Extracting the Meaning: Painless processing and analysis of image data with Fiji and Ruby

Slide 1

Slide 1 text

painless processing and analysis of image data with Fiji and Ruby extracting the meaning Based on the experience from the summer project by Gregory Goltsov melville f o u n d a t i o n Supervisor Dr Anne Savage

Slide 2

Slide 2 text

Slide 3

Slide 3 text

Slide 4

Slide 4 text

who am i?

Slide 5

Slide 5 text

who am i? 4th year Computer Games Technology

Slide 6

Slide 6 text

who am i? 4th year Computer Games Technology Ruby fan

Slide 7

Slide 7 text

who am i? 4th year Computer Games Technology Ruby fan (basically) a programmer

Slide 8

Slide 8 text

the project

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

Data

Slide 11

Slide 11 text

Breakthrough Breast Cancer Research Unit and Division of Pathology, University of Edinburgh

Slide 12

Slide 12 text

Breakthrough Breast Cancer Research Unit and Division of Pathology, University of Edinburgh Breast cancer spheroid

Slide 13

Slide 13 text

Breakthrough Breast Cancer Research Unit and Division of Pathology, University of Edinburgh Breast cancer spheroid Drug 396 403 S4 ...

Slide 14

Slide 14 text

Breakthrough Breast Cancer Research Unit and Division of Pathology, University of Edinburgh Breast cancer spheroid Drug 396 403 S4 ... ?

Slide 15

Slide 15 text

Breakthrough Breast Cancer Research Unit and Division of Pathology, University of Edinburgh Aim To develop a metric to objectively measure the eﬀectiveness of said drugs on the spheroids Breast cancer spheroid Drug 396 403 S4 ... ?

Slide 16

Slide 16 text

Data

Slide 17

Slide 17 text

Data

Slide 18

Slide 18 text

pre-processing • processing • analysis Data

Slide 19

Slide 19 text

pre-processing • processing • analysis Data Results

Slide 20

Slide 20 text

pre-processing • processing • analysis Data Results ──────

Slide 21

Slide 21 text

pre-processing • processing • analysis Data Results ────── Scalable

Slide 22

Slide 22 text

pre-processing • processing • analysis Data Results ────── Scalable Robust

Slide 23

Slide 23 text

pre-processing • processing • analysis Data Results ────── Scalable Robust Fully automatic

Slide 24

Slide 24 text

pre-processing • processing • analysis Data Results ────── Scalable Robust Fully automatic ──────

Slide 25

Slide 25 text

pre-processing • processing • analysis

Slide 26

Slide 26 text

pre-processing • processing • analysis Filtering invalid data

Slide 27

Slide 27 text

pre-processing • processing • analysis Filtering invalid data Renaming

Slide 28

Slide 28 text

396 3um well1 time0 403Contwell1time0 409BCont4time0 616inact 30um well2 time0 403 Cont 1 24h S4 1uM 1 24h 616 (inact) 1uM 1 time 48h S4 3 uM 4a time 48h

Slide 29

Slide 29 text

396 3um well1 time0 403Contwell1time0 409BCont4time0 616inact 30um well2 time0 403 Cont 1 24h S4 1uM 1 24h 616 (inact) 1uM 1 time 48h S4 3 uM 4a time 48h

Slide 30

Slide 30 text

396 3um well1 time0 403Contwell1time0 409BCont4time0 616inact 30um well2 time0 403 Cont 1 24h S4 1uM 1 24h 616 (inact) 1uM 1 time 48h S4 3 uM 4a time 48h How to retrieve a particular image?

Slide 31

Slide 31 text

396 3um well1 time0 403Contwell1time0 409BCont4time0 616inact 30um well2 time0 403 Cont 1 24h S4 1uM 1 24h 616 (inact) 1uM 1 time 48h S4 3 uM 4a time 48h How to retrieve a particular image? W hat is the id?

Slide 32

Slide 32 text

Slide 33

Slide 33 text

Slide 34

Slide 34 text

So I implemented the RENAMER

Slide 35

Slide 35 text

RENAMER

Slide 36

Slide 36 text

RENAMER

Slide 37

Slide 37 text

RENAMER 396 3um well1 time0 403Contwell1time0 409BCont4time0 616inact 30um well2 time0 403 Cont 1 24h S4 1uM 1 24h 616 (inact) 1uM 1 time 48h S4 3 uM 4a time 48h

Slide 38

Slide 38 text

RENAMER 396 3um well1 time0 403Contwell1time0 409BCont4time0 616inact 30um well2 time0 403 Cont 1 24h S4 1uM 1 24h 616 (inact) 1uM 1 time 48h S4 3 uM 4a time 48h 396 3uM 1 0h 403 (cont) 1 0h 409 (cont) 4 0h 616 (inact) 30uM 2 0h 403 (cont) 1 24h S4 1uM 1 24h 616 (inact) 1uM 1 48h S4 3uM 4 48h

Slide 39

Slide 39 text

But why bother with RENAMER ?

Slide 40

Slide 40 text

But why bother with RENAMER ? Consistent

Slide 41

Slide 41 text

But why bother with RENAMER ? Consistent Scalable

Slide 42

Slide 42 text

But why bother with RENAMER ? Consistent Scalable DB-like querying

Slide 43

Slide 43 text

But why bother with RENAMER ? Consistent Time-series analysis Scalable DB-like querying

Slide 44

Slide 44 text

But why bother with RENAMER ? Consistent Time-series analysis Robust Scalable DB-like querying

Slide 45

Slide 45 text

pre-processing • processing • analysis

Slide 46

Slide 46 text

pre-processing • processing • analysis Brightness/contrast correction

Slide 47

Slide 47 text

pre-processing • processing • analysis Brightness/contrast correction Pseudo Flat-Field correction

Slide 48

Slide 48 text

pre-processing • processing • analysis Brightness/contrast correction Pseudo Flat-Field correction Background subtraction

Slide 49

Slide 49 text

pre-processing • processing • analysis Brightness/contrast correction Pseudo Flat-Field correction Background subtraction Thresholding

Slide 50

Slide 50 text

Unprocessed data varies a lot 0 h 24 h 48 h

Slide 51

Slide 51 text

Unprocessed data varies a lot, making it impossible to threshold 48 h 0 h 24 h

Slide 52

Slide 52 text

Original

Slide 53

Slide 53 text

Brightness/ contrast correction # Code: Computations.enhance_contrast imp, la

Slide 54

Slide 54 text

Pseudo Flat-Field correction # Code: # This is the kernel size for Gaussian blur. # Smaller kernel removes more background kernel_size = 100 pffc = Computations.pffc imp, kernel_size, true

Slide 55

Slide 55 text

Background subtraction # Code: IJUtils.run 'Subtract Background...', :rolling => 100, :light_disable => ''

Slide 56

Slide 56 text

Otsu thresholding # Code: processor.set_auto_threshold AutoThresholder::Method::Otsu,false IJUtils.run 'Convert to Mask' IJUtils.run 'Despeckle'

Slide 57

Slide 57 text

Original Brightness/contrast adjustment + PFFC Subtract background Otsu thresholding

Slide 58

Slide 58 text

pre-processing • processing • analysis

Slide 59

Slide 59 text

pre-processing • processing • analysis Klonowski landscapes

Slide 60

Slide 60 text

pre-processing • processing • analysis Klonowski landscapes Higuchi fractal dimension

Slide 61

Slide 61 text

pre-processing • processing • analysis Klonowski landscapes Higuchi fractal dimension Area and perimeter

Slide 62

Slide 62 text

klonowski landscapes

Slide 63

Slide 63 text

klonowski landscapes 2D image ➝ 1D “signal”

Slide 64

Slide 64 text

No content

Slide 65

Slide 65 text

No content

Slide 66

Slide 66 text

No content

Slide 67

Slide 67 text

Vertical landscape Horizontal landscape

Slide 68

Slide 68 text

Klonowski et al. Nonlinear Biomedical Physics 2010, 4:7 http://www.nonlinearbiomedphys.com/content/4/1/7 Vertical landscape Horizontal landscape

Slide 69

Slide 69 text

Vertical landscape Horizontal landscape

Slide 70

Slide 70 text

higuchi fractal dimension

Slide 71

Slide 71 text

higuchi fractal dimension Complexity, D, of a “signal”/curve

Slide 72

Slide 72 text

No content

Slide 73

Slide 73 text

“Jagged, uneven, rough edge” “Smooth, gradual, round” D = 1.84 D = 1.07

Slide 74

Slide 74 text

“Jagged, uneven, rough edge” “Smooth, gradual, round” D = 1.84 D = 1.07

Slide 75

Slide 75 text

No content

Slide 76

Slide 76 text

Untreated

Slide 77

Slide 77 text

Untreated Treated (1μM 396)

Slide 78

Slide 78 text

D ≃ 1 Untreated Treated (1μM 396)

Slide 79

Slide 79 text

D ≃ 1 Untreated Treated (1μM 396) D ≃ 1.3

Slide 80

Slide 80 text

D ≃ 1 Untreated Treated (1μM 396) D ≃ 1.3 D ≃ 1.32

Slide 81

Slide 81 text

D ≃ 1 Untreated Treated (1μM 396) D ≃ 1.3 D ≃ 1.32 D ≃ 1

Slide 82

Slide 82 text

D ≃ 1 Untreated Treated (1μM 396) D ≃ 1.3 D ≃ 1.32 D ≃ 1 D ≃ 1.15

Slide 83

Slide 83 text

D ≃ 1 Untreated Treated (1μM 396) D ≃ 1.3 D ≃ 1.32 D ≃ 1 D ≃ 1.15 D ≃ 1.28

Slide 84

Slide 84 text

results

Slide 85

Slide 85 text

resultsalpha

Slide 86

Slide 86 text

Lower % is better

Slide 87

Slide 87 text

Lower % is better

Slide 88

Slide 88 text

what next?

Slide 89

Slide 89 text

what next? ──────

Slide 90

Slide 90 text

what next? ────── Open-source Higuchi fractal dimension code

Slide 91

Slide 91 text

what next? ────── Open-source Higuchi fractal dimension code More analysis

Slide 92

Slide 92 text

what next? ────── Open-source Higuchi fractal dimension code More analysis Paper

Slide 93

Slide 93 text

what next? ────── Open-source Higuchi fractal dimension code More analysis Paper ──────

Slide 94

Slide 94 text

thank you! [email protected] www.gregory.goltsov.info Questions? melville f o u n d a t i o n Big thanks to: School of CONTEMPORARY SCIENCES