Imaging Considerations to Enhance Data Post- Processing Trevor - PowerPoint PPT Presentation

Imaging Considerations to Enhance Data Post- Processing Trevor Lancon

2/29/2016 Commonly Asked Questions • What are the hardware requirements for Amira/Avizo? • How much RAM do I need? • How can I work more efficiently with large data? → File size Confidential 2

Example of How Avizo and Amira Utilize Hardware How are hardware components utilized in this example?

Example of How Avizo and Amira Utilize Hardware Hard Drive Copy of 1 Dataset Original Filtered Dataset Dataset RAM 3 2 5 4 3D Visualization Filtering Algorithm 2D Visualization CPU VRAM

2/29/2016 1 2 3 4 1. Open data 2. Ortho Slice 3. Median Filter 4. Volren Confidential 5

2/29/2016 File Size Example DualBeam Workflow 1. Load data 2. Slice alignment 3. Smoothing filter 4. Sharpening filter 5. Background correction 6. FFT filter 6x original in RAM! Confidential 6

2/29/2016 File Size FOV / Resolution Bit Depth • Small structures: high resolution • Coarseness of histogram bins • Statistics: large FOV • Image contrast – Reject poor samples in post- • Larger affect on file size analysis filtering – Reject 70% of 15 grains vs. – Reject 70% of 4000 grains Confidential 7

2/29/2016 File Size Consider 1500 x 1286 x 1040 Voxels 8bit 16bit float 8 7.5 7 6 File Size [GB] 5 4 3.7 3 2 1.9 1 0 100% 90% 80% 70% 60% 50% 40% FOV or Resolution [% of original size] Confidential 8

2/29/2016 Convert to 8bit Crop 60% None File Size 25 22.2 Example DualBeam Workflow 20 1. Load 16-bit data 16.9 2. Manage file size 15 15.1 3. Slice alignment 4. Smoothing filter 10 5. Sharpening filter 6. Background correction 5 7. FFT filter 0 Confidential 9

2/29/2016 Commonly Asked Questions • What are the hardware requirements for Amira/Avizo? • How much RAM do I need? • How can I work more efficiently with large data? → File size • WHY IS SEGMENTATION SO HARD?! → Contrast Confidential 10

2/29/2016 Contrast • Segmentation depends on contrast – Thresholding, watershed, morphology, etc. • Other factors affect contrast – Probe (e.g. electrons) – Signal collection (e.g. camera exposure) • Optimize current vs. stage drift • Optimize dwell time / frame averaging vs. scan time – Scan time vs. segmentation time • Reducing bit depth may reduce contrast? Confidential 11

2/29/2016 Contrast Confidential 12

2/29/2016 Contrast 8 Bit – Normalized 16 Bit – Normalized 255 65535 238 61166 221 56797 204 52428 187 48059 170 43690 153 39321 136 34952 119 30583 102 26214 85 21845 68 17476 51 13107 34 8738 17 4369 0 0 Confidential 13

2/29/2016 Contrast 8 Bit 16 Bit 220 4400 200 4000 180 3600 160 3200 140 2800 120 2400 100 2000 Confidential 14

2/29/2016 Contrast 8 Bit – Thresholded (168) 16 Bit – Thresholded (3341) Confidential 15

2/29/2016 Contrast 8 Bit – Thresholded (168) 16 Bit – Thresholded (3341) Confidential 16

2/29/2016 Contrast 8 Bit – Watershed 16 Bit – Watershed Confidential 17

Thank you FIBSEM UGM!