KEYWORDS: COVID 19, Medical imaging, Algorithm development, Diseases and disorders, Artificial intelligence, Data modeling, Gold, Medicine, Medical research, Evolutionary algorithms
PurposeThe Medical Imaging and Data Resource Center (MIDRC) open data commons was launched to accelerate the development of artificial intelligence (AI) algorithms to help address the COVID-19 pandemic. The purpose of this study was to quantify longitudinal representativeness of the demographic characteristics of the primary MIDRC dataset compared to the United States general population (US Census) and COVID-19 positive case counts from the Centers for Disease Control and Prevention (CDC).ApproachThe Jensen-Shannon distance (JSD), a measure of similarity of two distributions, was used to longitudinally measure the representativeness of the distribution of (1) all unique patients in the MIDRC data to the 2020 US Census and (2) all unique COVID-19 positive patients in the MIDRC data to the case counts reported by the CDC. The distributions were evaluated in the demographic categories of age at index, sex, race, ethnicity, and the combination of race and ethnicity.ResultsRepresentativeness of the MIDRC data by ethnicity and the combination of race and ethnicity was impacted by the percentage of CDC case counts for which this was not reported. The distributions by sex and race have retained their level of representativeness over time.ConclusionThe representativeness of the open medical imaging datasets in the curated public data commons at MIDRC has evolved over time as the number of contributing institutions and overall number of subjects have grown. The use of metrics, such as the JSD support measurement of representativeness, is one step needed for fair and generalizable AI algorithm development.
Spectral Domain Phase Microscopy (SDPM) is a recent extension of Spectral Domain Optical Coherence Tomography
(SDOCT) that exploits the extraordinary phase stability of spectrometer-based systems with common-path geometry to
resolve sub-wavelength displacements within a sample volume. This technique has been implemented for high resolution
axial displacement and velocity measurements in biological samples, but since axial displacement information is
acquired serially, has been unable to measure fast temporal dynamics in extended samples. Depth-Encoded SDPM
(DESDPM) uses multiple sample arms with unevenly spaced common path reference reflectors to multiplex independent
SDPM signals from separate lateral positions on a sample simultaneously using a single interferometer, thus limiting the
time required to detect unique optical events to the integration time of the detector. The minimum measured sample
displacements determined from the standard deviation of the detected phase as a function of time two ideal reflectors
were 407 and 730 pm. Heat-induced expansion in a microscope slide was measured at two sites simultaneously. A 51 ms
delay in 50% rise time of the surface displacement was measured. Further application of this technique to biological
samples could yield insight into temporal dynamics of activation signals.
We present in vivo human total retinal blood flow measurements using Doppler Fourier domain optical coherence tomography (OCT). The scan pattern consisted of two concentric circles around the optic nerve head, transecting all retinal branch arteries and veins. The relative positions of each blood vessel in the two OCT conic cross sections were measured and used to determine the angle between the OCT beam and the vessel. The measured angle and the Doppler shift profile were used to compute blood flow in the blood vessel. The flows in the branch veins was summed to give the total retinal blood flow at one time point. Each measurement of total retinal blood flow was completed within 2 s and averaged. The total retinal venous flow was measured in one eye each of two volunteers. The results were 52.90±2.75 and 45.23±3.18 µl/min, respectively. Volumetric flow rate positively correlated with vessel diameter. This new technique may be useful in the diagnosis and treatment of optic nerve and retinal diseases that are associated with poor blood flow, such as glaucoma and diabetic retinopathy.
There is considerable interest in new methods for the assessment of retinal blood flow for the diagnosis of eye diseases. We present in vivo normal human volumetric retinal flow measurement using Fourier domain Doppler optical coherence tomography. We used a dual-plane scanning pattern to determine the angle between the blood flow and the scanning beam in order to measure total flow velocity. Volumetric flow in each blood vessel around the optic nerve head was integrated in one cardiac cycle in each measurement. Measurements were performed in the right eye of one human subject. The measured venous flow velocity ranged from 16.26 mm/s to 29.7 mm/s. The arterial flow velocity ranged from 38.35 mm/s to 51.13 mm/s. The total retinal venous and arterial flow both added up to approximately 54 µl/min. We believe this is the first demonstration of total retinal blood flow measurement using the OCT technique.
Investigation of the autoregulatory mechanism of human retinal perfusion is conducted with a real-time spectral domain Doppler optical coherence tomography (SDOCT) system. Volumetric, time-sequential, and Doppler flow imaging are performed in the inferior arcade region on normal healthy subjects breathing normal room air and 100% oxygen. The real-time Doppler SDOCT system displays fully processed, high-resolution [512 (axial)×1000 (lateral) pixels] B scans at 17 frames/sec in volumetric and time-sequential imaging modes, and also displays fully processed overlaid color Doppler flow images comprising 512 (axial)×500 (lateral) pixels at 6 frames/sec. Data acquired following 5 min of 100% oxygen inhalation is compared with that acquired 5 min postinhalation for four healthy subjects. The average vessel constriction across the population is −16±26% after oxygen inhalation with a dilation of 36±54% after a return to room air. The flow decreases by −6±20% in response to oxygen and in turn increases by 21±28% as flow returns to normal in response to room air. These trends are in agreement with those previously reported using laser Doppler velocimetry to study retinal vessel autoregulation. Doppler flow repeatability data are presented to address the high standard deviations in the measurements.
KEYWORDS: Optical coherence tomography, Data acquisition, Diagnostics, Volume rendering, 3D image processing, Image segmentation, Retinal scanning, 3D acquisition, Data centers, Signal detection
We report on the development of quantitative, reproducible diagnostic observables for age-related macular degeneration
(AMD) based on high speed spectral domain optical coherence tomography (SDOCT). 3D SDOCT volumetric data sets
(512 x 1000 x 100 voxels) were collected (5.7 seconds acquisition time) in over 50 patients with age-related macular
degeneration and geographic atrophy using a state-of-the-art SDOCT scanner. Commercial and custom software utilities
were used for manual and semi-automated segmentation of photoreceptor layer thickness, total drusen volume, and
geographic atrophy cross-sectional area. In a preliminary test of reproducibility in segmentation of total drusen volume
and geographic atrophy surface area, inter-observer error was less than 5%. Extracted volume and surface area of AMD-related
drusen and geographic atrophy, respectively, may serve as useful observables for tracking disease state that were
not accessible without the rapid 3D volumetric imaging capability unique to retinal SDOCT.
KEYWORDS: Image segmentation, Optical coherence tomography, 3D image processing, Eye, 3D acquisition, Visualization, Retina, 3D metrology, Data acquisition, 3D visualizations
The acquisition speed of current FD-OCT (Fourier Domain - Optical Coherence Tomography) instruments allows rapid
screening of three-dimensional (3D) volumes of human retinas in clinical settings. To take advantage of this ability
requires software used by physicians to be capable of displaying and accessing volumetric data as well as supporting
post processing in order to access important quantitative information such as thickness maps and segmented volumes.
We describe our clinical FD-OCT system used to acquire 3D data from the human retina over the macula and optic
nerve head. B-scans are registered to remove motion artifacts and post-processed with customized 3D visualization and
analysis software. Our analysis software includes standard 3D visualization techniques along with a machine learning
support vector machine (SVM) algorithm that allows a user to semi-automatically segment different retinal structures
and layers. Our program makes possible measurements of the retinal layer thickness as well as volumes of structures of
interest, despite the presence of noise and structural deformations associated with retinal pathology. Our software has
been tested successfully in clinical settings for its efficacy in assessing 3D retinal structures in healthy as well as
diseased cases. Our tool facilitates diagnosis and treatment monitoring of retinal diseases.
Investigation of the autoregulatory mechanism of human retinal perfusion was conducted with a novel real-time spectral domain Doppler optical coherence tomography (SDOCT) system. Volumetric, time-sequential, and Doppler flow imaging was performed in the superior arcade region on normal healthy subjects breathing normal room air and 100% oxygen. The real-time Doppler SDOCT system displays fully processed, high-resolution [512 (axial) x 1000 (lateral) pixels] B-scans at 17 frames/sec in volumetric and time-sequential imaging modes, and also displays fully processed overlaid color Doppler flow images comprising 512 (axial) x 500 (lateral) pixels at 6 frames/sec. OCT fundus images generated from volumetric datasets updated in real time (up to 2 fundus images/sec for 100 x 100 pixel volumes) were used to image and localize retinal vessels for time-sequential and Doppler flow analysis. In preliminary measurements, data acquired following 5 minutes of 100% oxygen inhalation was compared with that acquired 5 minutes post-inhalation. The same arterial segments examined at both time points exhibit constriction in vessel diameter under pure oxygen inhalation of up to 7% and reduction in peak flow velocity as great as 38%, both of which are in good agreement with previous laser Doppler velocimetry studies.
We have combined Fourier-domain optical coherence tomography (OCT) with a closed-loop Adaptive Optics (AO) system. The AO-OCT instrument has been used for in vivo retinal imaging. High-lateral resolution of our AO-OCT system allows visualization of the microscopic retinal structures not accessible by standard OCT instruments.
KEYWORDS: Optical coherence tomography, Image quality, In vivo imaging, Signal to noise ratio, Mirrors, Spectroscopy, Charge-coupled devices, Retina, Eye, Imaging systems
We built a Fourier domain optical coherence tomography (FD-OCT) system using a line scan CCD camera that allows real time data display and acquisition. This instrument is able to produce 2D B-scans as well as 3D data sets with human subjects in vivo in clinical settings. In this paper we analyze the influence of varying exposure times of the CCD detector on image quality. Sensitivity values derived from theoretical predictions have been compared with measurements (obtained with mirrors and neutral density filters placed in both interferometer arms). The results of these experiments, discussion about differences between sensitivity values, potential sources of discrepancies, and recommendations for optimal exposure times will be described in this paper. A short discussion of observed artifacts as well as possible ways to remove them is presented. The influence of relative retinal position with respect to reference mirror position will also be described.
We present a low-cost, high resolution, real-time Spectral Domain Optical Coherence Tomography (SDOCT) system optimized for rapid 3D imaging of the human retina in vivo. Using a source with an 841nm center wavelength and a FWHM bandwidth of 49nm, 6.67 second length bursts of 100 512 x 1000 pixel images were acquired with an integration time of 50 microseconds/line and a frame rate of 16 frames/sec. Three-dimensional data sets comprising up to 4.0mm x 1.2mm x 2.45mm retinal volumes were streamed to hard disk during this brief ocular fixation interval and post-processed to create 3D volumetric images of the optic nerve head and fovea.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.