ-GAN is used to provide a highly compressed latent representation of the input volume, and a transformer is then used to estimate the likelihood of this compressed representation of the input. We demonstrate this approach can identify images that are both far- and near- out-of-distribution, as well as provide spatial maps that highlight the regions considered to be out-of-distribution. Furthermore, we find a strong relationship between an image's likelihood and the quality of a model's segmentation on it, demonstrating that this approach is viable for filtering out unsuitable images.

Refined flicker photometry technique to measure ocular lens density.

Many physiological and pathological conditions are associated with a change in the crystalline lens transmittance. Estimates of lens opacification, however, generally rely on subjective rather than objective measures in clinical practice. The goal of our study was to develop an improved psychophysical heterochromatic flicker photometry technique combined with existing mathematical models to evaluate the spectral transmittance of the human ocular media noninvasively. Our results show that it is possible to accurately estimate ocular media density in vivo in humans. Potential applications of our approach include basic research and clinical settings on visual and nonimage-forming visual systems.

A Purkinje image-based system for an assessment of the density and transmittance spectra of the human crystalline lens in vivo.

A method for rapid and objective assessment of ocular lens density and transmittance is needed for research and clinical practice. The aim of this study was to determine whether the Purkinje image-based technique can be used for objective and accurate quantification of spectral density and transmittance of ocular media (the mainly crystalline lens) in visible light. Twenty-six individuals (10 young, 9 middle-aged and 7 older individuals) participated in this study. Spectral lens density was evaluated by detecting the intensity of the IVth Purkinje image for different wavelengths. Subsequently, optical density index (ODI), the area under the curve in the lens density spectrum, was calculated and ODIs were compared with clinical lens opacification scales assessed subjectively using a slit lamp. Spectral lens transmittance was estimated from the lens density spectrum. Lens densities were higher in the short wavelength region of the visible spectrum across all age groups. ODI was highly correlated with the clinical opacification scale, while lens transmittance decreased with aging. Our results showed that spectral transmittance of the human crystalline lens can be easily estimated from optical density spectra evaluated objectively and rapidly using the Purkinje image-based technique. Our results provide clinicians and scientists with an accurate, rapid and objective technique for quantification of lens transmittance.

Embedded deep learning in ophthalmology: making ophthalmic imaging smarter.

Deep learning has recently gained high interest in ophthalmology due to its ability to detect clinically significant features for diagnosis and prognosis. Despite these significant advances, little is known about the ability of various deep learning systems to be embedded within ophthalmic imaging devices, allowing automated image acquisition. In this work, we will review the existing and future directions for 'active acquisition'-embedded deep learning, leading to as high-quality images with little intervention by the human operator. In clinical practice, the improved image quality should translate into more robust deep learning-based clinical diagnostics. Embedded deep learning will be enabled by the constantly improving hardware performance with low cost. We will briefly review possible computation methods in larger clinical systems. Briefly, they can be included in a three-layer framework composed of edge, fog, and cloud layers, the former being performed at a device level. Improved egde-layer performance via 'active acquisition' serves as an automatic data curation operator translating to better quality data in electronic health records, as well as on the cloud layer, for improved deep learning-based clinical data mining.

Heterochromatic Flicker Photometry for Objective Lens Density Quantification.

PURPOSE: Although several methods have been proposed to evaluate lens transmittance, to date there is no consensual in vivo approach in clinical practice. The aim of this study was to compare ocular lens density and transmittance measurements obtained by an improved psychophysical scotopic heterochromatic flicker photometry (sHFP) technique to the results obtained by three other measures: a psychophysical threshold technique, a Scheimpflug imaging technique, and a clinical assessment using a validated subjective scale. METHODS: Forty-three subjects (18 young, 9 middle aged, and 16 older) were included in the study. Individual lens densities were measured and transmittance curves were derived from sHFP indexes. Ocular lens densities were compared across methods by using linear regression analysis. RESULTS: The four approaches showed a quadratic increase in lens opacification with age. The sHFP technique revealed that transmittance decreased with age over the entire visual spectrum. This decrease was particularly pronounced between young and older participants in the short (53.03% decrease in the 400-500 nm range) wavelength regions of the light spectrum. Lens density derived from sHFP highly correlated with the values obtained with the other approaches. Compared to other objective measures, sHFP also showed the lowest variability and the best fit with a quadratic trend (r2 = 0.71) of lens density increase as a function of age. CONCLUSIONS: The sHFP technique offers a practical, reliable, and accurate method to measure lens density in vivo and predict lens transmittance over the visible spectrum. An accurate quantification of lens transmittance should be obtained in clinical practice, but also in research in visual and nonvisual photoreception.

Aging of non-visual spectral sensitivity to light in humans: compensatory mechanisms?

The deterioration of sleep in the older population is a prevalent feature that contributes to a decrease in quality of life. Inappropriate entrainment of the circadian clock by light is considered to contribute to the alteration of sleep structure and circadian rhythms in the elderly. The present study investigates the effects of aging on non-visual spectral sensitivity to light and tests the hypothesis that circadian disturbances are related to a decreased light transmittance. In a within-subject design, eight aged and five young subjects were exposed at night to 60 minute monochromatic light stimulations at 9 different wavelengths (420-620 nm). Individual sensitivity spectra were derived from measures of melatonin suppression. Lens density was assessed using a validated psychophysical technique. Although lens transmittance was decreased for short wavelength light in the older participants, melatonin suppression was not reduced. Peak of non-visual sensitivity was, however, shifted to longer wavelengths in the aged participants (494 nm) compared to young (484 nm). Our results indicate that increased lens filtering does not necessarily lead to a decreased non-visual sensitivity to light. The lack of age-related decrease in non-visual sensitivity to light may involve as yet undefined adaptive mechanisms.

An inexpensive Arduino-based LED stimulator system for vision research.

Light emitting diodes (LEDs) are being used increasingly as light sources in life sciences applications such as in vision research, fluorescence microscopy and in brain-computer interfacing. Here we present an inexpensive but effective visual stimulator based on light emitting diodes (LEDs) and open-source Arduino microcontroller prototyping platform. The main design goal of our system was to use off-the-shelf and open-source components as much as possible, and to reduce design complexity allowing use of the system to end-users without advanced electronics skills. The main core of the system is a USB-connected Arduino microcontroller platform designed initially with a specific emphasis on the ease-of-use creating interactive physical computing environments. The pulse-width modulation (PWM) signal of Arduino was used to drive LEDs allowing linear light intensity control. The visual stimulator was demonstrated in applications such as murine pupillometry, rodent models for cognitive research, and heterochromatic flicker photometry in human psychophysics. These examples illustrate some of the possible applications that can be easily implemented and that are advantageous for students, educational purposes and universities with limited resources. The LED stimulator system was developed as an open-source project. Software interface was developed using Python with simplified examples provided for Matlab and LabVIEW. Source code and hardware information are distributed under the GNU General Public Licence (GPL, version 3).

Does pupil constriction under blue and green monochromatic light exposure change with age?

Many nonvisual functions are regulated by light through a photoreceptive system involving melanopsin-expressing retinal ganglion cells that are maximally sensitive to blue light. Several studies have suggested that the ability of light to modulate circadian entrainment and to induce acute effects on melatonin secretion, subjective alertness, and gene expression decreases during aging, particularly for blue light. This could contribute to the documented changes in sleep and circadian regulatory processes with aging. However, age-related modification in the impact of light on steady-state pupil constriction, which regulates the amount of light reaching the retina, is not demonstrated. We measured pupil size in 16 young (22.8±4 years) and 14 older (61±4.4 years) healthy subjects during 45-second exposures to blue (480 nm) and green (550 nm) monochromatic lights at low (7×10(12) photons/cm2/s), medium (3×10(13) photons/cm2/s), and high (10(14) photons/cm2/s) irradiance levels. Results showed that young subjects had consistently larger pupils than older subjects for dark adaptation and during all light exposures. Steady-state pupil constriction was greater under blue than green light exposure in both age groups and increased with increasing irradiance. Surprisingly, when expressed in relation to baseline pupil size, no significant age-related differences were observed in pupil constriction. The observed reduction in pupil size in older individuals, both in darkness and during light exposure, may reduce retinal illumination and consequently affect nonvisual responses to light. The absence of a significant difference between age groups for relative steady-state pupil constriction suggests that other factors such as tonic, sympathetic control of pupil dilation, rather than light sensitivity per se, account for the observed age difference in pupil size regulation. Compared to other nonvisual functions, the light sensitivity of steady-state pupil constriction appears to remain relatively intact and is not profoundly altered by age.