Stereo-seq Image QC
Last updated
Last updated
Image QC serves to evaluate the appropriateness of microscope images obtained from the Stereo-seq experiment for precise automated analysis within the Stereo-seq Analysis Workflow (SAW).
StereoMap's Image QC can be accessed from the Tools page.
The acceptable types of images include nuclei-staining images (e.g., ssDNA or DAPI), a set of nuclei-staining images (e.g. DAPI) and up to 6 immunofluorescence images (IF images), as well as color images (specifically H&E images). The recommended file format for images is TIFF. Additionally, some microscope file formats such as CZI files from ZEISS or image tiles organized in a folder from Motic are also acceptable. If you are unsure whether your imaging system is suitable for the Stereo-seq experiment, please refer to the Microscope Assessment Guideline to assess your imaging system. The guideline can be accessed from STOmics > Documentation.
Nuclei-staining
8 or 16-bit grayscale image stores in a single single-page image file. Highly recommended format.
TIFF file (.tif or .tiff)
CZI file (.czi)
Image tiles
Nuclei-staining + IF
8 or 16-bit grayscale images store in multiple single-page image files.
TIFF file (.tif or .tiff)
Nuclei-staining + IF
8 or 16-bit grayscale images store in a multi-page image file. Only valid for the Zeiss microscope image.
CZI file (.czi)
H&E
24 or 48-bit colored image stored in a single image file.
TIFF file (.tif or .tiff)
CZI file (.czi)
Image tiles
.tif or .tiff file
Any TIFF
.tar.gz QC output
The output .tar.gz compressed file from Image QC
.tif format image tiles
Customized Motic
MGI STOmics Microscope Go Optical
Leica DM6 B
Other format
.czi file from ZEISS Axio Scan.Z1
.czi file from ZEISS Axioscan 7
Visit Image QC Navigation for each staining type to get more examples.
The quality of an image is a direct reflection of both the stability of the microscope and the excellence of the imaging. The basic assessment includes detecting qualified tracklines and evaluating stitching. Depending on the specific scenarios, the image may also be assessed for clarity and calibration confidence. The specific combination and score threshold vary slightly depending on the image type.
The assessments include,
Trackline Detection: The tracklines are fiducial markers etched on the surface of the Stereo-seq chip. They can be detected in the microscope image and the sequencing-based spatial feature expression density heatmap, linking the two data modalities. Reliable detection of tracklines from the image is crucial for aligning the image with the spatial feature expression density map. A detection score combines several factors to evaluate the likelihood of successful image registration. The score takes into account the visibility of tracklines and a sufficient number of neighboring tracklines that qualified for deducing a periodic trackline grid, or trackline template. The inferred grid is the key to adjusting the scale, rotation, and position of the image.
Image Clarity: Image clarity evaluates the likelihood of getting a clear cell segmentation result. The image is split into tiles and classified into good, moderately blurred, severely blurred, slightly overexposed, or severely overexposed by a deep learning model based on a convolutional neural network.
Microscope Stitching Stability: Microscope stitches image FOVs to get a panoramic image that reveals the entire tissue morphology. Stitching errors may happen due to several reasons, vibrations or movements of the microscope stage during image acquisition, external disturbances on the lab bench, or poor stitching algorithm. Stitching errors impact the accuracy and reliability of the resulting stitched image. The visible seams or misaligned features in the overlapped region will lead to misidentified features or boundaries which affect the result of image registration and segmentation. In Image QC, the assessed image is split into FOVs to measure the morphology feature similarity of adjacent overlapping regions.
Image Calibration: In experiments involving the acquisition of multiple images of the same chip, the images necessarily have consistent stitching, registration, and tissue region. Due to the limitation of staining and imaging, the tracklines - critical for accurate image registration - are only visible in the nuclei-staining image. Therefore, the other images must first align with the nuclei-staining image and then be registered with the spatial feature expression density map using the same actions. The similarity and feature offset are computed to evaluate the confidence that the two images are aligned. This assessment is only applied in nuclei-staining + mIF scenarios and is only applied for single-channel images.
Trackline Detection
≥ 60, sufficient amount of qualified tracklines is detected. The image can be automatically registered with the sequencing-based spatial feature expression matrix.
< 60, lack or qualified tracklines to deduce a trackline template or the detected tracklines cannot overlap with the deduced trackline template. You may need to retake an image.
Necessary assessment indicator to pass QC.
Image Clarity
≥ 80, a good image that is highly likely to get accurate cell identification and segmentation.
< 80, does not necessarily mean the cells cannot be identified, but the SAW build-in cell segmentation algorithm may not perform well.
Prefer-to-pass assessment indicator in QC.
Microscope Stitching Stability (Stitching Evaluation)
≥ 60, more than 30% of FOVs exhibit clearly similar morphology features in the adjacent overlapping regions, and the mean feature offset score of these FOVs is greater than 0.7.
< 60, more than 30% of FOVs have blurry features in the adjacent overlapping regions and the mean feature offset score of these FOVs is less than 0.7.
Conditional assessment indicator.
Image Calibration
Pass, the maximum feature offset between nuclei-staining image and another image ≤ 20 pixels and the feature morphology similarity ≥ 1%.
Fail, the maximum feature offset between nuclei-staining image and another image > 20 pixels or the feature morphology similarity < 1%.
Conditional assessment indicator.
*Refer to STOmics Microscope Assessment Guideline - Chapter 4 Microscope Image Assessment - 4.3. Image Examples for more information about image quality.
Image QC interface components are shown here:
Open the image QC tool and drag the image to the window, the Image Information section will be automatically filled. Make sure at least the Chip SN, Operator, Image Path, and Staining Types have been accurately filled. The Run button will be available for clicking, allowing you to start a QC assessment.
Chip SN
Stereo-seq chip serial number. You can find SN from the bottom of the Stereo-seq chip. E.g. S1 (1x1) chip: SS200000135TL_D1, C02533C S0.5 (0.5x0.5) chip: FP200009107_E414, B03210C211 Large chip: SS200000108BR_A3A4, D02070C3D3 Please ensure that the SN and the image correspond accurately.
Operator
User information. Recommend to enter your email address.
Image Path
Path of your image which is prepared to be checked. The path will be auto-filled in once you have dragged and dropped your image into the window. To avoid any foreseeable errors, refrain from using any characters other than English letters, numbers, or underscores.
Staining Types
Image staining type. Valid options are ssDNA, DAPI, DAPI+mIF, and H&E.
Upload
No: do not upload any image.
QC Input Files (Microscope image): upload microscope image in its original format.
QC Output Files (TAR.GZ): upload image that has been checked by image QC.
Select all: upload both input files and output files.
Remark
Any comments to the image or this QC process.
If your image is in TIFF format, you will be prompted to provide additional microscope details. This information is crucial for accurately measuring the image scale and generating the trackline template. If you are unable to retrieve the specific microscope settings, you can use the default value. However, it is strongly recommended to input the precise information whenever possible.
A comprehensive QC process involves two steps, QC Index Evaluation and Image Compression. Additionally, if you choose to Upload your QC files, successful completion of the image uploading becomes an integral part of your QC progression. The progression, result of each QC assessment, and overall QC conclusion along with further analysis suggestions will be displayed on the screen.
Once the QC progression concludes successfully, you’ll have access to the QC result file in your file system in StereoMapWorkspace -> QC folder. The saving path can be changed in Setting.
If your image successfully pass the QC, the QC output .tar.gz can be transferred to the SAW count pipeline through the --image-tar argument. This will enable it to be automatically co-processed with the feature expression matrix. Conversely, if your image does not pass QC but you still want to co-visualize it with the feature density map, you will have to input your image file to the StereoMap Image Processing module and manually manipulate rectify the issue steps before transferring to SAW.
Input example:
.tif or .tiff file
.tar.gz QC output
.tif format image tiles
Motic:
STOmics Go Optical:
Leica:
Other format
Zeiss .czi
The nuclei-staining image undergoes assessment based on two factors: Trackline Detection and Image Clarity. Notably, Trackline Detection serves as the direct indicator of the overall QC result.
Input example:
.tif or .tiff file
.tar.gz QC output
.tif format image tiles
Motic:
STOmics Go Optical:
Other format
Zeiss multi-page .czi
The evaluation of a set of images, which includes a nuclei-staining image and multiple IF images, relies on four factors: Trackline Detection, Image Clarity, Stitching Evaluation, and Image Calibration. It's important to note that indicators other than Image Clarity directly impact the overall QC result.
Input example:
.tif or .tiff file
.tar.gz QC output
.tif format image tiles
Motic:
STOmics Go Spatial:
Other format
The quality of the H&E image is evaluated based on Trackline Detection. Identifying tracklines in an H&E image is more challenging than in a grayscale image, so it is important to ensure that the lines are as visible as possible under the microscope.
Uploading settings setup configurations for transferring image files to HPC or cloud. If you have any queries about which mode is most suitable for your situation, please don’t hesitate to reach out to the FAS.
ALICLOUD
SINGAPORE, AP
AWS
CALIFORNIA, US
RIGA, EU
SINGAPORE, AP
HPC
HPC mode is a configuration used for the laboratory’s internal network, enables the uploading of image files to the local cluster. Please contact FAS to arrange the configuration in advance. Preferreably used for STOmics Tech internal network.
CHONGQING, CN
SHENZHEN, CN
RAYSYNC
CHONGQING, CN
QINGDAO, CN
SHENZHEN, CN
If you have purchased your own Alibaba Cloud or AWS cloud service, you can transfer your image to your personal cloud storage bucket by setting a customized upload path.
Select your cloud service type, which can be either ALICLOUD.CUSTOM or AWS.CUSTOM, and set the region to OWNER. Enter your remote path, keyID, password, and the name of your S3 bucket. If your cloud is Alibaba Cloud, you will need to fill in the domain information as well. Finally, click on Confirm to complete editing.
You will find your customized configuration in the Upload information configuration window.
Upload image to or customized path. Valid options:
See for details.
ALICLOUD mode transfer image through . This mode is ideal for users in regions where STOmics Cloud has been deployed on Alibaba Clous. If there are newly established regions, please reach out to FAS to set up the necessary configurations.
AWS mode allows for transferring image files via . This mode is ideal for users in regions where STOmics Cloud has been deployed on AWS. If there are newly established regions, please reach out to FAS to set up the necessary configurations.
Transfer image file with . By connecting to internet, you can transfer files to the region's cluster without any additional setup.
