Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V10/00—Arrangements for image or video recognition or understanding
  • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
  • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
  • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
  • A61B3/0016—Operational features thereof
  • A61B3/0025—Operational features thereof characterised by electronic signal processing, e.g. eye models
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
  • A61B3/0016—Operational features thereof
  • A61B3/0041—Operational features thereof characterised by display arrangements
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
  • A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
  • A61B3/102—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for optical coherence tomography [OCT]
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
  • A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
  • A61B3/12—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for looking at the eye fundus, e.g. ophthalmoscopes
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
  • A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
  • A61B3/12—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for looking at the eye fundus, e.g. ophthalmoscopes
  • A61B3/1225—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for looking at the eye fundus, e.g. ophthalmoscopes using coherent radiation
  • A61B3/1233—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for looking at the eye fundus, e.g. ophthalmoscopes using coherent radiation for measuring blood flow, e.g. at the retina
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
  • A61B5/0062—Arrangements for scanning
  • A61B5/0066—Optical coherence imaging
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
  • A61B5/02007—Evaluating blood vessel condition, e.g. elasticity, compliance
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
  • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
  • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
  • A61B5/7207—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal of noise induced by motion artifacts
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
  • A61B5/7235—Details of waveform analysis
  • A61B5/7253—Details of waveform analysis characterised by using transforms
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
  • A61B5/7235—Details of waveform analysis
  • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
  • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
  • A61B5/7271—Specific aspects of physiological measurement analysis
  • A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T5/00—Image enhancement or restoration
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/0002—Inspection of images, e.g. flaw detection
  • G06T7/0012—Biomedical image inspection
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V10/00—Arrangements for image or video recognition or understanding
  • G06V10/20—Image preprocessing
  • G06V10/30—Noise filtering
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V10/00—Arrangements for image or video recognition or understanding
  • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
  • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
  • A61B5/00—Measuring for diagnostic purposes; Identification of persons
  • A61B5/02—Detecting, measuring or recording for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
  • A61B5/026—Measuring blood flow
  • A61B5/0261—Measuring blood flow using optical means, e.g. infrared light
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
  • G06F2218/02—Preprocessing
  • G06F2218/04—Denoising
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10028—Range image; Depth image; 3D point clouds
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10072—Tomographic images
  • G06T2207/10101—Optical tomography; Optical coherence tomography [OCT]
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20172—Image enhancement details
  • G06T2207/20182—Noise reduction or smoothing in the temporal domain; Spatio-temporal filtering
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30004—Biomedical image processing
  • G06T2207/30041—Eye; Retina; Ophthalmic
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30004—Biomedical image processing
  • G06T2207/30101—Blood vessel; Artery; Vein; Vascular
  • G06T2207/30104—Vascular flow; Blood flow; Perfusion
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
  • G06V2201/03—Recognition of patterns in medical or anatomical images
• • G—PHYSICS
  • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
  • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
  • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
  • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

• Embodiments of the present invention relate generally to the field of optical coherence tomography (OCT) angiography and applications thereof, and specifically methods and apparatus for improved processing of OCT angiography (OCTA) data and reducing the effects of various errors caused by, for example, projection artifacts, low OCT signal, and noise.
• OCT optical coherence tomography
• OCTA OCT angiography
• Optical coherence tomography angiography is a non-invasive vascular imaging modality to visualize flow by detecting motion contrast using repeated OCT measurements at the same location.
• OCTA optical coherence tomography angiography
• FA fluorescein angiography
• ICG indocyanine green
• OCTA imaging is injection-free and provides depth- resolved three dimensional (3D) vascular information of the blood flow or vasculature in the tissue such as in an eye.
• 3D three dimensional
• a method of reducing artifacts includes obtaining OCT/OCTA data from an OCT/OCTA imager; preprocessing OCTA/OCT volume data; extracting features from the preprocessed OCTA/OCT volume data; classifying the OCTA/OCT volume data to provide a probability determination data; determining a percentage data from the probability determination data; and reducing artifacts in response to the percentage data.
• Figure 1 illustrates an exemplary image of the OCTA B-scan with projection artifacts.
• Figures 2A through 2D illustrate exemplary images of OCTA imaging of a normal subject.
• Figure 3 shows a block diagram illustrating the steps to reduce projection artifacts in OCTA 3D volume.
• Figure 4 illustrates an exemplary image of the OCTA B-scan at the same location as Fig. 1 after projection artifacts are reduced.
• Figures 5 A through 5D illustrate exemplary images of OCTA imaging of the same normal subject shown in Fig. 2 after the projection artifacts are reduced.
• Figures 6A and 6B illustrate exemplary images of the outer retina of an age-related macular degeneration (AMD) patient with choroidal neovascularization (CNV) before and after projection artifacts are reduced in the OCTA volume.
• AMD age-related macular degeneration
• CNV choroidal neovascularization
• Figure 7 shows a block diagram illustrating the steps to train the classifier used for projection artifacts reduction.
• OCTA imaging detects vessels with blood flow.
• flow and "vessel” are therefore used interchangeably in the following descriptions.
• OCTA employs motion contrast imaging in order to generate images that show flow, in particular blood flow.
• an OCTA imager compares the differences in the backscattered OCT signal intensity between sequential OCT B-scans taken at the same cross-section of the sample in order to construct a map of blood flow.
• the OCT scans can be corrected for eye movement between sequential images.
• both the OCT images and the derivative OCTA images can be provided.
• Projection or decorrelati on-tail artifacts are one of the most important artifacts that could limit the clinical utility and accuracy of OCTA results.
• Current OCTA processing techniques can generate false motion contrast signals in tissue that falls underneath a blood flow region, even when the underlying tissue is static.
• OCTA techniques are based on the principle of obtaining motion contrast, i.e. identifying and quantifying the change in OCT signal at different depths in the tissue. When the light passes through a blood vessel or a flow region, various factors such as forward scattering, refraction, absorption and path length variations cause unpredictable changes to the light field (and signal) at subsequent depths.
• the backscattered light (and hence the signal) that comes from underneath a region of flow inherits the changes in light field and signal from above, and hence may show a false motion contrast signal, depending on the level of backscattered light and change imparted by the disturbance above. It is very difficult to quantify or correct these changes as they are variable in nature and change in each
• Figure 1 illustrates an exemplary image of the OCTA B-scan illustrating projection artifacts.
• some projection artifacts are illustrated by arrows.
• Figure 1 further shows the internal limiting membrane (ILM) and retinal pigment epithelial (RPE).
• Figure 1 shows OCTA signal in the human retinal layers.
• the arrows indicate the projection artifacts at different retinal levels, whereas the true location of the blood vessels is in the retina above. Hence, any quantitative analysis that occurs without removing the projection artifacts will be misleading, inaccurate, or sub-optimal at best.
• Shadowing artifacts occur when the OCT signal is attenuated behind an absorbing of scattering opacity or obstruction. No or low OCT signal results in no or low OCTA signal. These artifacts can be due to the pathologies of patients such as epi-retinal membranes (floaters) and cataracts. The artifacts can also be due to strong light absorption in the upper tissue layers. Some imaging and processing techniques may be applied to alleviate the shadowing effect. Subsequent image processing and analysis for OCTA can be adjusted accordingly to offset the shadowing effect.
• OCTA noise Another artifact is noise.
• System noise and fluctuations in OCT incident light intensity can result in high OCTA signal even at locations of static tissue with no flow.
• OCTA noise, or false-positive flow can be visually identified by its short segment and isolation from neighboring structured vessels.
• the presence of noise affects subsequent quantification and visualization of small capillaries.
• OCTA volume data may consist of artificial signals that are not related to flow.
• the errors in the OCTA data caused by factors including projection artifacts, shadowing, and noise can be detected and reduced through several methods and techniques according to some embodiments and discussed in this disclosure. This reduction in the artifacts can result in improving the image quality of retinal microvasculature visualization and accuracy of the subsequent quantitative measurements for blood flow.
• the methods used to reduce OCTA artifacts can be generalized to process both OCT and OCTA 3D volume, 2D plane (B-scan), and ID line (A-line) data.
• the artifacts-reduced volume can be used for true 3D visualization.
• the artifacts-reduced volume can be used to generate 2D en face projection images.
• the methods to generate en face images have been disclosed in previous applications. See, e.g., John Davis et al. "Enhanced imaging for optical coherence tomography," US Patent US 8,781,214 B2, July 2014, which is herein incorporated by reference in its entirety.
• the OCTA data can be visualized in 3D and/or 2D by using a different color scheme for pre-processed original signals and artifact signals. For example, voxels/pixels with true signals can be color-coded in grayscale, while projection artifacts color- coded in red, and shadowing artifacts in blue.
• vascular parameters can be calculated from the artifacts-reduced OCTA volume.
• quantitative measurements can be calculated with 3D volume- based parameters and/or 2D en face image-based parameters.
• the parameters include, but are not limited to, flow volume/area, non-flow volume/area, flow density (volume/area/length density), vessel caliber, vessel branching, and tortuosity.
• Figures 1 and 2 A through 2D illustrate projection artifacts in a normal subject with no retinal pathologies based on clinical evaluation, as demonstrated by the B-scan (Fig. 1) and en face (Figs. 2A through 2D) images.
• Figures 2A through 2D illustrate exemplary images of OCTA imaging of a normal subject, with Figure 2 A illustrating en face images of four retinal layers, superficial capillary plexus, with Figure 2B illustrating the deep capillary plexus, with Figure 2C illustrating the outer retina, and with Figure 2D illustrating the choriocapillaris.
• Figures 2A through 2D have been generated from the pre-processed OCTA volume before projection artifacts are reduced.
• the projection artifacts appear at different retinal layers, as indicated by the arrows in Figure 1.
• the projection artifacts coming from the superficial capillary plexus (Figure 2 A) are most noticeable, causing false OCTA signals with similar vascular pattern in the deep capillary plexus ( Figure 2B), outer retina ( Figure 2C), and choriocapillaris (Figure 2D) layers, where no capillaries actually exist.
• FIG. 3 illustrates an exemplary flow diagram demonstrating the steps to reduce artifacts in an OCTA 3D volume.
• An OCTA imager (block 301) generates OCTA volume from OCT data using methods described in previously filed applications. See, e.g., Yali Jia et al. "Split-spectrum amplitude-decorrelation angiography with optical coherence tomography," Optics Express, Feb 2012, which is herein incorporated by reference in its entirety.
• an OCT imager can also be used to provide the structural OCT volume for additional information.
• the OCTA and OCT imager can also be combined to a single
• the OCTA volume and OCT volume data 302 is first passed to an optional preprocessing processer 303.
• the pre-processing processer 303 first detects regions with OCT or OCTA signals above background noise. Background regions can be excluded in the later processing steps to speed up the processing time.
• landmarks are detected along each OCT/OCTA A-line (depth-direction). These landmarks may include peaks and valleys along the ID A-line signal profile, and are often associated with retinal layer boundaries. For example, inner limiting membrane (ILM), junction of inner and outer photoreceptor segments (IS/OS), and retinal pigment epithelium (RPE) usually have stronger OCT intensities and appear as peak points along OCT A-lines.
• ILM inner limiting membrane
• IS/OS junction of inner and outer photoreceptor segments
• RPE retinal pigment epithelium
• the locations or depths of these landmarks can be further refined by averaging over neighboring landmarks (across A-lines and across B-scans). Next, flattening is performed to align all A-scans to a chosen landmark in depth. This is a common step performed for retina segmentation and has been disclosed previously. See, e.g., Mona K. Garvin et al. "Automated 3-D Intraretinal Layer Segmentation of Macular Spectral -Domain Optical
• the OCTA and OCT volume 302 are passed to a feature extraction processer 304. If the optional pre-processing processer 303 is applied, the pre-processed OCTA and OCT volume, along with outputs from the preprocessing processer 303 (for example, detected landmarks) are passed to the feature extraction processer 304.
• Feature-extraction processer 304 extracts features for each base unit.
• the base unit can be one single voxel or a localized region formed by a small number of voxels.
• Feature extraction involving spatial location or depth of the current base unit can include, for example, distance to landmarks (measured in pixels or in microns). Such extraction may also include relative distance (RD) to landmarks, for example, the relative distance from the current base unit (z current ) to landmark A (z A ) can be computed by normalizing with the distance between landmark A and B (z B ). This can be given by the following relation:
• RDA( z ) ⁇ z current ⁇ Z A ⁇ / ⁇ Z A ⁇ Z B ⁇ -
• Feature extraction involving pre-processed OCT and OCTA intensity can include the OCT intensity of the current base unit and the OCTA intensity of the current base unit. Furtherm ore, derivatives (1 st , 2 nd , . . . ) of OCT intensity in each x-, y-, z-direction from the current base uni t and derivatives (1 st , 2 nd , . . . ) of OCTA intensity in each x-, y-, z-direction from the current base unit can be included. Furthermore, intensities and derivatives neighboring base units can be used The kernel size of the neighboring base units to be included as features can be fixed.
• the surrounding 26 voxels in a 3 x3 x3 kernel can be defined as neighbors.
• the kernel size can also be dynamically determined. For example, a bigger kernel s ize can be assigned to a voxel with a higher OCTA intensity.
• Such features may also include a c orresponding OCT intensity at the same depth location where maximum OCTA intensity along A -line is detected.
• Yet another example of these features includes One-dimensional (ID) derivativ e oi OCTA max (z) .
• Feature extraction may also include information related to vessel caliber.
• Such features include the distance to a closet base unit with half the OCTA intensity of the current base unit in the x-direction, the distance to the closet base unit with half the OCTA intensity of the current base unit in the y-direction, the distance to the closet base unit with half the OCTA intensity of the current base unit in the +z-direction, or the distance to the closet base unit with half the OCTA intensity of the current base unit in -z-direction.
• the extracted features are passed to a classifier (block 305).
• the classifier is trained with a sufficiently large dataset where each OCTA voxels are manually labeled by human experts to indicate the presence of different types of artifacts, including projection artifacts. The details of the how the classifier can be trained is described in the Training classifier section and Figure 7.
• the classifier can also be designed with observational criteria.
• the classifier then returns the probability or score of each base unit belonging to one of the classification categories. For example, three categories can be used: a purely true flow signal, a purely artifact signal, and a mixture of both true flow and artifact.
• the classifier can return a hard classification which predicts which categories the base unit belongs to, without providing the probability.
• the probability volume or categorical results provided by the classifier is passed to a transform processer (block 306).
• the processer transforms the probability or categorical results to the percentage of true signal in each base unit.
• the transform function can be a linear transformation determined empirically by phantom studies or by optimizing human retinal scan data to meet clinical understanding. For example,
• Percentage ⁇ wi ⁇ Prob tme + w 2 ⁇ Probe d + w 0 , where Percentage ⁇ is the percentage of true signal in the base unit, Prob s and Prob ⁇ is the probability of belonging to a purely true flow signal group and the probability of belonging to a mixed signal group, respectively.
• the parameters w 0 , wi, and w 2 are the linear weighting factors, which may be determined empirically.
• OC73 ⁇ 4 post (x,y,z) OCTA (x,y,z) ⁇ Percentage ⁇ (x,y,z), and the post-processed artifacts-reduced OCTA volume (OCTA post ) is obtained.
• the artifacts- reduced OCTA data can then be utilized for display (block 308) including but not limited to 3D visualization with volume rendering, 2D visualization of en face projection images and B-scans.
• the artifacts-reduced OCTA data can also be used for further analysis (block 309) to calculate flow or vasculature-related quantitative parameters.
• Figure 4 illustrates an exemplary image of the OCTA B-scan at the same location as Fig 1 after projection artifacts are reduced.
• the arrows indicate a few locations where the projection artifacts are reduced after processing.
• Elongated inner retinal vessels which appear in the pre- processed B-scan ( Figure 1) are shortened. This circular shape of vessels is more consistent with their physical dimensions. Projection artifacts at the IS/OS and RPE layers are also significantly reduced.
• Figures 5 A through 5D illustrate exemplary images of OCTA imaging of the same normal subject shown in Figures 2 A through 2D after the projection artifacts are reduced.
• the four en face images include superficial capillary plexus (Figure 5A), deep capillary plexus (Figure 5B), outer retina (Figure 5C), and choriocapillaris ( Figure 5D).
• Figures 5 A through 5D show the post-processed en face images as compared to the pre-processed en face images in Figures 2A through 2D.
• the duplicated vascular networks are removed from the bottom layers, while the remaining networks are preserved and well-connected.
• Figures 6A and 6B illustrates exemplary images of the outer retina of an AMD patient with CNV before ( Figure 6 A) and after ( Figure 6B) projection artifacts are reduced in the OCTA volume. After processing, the projection artifacts are reduced and the CNV network is better visualized (Fig. 6B). The CNV boundaries are also easier to outline which allows more reliable patient follow-up to assess treatment outcome.
• Figure 7 shows a block diagram illustrating the steps to train the classifier used for projection artifacts reduction.
• the classifier used in the projection artifacts reduction process illustrated as block 305, can be pre-trained with a sufficiently large amount of data.
• Figure 7 is an exemplary flow diagram demonstrating the steps to train the classifier.
• First a training dataset with co-acquired OCT and OCTA volume data from subjects with varying ages, genders and retinal pathologies are collected by an OCT/OCTA imager in block 701.
• A-lines are randomly selected from the OCT/OCTA volume for normal subjects, and randomly selected within pathological areas in scans of pathological patients.
• human experts grade every base unit of these A-lines in block 702.
• Each base unit is labeled with a category.
• the categories can include pure flow signal, pure projection artifacts signal, mixed signal, noise, and unknown signal.
• a subset of dataset is used as testing dataset in block 707 and not used during the training process.
• the OCT and OCTA volume data goes through the pre-processing and feature extraction step in block 703 as described in the previous sections.
• the volume data, features and the human graded label are then passed to a classifier in block 704.
• the machine learning model for example, can be based on logistic regression, ensemble models such as random forest, naive bayes, support vector machine, or combinations of different models.
• the training error is calculated during the training process in block 705.
• the testing dataset (block 707) is inputted to the classifier and the testing error is calculated in block 706.
• the training error (block 705) and testing error (block 706) are then used to refine the classifier in block 708. During this step, the parameters and features in the classifier are refined to minimize while balancing the error from the training dataset and from the testing dataset.
• the method described herein is applied to reduce projection artifacts in OCTA volume, but other artifacts such as noise and shadowing artifacts, can also be reduced through the same processing.
• the method can also be applied to detect artifacts in OCT volume, such as shadowing artifacts.

Landscapes

• Health & Medical Sciences (AREA)
• Life Sciences & Earth Sciences (AREA)
• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Health & Medical Sciences (AREA)
• Medical Informatics (AREA)
• Public Health (AREA)
• Veterinary Medicine (AREA)
• Biophysics (AREA)
• Biomedical Technology (AREA)
• Heart & Thoracic Surgery (AREA)
• Molecular Biology (AREA)
• Surgery (AREA)
• Animal Behavior & Ethology (AREA)
• Artificial Intelligence (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Signal Processing (AREA)
• Pathology (AREA)
• Theoretical Computer Science (AREA)
• Physiology (AREA)
• Psychiatry (AREA)
• General Physics & Mathematics (AREA)
• Ophthalmology & Optometry (AREA)
• Evolutionary Computation (AREA)
• Multimedia (AREA)
• Radiology & Medical Imaging (AREA)
• Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
• Software Systems (AREA)
• Databases & Information Systems (AREA)
• Computing Systems (AREA)
• Fuzzy Systems (AREA)
• Mathematical Physics (AREA)
• Cardiology (AREA)
• Vascular Medicine (AREA)
• Quality & Reliability (AREA)
• Hematology (AREA)
• Eye Examination Apparatus (AREA)
• Investigating Or Analysing Materials By Optical Means (AREA)
• Image Analysis (AREA)

Priority Applications (6)

Applications Claiming Priority (4)

Publications (1)

Family

ID=59625416

Family Applications (1)

Country Status (4)

Cited By (17)

Families Citing this family (18)

Citations (3)

Family Cites Families (5)

• 2017
  • 2017-02-17 US US15/436,704 patent/US10194866B2/en active Active
  • 2017-02-17 WO PCT/US2017/018521 patent/WO2017143300A1/en not_active Ceased
  • 2017-02-17 CA CA3014998A patent/CA3014998C/en active Active
  • 2017-02-17 JP JP2018543677A patent/JP7193343B2/ja active Active

Patent Citations (3)

Also Published As

Similar Documents

Legal Events