IMAGE PROCESSING PRACTICALS

Transcription

1 EPFL PTBIOP IMAGE PROCESSING PRACTICALS

2 ACKNOWLEDGEMENTS This presentation and the exercises are based on the script CMCI Image processing & Analysis Course Series I which was kindly provided by Kota Miura (CMCI, EMBL Heidelberg).

3 CONVERTING AN IMAGE TO A TEXT FILE Goal: Understand that images are simply matrices of numbers. 1. Make a new image using File New Image And these settings 2. Click OK 3. Draw on it using the Pen tool 4. Use File Save As Text Image 5. Open image in Explorer

4 CONVERTING IMAGE DEPTHS (8,16,32 BITS) Goal: Intelligently using proper bit depth and changing it with Fiji / ImageJ 1. Open a 16-bit image from sample File Open m51.tif 2. Choose Line selection tool and draw a line 3. Look at the line profile using Analyze Plot Profile 4. Convert the image to 8-bit First check Edit Options Conversions Scale when converting : OFF Go to Image Type 8-bit 5. Look at the new line profile. 6. Save your ROI Analyze Tools ROI Manager : Click Add 7. Re-open the image m51.tif 8. Set Scale when converting to ON 9. Convert the image to 8-bit again 10. Load your ROI by clicking on its name 11. Use: Analyze Plot Profile

5 CONVERTING IMAGE DEPTHS (8,16,32 BITS) 16-bit 8-bit no scaling 8-bit scaling

6 WORKING WITH RGB IMAGES Goal: Understanding and managing multiple channels on an image 1. Open an RGB image File Open hela-cells.tif 2. Split the channels Image Color Split Channels 3. Merge channels Image Color Merge Channels (Create Composite: OFF, Keep Source Images: ON 4. Try again with a different color configuration 5. Working with separate channels: Use Create Composite: ON when merging OR Use: Image Color Channel Tools More Make composite. 6. Setting the mode to Composite allows to view any channel combination 7. Set the mode to Color and draw a Rectangular ROI 8. Select Edit Clear and return to Composite

7 IMAGE MATH: ADD, SUBTRACT, MULTIPLY AND DIVIDE Goal: Understand add, subtract, multiply and divide from a digital image point of view 1. Create a new 8-bit image and draw on it Put the pointer on the image and look at the values. 2. Add 10 to the image: Process Math Add 3. Place cursor again and notice the new values. What happens to values already at 255? 4. Create a new 8-bit image Add 100 Process Math Add Divide by 3 Process Math Divide 5. Look at the cursor values. 8-bits and 16-bits are integer values for images 6. Try 4. on a 32-bit image NOTE: Although the color picker offers you colors, images in 8-bit, 16-bit and 32-bit depth are grayscale images and color will not show up on them. The only way to see grayscale in color is by using lookup tables Drawing tool Color picker Foreground color Background color

8 IMAGE MATH: IMAGE TO IMAGE Goal: Understanding how to make calculations on images using other images and its uses 1. Open Fluorescent-cells.tif 2. Split the channels Image Color Split Channels 3. Subtract the DAPI channel to the tubulin channel using Process Image Calculator And selecting the right source and destination

9 LOOKUP TABLES (LUT) Goal: Understanding how pixel values are represented in color and how to make use of it 1. Open Cell_Colony.tif File Open Cell_Colony.tif 2. Convert it to 8-bit (See what happens if you don t) 3. Show the Lookup Table (LUT) Image Color Show LUT 4. Change the LUT to spectrum Image Lookup Tables Spectrum 5. Look at the new LUT as in 2.

10 FILE FORMATS, HEADERS, MULTIDIMENSIONAL IMAGES Goal: Knowing what a header is and why avoid lossy formats 1. Open Boats.tif and confocal_series.tif 2. Use Image Show Info on both 3. Use Image Properties Show Info contains the data from the header which can be used by ImageJ to get the image s calibration, i.e. how much does one pixel measure in microns/mm/meters, etc 1. Open Hela-cells.tif (RGB) 2. Use File Save As jpeg 3. Open the jpeg file 4. Subtract it to the original using a 32-bit result The difference between the two images are the artifacts introduced by lossy compression. JPEG and GIF should only be used for qualitative visualization but NEVER for quantitative analysis.

11 Original file (800%) JPEG file (800%) Differences (800%) 1. With the hela-cells (RGB) tiff file, the jpeg file and the subtraction result open, do: Image Stacks Images to Stack 2. Navigate through the stack to see the differences up close.

12 RESIZING IMAGES (SHRINKING AND ENLARGING) Goal: Understand the difference between zooming and resizing, interpolation 1. Open 4pixels.tif 2. Zoom in at maximum 3. Duplicate the image: Image Duplicate 4. Select Image Adjust Size Width: 4, No Interpolation (200%) 5. Duplicate the original again and adjust the size to a width of 3 (150%) 6. Repeat 3-5 with interpolation. Scaling modifies the values of the pixels of the image, so you should be careful when scaling up as much as when scaling down. Interpolation methods can alter scientific results depending on the method.

13 INTENSITY AND HISTOGRAM Goal: Understand the relation between the histogram and the pixel values on the image 1. Open Cell-Colony.tif 2. Use Analyze Histogram 3. Hover the mouse over the histogram plot to see the pixel counts at a given position. 4. Now hover over the image and look at the black dots 5. What is the range of pixel values which correspond to the spot signal? Cursor position Pixel intensity values # Pixels Pixel position Pixel intensity value

14 REGIONS OF INTEREST (ROI) Goal: Knowing how to use, save, and recall ROIs Exercises: 1. CROP: Open any image, select a rectangular ROI, then Image Crop 2. MASKING Open any image, select a rectangular ROI, then Edit Clear, then, Edit Clear Outside Finally Edit Fill. See what happens at each step The same operation works with Edit Selection Create Mask 3. INVERT ROI Open any image, make a rectangular selection, then Edit Selection Make Inverse 4. REDIRECTING ROI Open 2 images, make a selection on one image On the second image, do Edit Selection Restore Selection 5. ROI MANAGER Open any image, and select a ROI Use Analysis Tools Roi Manager (or press t ) and click Add You can save ROIs to a file and restore them later when you reopen your image.

15 INTENSITY MEASUREMENTS Goal: Be able to acquire and understand statistical information of pixels inside a ROI 1. Open Actin.tif, split the channels and keep only the red channel (Actin) 2. Select the Polygon tool and draw a ROI around a cell. 3. Use Analyze Set Measurements and select only Area, Integrated Density and Mean Gray Value 4. Do Analyze Measure (or CTRL+M) to see the results Is that the actual intensity of the actin in the cell? 5. Measure the background intensity using a ROI

16 CONTRAST ENHANCEMENT Goal: Be able to adjust contrast, understand the dangers and pitfalls (Saturation) 1. Open Gel.tif, and do Image Adjust Brightness/Contrast 2. The diagonal line is the LUT (Grays here) The y axis of the line is the brightness on the screen. 3. Change the sliding bars and look at the result on the image. 4. When you push Apply, the histogram changes: The pixel values were altered! Undo: Edit Undo or File Revert 5. The same can be done with RGB images using: Image Adjust Color Balance Try this now on hela-cells(rgb).tif Original Settings What is wrong here? After Apply

17 SPATIAL FILTERING: CONVOLUTION Goal: Understand the bases of convolution and how it is done on images 1. Open neuron_bit.tif Zoom in till you see pixels 2. Launch Process Filter Convolve Match the kernel to the examples and use preview 3. What happens as the kernel gets larger? 4. Try the non-linear filter Median Process Filter Median 3x3 5x5 9x9 Sharpen then Edge Detector Gaussian Blur Sigma = 2 (5x5) Linear filters are simply convolutions of the image values with particular matrices

18 MORPHOLOGICAL IMAGE PROCESSING Goal: Understand and use morphological operations to extract features from an image Open fingerprint.tif and text1.tif 1. Launch Process Binary Options Play with erosion and dilation operations with different iterations using preview 2. Use openings and closings on the image to see how they affect it. 3. Can you find the best parameters to recover the text?

19 FILL HOLES, SKELETONIZE, OUTLINE Goal: Understand and use morphological operations to extract features from an image 1. Open fingerprint.tif, text1.tif and text2.tif 2. Use openings and closings to remove unwanted artifacts if any. 3. Use Outline, Fill Holes and Skeletonize on the images and see the results Closing Outlines Fill Holes Skeletonize Iterations: 2 Count: 3

20 MORPHOLOGY ON GRAY LEVEL IMAGES Goal: Understand and use morphological operations on grey level images 1. Open rice.tif 2. Duplicate the image using Image Duplicate (Shortcut CTRL+SHIFT+D) and name it Background 3. Launch Process Filters Minimum Find radius for which rice grains disappear 4. Use this image for background subtraction using the Image Calculator - After Adjusting B&C =

21 BACKGROUND SUBTRACTION Goal: Be able to use background subtraction methods and know their limitations 1. Open rice.tif 2. Launch Process Subtract Background With different sizes for Rolling Ball Radius 3. Draw a line ROI along the image and use: Analyze Plot Profile for each radius you used. The rolling ball radius should be at least as large as the radius of the largest object in the image that is not part of the BG. What happens if it s too small or too big? Too big Original Good radius

22 SEGMENTATION: THRESHOLDING Goal: Be able to use threshold based image segmentation 1. Open Gel.tif 2. Go to the Adjust Brightness/Contrast menu first Set minimum and maximum to the same value and move brightness and look at the change. Any pixel above the vertical bar is white, any pixel below is black. 3. Now use File Revert and do Image Adjust Threshold Red: White pixels Play with the settings and the auto threshold methods to see the differences All methods explained at: Black White

23 SEGMENTATION: EDGE DETECTION Goal: Understand how to detect the edges of objects within an image 1. Open Rice.tif and transform them to 32-bit 2. Correct for background using the rolling ball algorithm 3. Duplicate the image twice 4. On one duplicate use the following filter: On the other duplicate, use: Now get the root square of the sum of the squares of the images Process Math Square, then Image Calculator (add) then Process Math Square Root 7. Adjust the Brightness to see the result 8. Merge the resulting image using Image Color Merge Channels Merged result

24 SEGMENTATION: MORPHOLOGICAL WATERSHED Goal: Understand and perform segmentation to separate touching objects 1. Open Blood.tif 2. Adjust the background, watch the size of the rolling ball 3. Threshold the image 4. Use Process Binary Fill Holes 5. Use Process Binary Watershed Look at the results. Watershed is powerful but cannot distinguish objects too clumped together Also, RBCs at the edges were not filled, so the watershed failed when trying to segment them

25 SEGMENTATION: PARTICLE ANALYSIS, SHAPES Goal: Be able to use the Particle Analyzer and understand the output 1. Open Blood.tif 2. Duplicate it (We ll need it later) 3. Adjust the background, watch the size of the rolling ball 4. Threshold the image. 5. Use Process Binary Fill Holes 6. Use Process Binary Watershed 7. Change Set Measurements to include the Shape Descriptors. 8. Go to Analyze Analyze Particles With the settings on the right. 9. Look at the results table. 10. Change the settings of Size to 7000 and use Show: Overlay Outlines 11. Play with the settings to filter your particles.

26 SEGMENTATION: PARTICLE ANALYSIS: INTENSITIES Goal: Use segmentation data to get information on intensities using complex masks 1. Open Nuclei.tif and Duplicate it 2. Do a background correction 3. Threshold the image 4. Use binary operations as necessary (Closing, Fill Holes, Watershed). 5. Run Analyze Particles and select the options on the right. 6. Now select your original image and the ROI manager 7. Select all the objects in the ROI manager and click Measure 8. All the intensity measurements are displayed in the Results table.

27 BATCH PROCESSING FILES Goal: Understand macros and be able to make your own macros with Macro Recorder 1. Open??? 2. Launch Plugins Macros Record 3. Select the image window and do: Process Subtract Background 4. Look at the Macro Recorder window: run("subtract Background...", "rolling=40 light"); Is a macro command 5. Copy it and then open Process Batch Macro 6. Paste the macro code on the new window 7. Set the input and output folders 8. Click Process

28 MACROS, FUNCTIONS Goal: Know how to use simple variables and functions 1. Open a file from Sequence folder (name ends in cherry ) 2. Launch Plugins Macros Record 3. Select the image window and duplicate the image rename it Mask 4. Then Threshold it using the automatic Triangle method 5. Dilate it with 5 iterations and a count of 2 6. Run a watershed 7. Run analyze particles 8. Re-select your original image 9. Look at the recorder window, delete any superfluous lines and name it Sample Macro.ijm choose Create

29 MACROS, FUNCTIONS Goal: Know how to use simple variables and functions Exercise (cont.): 1. We need to make sure that the macro works on the selected image. Add the lines on top: roimanager("reset"); name=gettitle() 2. After the run( Analyze ) line, add: selectwindow(name); n = roimanager("count"); for (i=1; i<n; i++) { roimanager("select",i); roimanager("measure"); } 3. Save the macro, open another image from the Sequence and run it

30 TIME SERIES: IMPORT Goal: Being able to import an image sequence 1. Use File Import Image Sequence And open the first file in the folder Sequence 2. Look at the options presented on the next dialog Under File name contains: write cherry 3. Play the sequence you just opened and go to Image Stacks Tools Animation Options And change the animation speed. Selecting only particular images within the folder Very useful in case of very large files (from ~1Gb)

31 TIME SERIES ANALYSIS: FRAME DIFFERENCES Goal: Perform Image Difference Analysis or Image Subtraction Analysis 1. Open the image sequence again 2. Duplicate the entire stack 3. Delete the first frame from the original stack Image Stacks Delete Slice 4. Delete the last frame from the duplicated stack 5. Subtract the original stack from the duplicated stack using the Image Calculator Using Create New Window and 32-bit Result 6. Examine the result. White: Initial Position Black: Position change in next frame 7. Gray: No change NOTE: This is a powerful method to quickly analyze the differences between subsequent time frames.

32 TIME SERIES ANALYSIS: PROJECTIONS Goal: Understand the use of the different projection methods for visualizing time series 1. Open: Speckles.tif 2. Create a ROI around a single cell and Duplicate it 3. Use Edit Clear Outside 4. Use Image Stacks Z-Project 5. Perform different projections and compare the results Original Mean (After B&C Adjustment) NOTE: This tool allows for a easy representation of the trajectories of our object of interest provided they are not too densely packed. Max Intensity Standard Deviation

33 TIME SERIES ANALYSIS: MEASURING INTENSITY DYNAMICS Goal: Be able to measure variations in intensity in time along a ROI 1. Open: FRAP1.tif 2. Create a ROI of of the area where bleaching takes place. 3. Use Image Stacks Plot Z-Axis Profile 4. Measure the baseline intensity of an area of background. 5. By clicking on List You can recover the xy coordinate of the curve to copy to your favorite Spreadsheet software

34 TIME SERIES ANALYSIS: KYMOGRAPHS Goal: Be able to produce a kymograph and generate a velocity plot 1. Open Speckles.tiff 2. Do a Z-Projection 3. Using the Segmented Line ROI (Right-click on line icon) trace the trajectory of a speckle. 4. On the original stack, do Edit Selection Restore Selection Then. Image Stacks Reslice + OK 5. Rename the new Image Kymo Image Rename 6. On the Kymograph, trace the bright signal with a line ROI 7. Install the macro Read_Line.txt Plugins Macros Install 8. Use Plugins Macros Read velocities from tsp

35 TIME SERIES ANALYSIS: MANUAL TRACKING Goal: Know how to use the Manual Tracking Plugin 1. Open Qdots.avi 2. Separate the channels 3. Use Image Stack Tools Reduce Factor: Open Plugins Tracking Manual Tracking 5. Check Centering Correction, use Local Maximum 6. Start tracking using Add Track 7. Click on the image where the dots are 8. End tracking with End Track 9. Look at the track using the Drawing 10. You can save your results from the Results window and File Save As

36 TIME SERIES ANALYSIS: AUTOMATIC TRACKING Goal: Understand the basics of Particle Tracker from MOSAIC. 1. Close Fiji/IJ, copy the plugin mosaic_plugins.jar to the Plugins Folder 2. Open SwimmingAlgae.tif 3. Do a BG subtraction 4. Open Plugins Mosaic Particle Tracker 5. Try different parameters and use Preview Detected until you have all particles 6. After clicking OK, you can see different trajectories ( Visualize All Trajectories ) 7. Clicking on a trajectory and clicking Focus on selected trajectory isolates it from the image 8. You can extract the trajectories to a table using All trajectories to table

37 TIME SERIES ANALYSIS: AUTOMATIC TRACKING Goal: Understand the basics of MTrack2 1. Open SwimmingAlgae.tif 2. Do a BG subtraction 3. Segment the image (Threshold) 4. Open Plugins Tracking MTrack2 5. Launch it with the criteria you want (If it fails to track, use Edit Invert before trying again) 6. You can export the results to Excel to work on them. NOTE: This is an object-based tracking plugin, whereas Particle Tracker is an intensity-based tracking plugin. In the current case you have to segment the image BEFORE launching the tracker, and you do not have intensity data

38 USING COLOCALIZATION Goal: Be able to use the Colocalization Threshold and the Colocalization Test Plugins 1. Open 63x 76pixel positif positif 16bit.lsm 2. Split the channels, keep only the red one and the green one 3. Rename the images for convenience 4. Run Analyze Colocalization Colocalization Threshold with the following options. 5. Look at the options available and check the results 6. Run Analyze Colocalization Colocalization Test with the settings: 7. Use a ROI to select a sub-region and perform the analysis again. 8. Open Sample_Bleed.lsm and colocalize the Green and Blue channels

39 USING COLOCALIZATION Green channel Results of Colocalization Threshold Red channel Mask ZeroZero Rtotal m b ROI0 excl Ch1 thresh Ch2 thresh Rcoloc R<threshold M1 M2 tm1 tm Ncoloc %Volume %Ch1 Vol %Ch2 Vol % 56.08% 75.89% %Ch1 Int %Ch2 Int %Ch1 Int > thresh %Ch2 Int >thresh 41.93% 64.65% 58.38% 81.45% Good colocalization: 1. High Rcoloc 2. Good thresholds 3. R<threshold close to 0 4. Good regression m (slope), b(intercept) 5. EXAMINE YOUR IMAGE FOR ARTIFACTS Perfect coloc is the sign of Fluorescence Emission Bleed Through

40 COLOCALIZATION DANGERS Bleedthrough Colocalization looks good (numbers) But only in nucleus? Bleed through

41 USING COLOCALIZATION Results of Colocalization Test Good colocalization: 1. All random values are close to 0 2. P-value > The more iterations the better NOTE: The Fiji Colocalization plugin has the following bugs which you should take into account: 1. The Costes Approximation generates a RANDOM image based on the PSF parameters. The original publication calls for using one of the two channels and randomizing PSFsized chunks of that image instead. This is being corrected. 2. The Rcoloc value as well as the Costes Approximation fail in 16-bit images below a certain size. This is also being corrected in the next generation of plugins.

42 USING COLOCALIZATION Other Softwares Just Another Colocalization Plugin (JACoP) (Search JACoP Wiki on Google) Intensity Correlation Analysis (ICA)

43 IMAGE STITCHING: STITCH 2D/3D Goal: Understand the built-in stitchers and be able to perform stitching Exercise 1. Open Stitching\Top_slice_1.tif and Top_slice_2.tif 2. Run Plugins Stitching 2D Stitching Leave the options as they are and click OK 3. Play with the options and see what they do. 4. Open Bottom_slice_1.tif and Bottom_slice_2.tif 5. Try the same protocol as before. Is the stitching correct? 6. Select a rectangular ROI in each image where they overlap and launch 2D Stitching again Now they are properly stitched.

44 IMAGE STITCHING: FLAT-FIELD CORRECTION Goal: Be able to correct non-uniform illumination of image due to geometry of optics Exercise 1. Open Pos004_S001_ch02.tif and Pos005_S001_ch02.tif from the Stitching/single image transmission folder 2. Perform the 2D stitching as before. 3. Now open Pos056_S001_ch02.tif 4. Transform it to 32-bits 5. Apply a Gaussian of ~60.0 to remove dust and small artifacts Process Filters Gaussian Blur 6. Divide the image by its maximum (Normalize it) 7. Divide the two first images by the Flat Field you just computed Make sure that the result is in 32-bit and create new window is ON 8. Perform the stitching again on the newly corrected images and compare them to the original stitch.

45 IMAGE STITCHING: FLAT-FIELD CORRECTION Goal: Be able to correct non-uniform illumination of image due to geometry of optics Non-corrected Corrected

46 IMAGE STITCHING: STITCH GRID Goal: Be able to perform stitching using Stitch Image Grid Plugin Exercise 1. Open Plugins Stitching Stitch Grid of Images 2. Use the following settings: x: 7, y: 8, overlap: 10% Directory: Stitching\Image To Stitch Filenames: Pos{iii}_S001_ch02.tif Leave the rest untouched 3. Perform the same stitching but with the folder Flat Field that contains the corrected images.