Quality Analysis of raw MRI scans using BRAINSImage

Quality Analysis of raw MRI scans using BRAINSImage. Eval This guide is intended to assist the user in the application of BRAINSImage. Eval and to clarify the reviewer’s evaluation of T 1, T 2 and PD images. Kathy Jones and Jacquie Marietta April 2012

A Brief Description of MRI An MR system consists of the following components: 1) a large magnet to generate the magnetic field 2) shim coils to make the magnetic field as homogeneous as possible 3) a radiofrequency (RF) coil to transmit a radio signal into the body part being imaged 4) a receiver coil to detect the returning radio signals 5) gradient coils to provide spatial localization of the signals 6) a computer to reconstruct the radio signals into the final image The signal intensity on the MR image is determined by four basic parameters: 1) proton density 2) T 1 relaxation time 3) T 2 relaxation time 4) flow Proton density is the concentration of protons in the tissue in the form of water and macromolecules (proteins, fat, etc). The T 1 and T 2 relaxation times define the way that the protons revert back to their resting states after the initial RF pulse. The most common effect of flow is loss of signal from rapidly flowing arterial blood. Source: Basic principles of MR imaging John R. Hesselink, MD, FACR http: //spinwarp. ucsd. edu/Neuro. Web/Text/br-100. htm

Examples of MRI scans Scan types: T 1 - longitudinal relaxation time T 2 - transverse relaxation time PD - proton density axial Note: T 2 and PD scans are acquired together coronal sagittal

A Brief Description of Scan Artifacts • An image artifact is a structure not normally present but visible as a result of a limitation or malfunction in the hardware or software of the MRI device, or in other cases a consequence of environmental influences as heat or humidity or it can be caused by the human body (blood flow, implants etc. ). The knowledge of MRI artifacts and noise producing factors is important for continuing maintenance of high image quality. Artifacts may be very noticeable or just a few pixels out of balance but can give confusing artifactual appearances with pathology that may be misdiagnosed. • Changes in patient position, different pulse sequences, metallic artifacts, or other imaging variables can cause image distortions, which can be reduced by the operator; artifacts due to the MR system may require a service engineer. • Many types of artifacts may occur in magnetic resonance imaging. Artifacts in magnetic resonance imaging are typically classified as to their basic principles, e. g. : Physiologic (motion, flow) Hardware (electromagnetic spikes, ringing) Inherent physics (chemical shift, susceptibility, metal) Source: Magnetic Resonance – Technology Information Portal http: //www. mr-tip. com

Raw scan review using BRAINSImage. EVAL Raw scan review is the process by which the quality of scanned images are individually reviewed and rated for either inclusion or exclusion in the Iowa automated neural network (ANN). BRAINSImage. Eval is a standardized system for reviewing raw scans to be used in the Iowa ANN. Ratings are given for T 1, T 2 and PD images based on Signal to Noise (SNR), Contrast to Noise (CNR) and each scan slice receives close visual inspection for lesions, artifacts and other issues in the coronal, axial and sagittal planes. These ratings are combined to create the overall quality assessment (QA) score ranging from 0 -10 and each scan must score a 6 or greater to be included in the ANN. Visual quality analysis is critical to quality control and there are significant reasons why scans may not be included in the ANN and/or data analysis, such as incomplete acquisition or significant scan artifacts. Each Predict-HD scan is reviewed using BRAINSImage. Eval software. NOTE: We provide no radiological or neurological reading of the MRI. The scans are reviewed for motion and other artifacts by research assistants as a matter of basic visual analysis. See the Ethics Core Question document on Share. Point for more information: Home - Imaging Quality Control

Using BRAINSImage. Eval Scripts to run BRAINSImage. Eval: LINUX: /hjohnson/NAMIC/BRAINSImage. Eval-bin/BRAINSImage. Eval MAC: /hjohnson/NAMIC/BRAINSImage. Eval-bin/BRAINSImage. Eval. OSX BRAINSImage. Eval basics: – All scans start out with a zero (0) overall QA rating. – All scans with a rating of 5 or less are excluded from the ANN and MUST include reasons for exclusion. – All scans must include the axial, coronal, and sagittal views. – All scans with lesions or ANY missing brain tissue are rated zero (0) Toolbar buttons: – FINISH EVALUATION – review of current scan is complete. Next scan automatically loads for review. – SKIP THIS IMAGE – skips current scan. Skipped scans reload the next time BRAINSImage. Eval is run. – SHOW COMPLETED EVALS – opens a temporary file that lists the scans currently being reviewed. – EXIT – exit BRAINSImage. Eval. INFO (scan label): The scan label includes the scan identifiers (e. g. PHD, FMRI, or NAMIC); the site number; the scan. ID; the session. ID; and the scan type (T 1, T 2, PD) Example: Scan. Identifier_site_scan. ID_session. ID_scantype PHD_XXXX_XXXXX-3 -T 1

The Image Evaluation form USING THE MOUSE BUTTONS: Click in the image window to activate • Right – zooms image • Middle – moves image • Left – adjusts image contrast BRAINSImage. Eval Toolbar scan_ID type notes here. . . username USING BRAINSImage. Eval: • Review the image using scroll bars • Click to select issues • Scroll to select SNR, CNR and QA scores • A QA score of 0 -5 requires notation USING THE TEXT BOX: • Required – reasons for exclusion • Optional – any notes of interest USING THE KEYBOARD TO SCROLL: Click in the image window to activate w. Left & right arrows – axial w. Up & down arrows – coronal w. Ctrl left & right arrows – sagittal

BRAINSImage. Eval rating scale 0 = bad -Any scan with disqualifying scan issues, e. g. incomplete coverage or lesions. Scan will be excluded from the ANN- Text box notation is required. 1 -5 = poor - A low quality scan and/or major scan issues. Scan will be excluded from the ANN – Text box notation is required. 6 -7 = acceptable - An average scan and/or minor scan issues. 8 -9 = good - A better than average scan with minimal scan issues. 10 = excellent – A high quality scan, the SNR and CNR are scored 10 each and there are no other issues.

BRAINSImage. Eval rating criteria Click on any scan issue for more information and links to several examples: Normal variants – non-lesion variations to anatomy, e. g. atrophy, large ventricles or cerebellar cysts. Lesions – variations to anatomy indicated by circumscribed brain tissue that appear to be lesions. SNR (0 -10) – Signal to Noise Ratio - the overall CLARITY of a scan, e. g. speckling or background noise. CNR (0 -10) – Contrast to Noise Ratio - the overall CONTRAST of the scan, e. g. gray to white tissue differentiation. Full Brain Coverage – complete brain coverage is required – scans are excluded if any portion of a brain tissue is missing. Misalignment – poor head positioning in scanner, e. g. tilting of head; any up & down or side-to-side motion of the head. Swap/Wrap Around – aliasing artifact due to improper field of view (FOV) or inhomogeneity, e. g. wrap or Moiré fringe. Ghosting / Motion – caused by movement and generally observed throughout the scan, e. g. blurriness, bands or rings. Inhomogeneity – caused by non-uniform signal intensity within a scan, e. g. banding or bias. Susceptibility / Metal –magnetic susceptibility resulting in artifact at tissue borders, e. g. metals, sinuses or eyes. Flow artifact – motion artifact due to fluid movement, e. g. blood flow in the temporal lobes. Truncation – abrupt signal intensity change artifact, e. g. ringing or ripple effect, usually observed at the edge of the brain. Overall QA assessment (0 -10) – the overall score is determined by SNR and CNR scores and any additional scan issues.

Normal Variants Unusual or minor non-lesion anatomical changes are considered normal variants and include: • • Cortical atrophy Enlarged Virchow-Robin spaces Enlarged ventricles Cerebellar cysts Normal variants are not generally exclusionary. Scans are rated as any other scan and the variant’s effect reduces the overall score of the scan accordingly. Include notation in the text box. Back to rating criteria

Severe atrophy - overall rating 8

Virchow-Robin - overall rating 8

Enlarged Ventricles in T 1 - overall rating 6

Cerebellar cyst - overall rating 6

Lesions are defined as anatomical abnormalities. They appear as circumscribed regions of tissue within the brain as bright white signal in T 2 and as dark signal in T 1. Lesions are exclusionary and should be marked as zero (0). A note must be included in the text box. Back to rating criteria

Lesion - overall rating 0

SNR = Signal to Noise Ratio is indicated by scan CLARITY. SNR rates the amount of speckling or noise in scan and is significant in evaluating the image quality for use in the ANN. Good SNR has little or no noise and the brain structures are clearly visible.

SNR with noise - rating 6

CNR = Contrast to Noise Ratio is indicated by scan CONTRAST. CNR rates the differentiation in tissue classifications which is vital in segmentation for the ANN. Good CNR has distinct gray/white tissue boundaries and the brain structures are clearly visible. Back to rating criteria

CNR with low contrast - rating 5

Full Brain Coverage Full brain coverage is required for inclusion into the ANN. Coverage is considered incomplete when any portion of the brain tissue is missing. Scroll carefully through each slice in all planes. Incomplete scans are rated zero (0). Must include a notation in text field. Back to rating criteria

Incomplete scan - overall rating 0

Misalignment can be identified by scrolling through the coronal slices and noting the up and down or side-to-side motion of the head. The sagittal view often shows the stair-step image acquisition. Misalignment can also be poor positioning of the head in the scanner, e. g. significant tilting of the head. Misalignment can make the scan unusable for the ANN. Back to rating criteria

Misalignment in PD - overall rating 2

Misalignment in PD - overall rating 1

Swap / Wraparound is indicated by structures appearing to be in the wrong location within the scan, e. g. nose wrapped into posterior brain tissue. Another example is Moiré fringe which appears as light and dark coils or cross-hatches within the brain. Aliasing is not exclusionary unless it interferes with brain tissue and/or structures. Back to rating criteria

Nose wrap - overall rating 0

Moiré fringe - overall rating 7

Moiré fringe - overall rating 5

Ghosting / Motion is a repetitive pattern extending across the entire field of view (FOV). It is visible inside and outside of the skull. This is one of the most common imaging artifacts. It is indicated by waves, blurring, mirror images, bright shadowing, symmetrical rings inside and outside of the brain. The severity determines the usability of the scan. Back to rating criteria

Motion in T 1 - overall rating 5

Motion in T 2 - overall rating 3

Motion in T 1 - overall rating 2

Motion in T 1 - overall rating 1

Motion in T 1 - overall rating 0

Inhomogeneity is inconsistent gray white tissue intensity within a scan. Unlike CNR, inhomogeneity is indicated by bias and banding in portions of the scan, while CNR is the overall contrast throughout the entire image. Look for variations of intensity, bias, which is most easily viewed as light and dark shadowing at the anterior/posterior poles and/or contrasting bands visible in all views. Loss of gray white tissue delineation can result in the exclusion of the scan. Back to rating criteria

Banding - overall rating 2

Banding - overall rating 1

Bias - overall rating 7

Bias - overall rating 6

Bias and banding in PD - overall rating 1

Susceptibility / Metal Susceptibility/metal artifact is indicated by a light or dark signal in regions of adjacent materials or tissues. Examples are: – metal artifact such as dental braces – the sinus bone junction – the tissue between the eyes – zipper artifact which is seen as alternating bright and dark pixels in a line resembling a zipper. The rating depends on the amount of interference within the tissue. Back to rating criteria

Susceptibility artifact in T 1 - overall rating 0

Artifact from sinus/bone junction - overall rating 0

Susceptibility artifact between eyes in T 1 - overall rating 8

Susceptibility artifact between eyes in T 1 - overall rating 7

Susceptibility / Zipper artifact - overall rating 7

Flow Artifact Flow artifact is susceptibility due to blood flow and is usually seen as hatched lines in the temporal lobes. Flow artifacts are infrequent and usually do not effect the rating. Back to rating criteria

Flow artifact in temporal lobes - overall rating 7

Truncation artifacts are rings or waves seen around the outside edge of the brain but unlike ghosting, they diminish as they move toward the center. Truncation does not extend outside the brain. It is infrequent and rarely effects the ratings. Back to rating criteria

T 1 Truncation artifact - overall rating 8

PD Truncation artifact - overall rating 7

Overall Quality Assessment The overall raw scan quality assessment (QA) is based on a combination of the SNR and CNR ratings in addition to the presence or absence of brain coverage, lesions, artifacts, or other scan issues. The ideal scan will rate a score of 10 each on SNR, CNR and overall QA and have no conditions selected in BRAINSImage. Eval. Back to rating criteria