Which of the six boxes below cannot be made from this unfolded box? (There may be more than ffi. ffi

Transcription

1 Which of the six boxes below cannot be made from this unfolded box? (There may be more than ffi c

2 YORK Peripheral sensory transduction Christopher Bergevin (York University, Dept. of Physics & Astronomy) CVR Summer School (2016)

3 Which of the six boxes below cannot be made from this unfolded box? (There may be more than one.) Question: Physiologically, how did you (try to) solve this ffi c à We ll come back to the basic units of this thing in a bit National Geographic!

4 Question: How do our sensory systems encode information about the world around us? Consider how you process this picture...

5 Question: How is information being transduced here? WebVision (Utah)

6 Palmer (1999)

7 Palmer (1999)

8 Question: What are the basic building blocks that make up these circuits? McIlwain (1996)

9 Aside: Images as numbers (i.e., a bitmap ) Note Even this basic picture is too simple for a jpeg file Question: Does your eye/nervous system process and store this image like a computer does?

10 Big Picture Theme How do our sensory systems encode information about the world around us? Not just vision... Pulkki & Karjalainen (2015)

11 Big Picture Theme How do our sensory systems encode information about the world around us? Two broad topics to cover What are the basic building blocks that make up these circuits? What are some basic (signal processing) considerations about transforming information?

12 à How might you go about measuring brain activity? 1 Basic neuroscience building blocks

13 1 Basic neuroscience building blocks What are these methods used to measure neural activity actually telling us? Epstein & Kanwisher (1998)

14 Why/how are direct neurophysiological responses commonly measured as spike rates? 1 Basic neuroscience building blocks

15 1 Basic neuroscience building blocks Voltage Time à Why/how are responses commonly measured as spike rates? Note: This work led to Hubel & Wiesel winning the 1981 Nobel Prize

16 1 Basic neuroscience building blocks

17 1 Basic neuroscience building blocks Neural code Aside Is our central nervous system essentially digitized?

18 1 Basic neuroscience building blocks Voltage Time Ø What is a spike? Ø What is being measured Ø How is such measured? Hodgkin & Huxley (1939)!

19 1 Basic neuroscience building blocks Hodgkin & Huxley (1939)!

20 Ex. Neural coding of sound 1 Basic neuroscience building blocks Cochlear nerve contains ~30000 fibers

21 1 Basic neuroscience building blocks Response of a single cochlear nerve fiber Voltage Time Neuron Mic Kiang!

22 Neurons 1 Basic neuroscience building blocks Neurons ( fibers ) = Information highway Key Point: Electrical properties of cells are important Weiss (1996)!

23 Action potentials 1 Basic neuroscience building blocks Weiss (1996)!

24 Action potentials 1 Basic neuroscience building blocks Neuron Stimulus Response à Neurons send info via electrical pulses (spikes) occurring across the cell membrane Weiss (1996)!

25 Cell membrane 1 Basic neuroscience building blocks Ø Membrane primarily consists of a lipid bilayer (to separate inside from outside) Ø All sorts of stuff embedded inside, to allow for communication across membrane zoom in on cell membrane inside cell outside cell Weiss (1996)!

26 Electrical excitability 1 Basic neuroscience building blocks electrically inexcitable cell electrically excitable cell Weiss (1996)!

27 Electrical Responses in Sensory Systems 1 Basic neuroscience building blocks Photoreceptors à Not always electrically excitable per se, but role as transducers critically tied to electrical responses Auditory Hair Cells

28 Cable model 1 Basic neuroscience building blocks - First solved by William Thomson (aka Lord Kelvin) in ~ Motivated by Atlantic submarine cable for intercontinental telegraphy Weiss (1996)!

29 Biophysical model of a neuron 1 Basic neuroscience building blocks à Model via an electric circuit Freeman!

30 Biophysical model of a neuron 1 Basic neuroscience building blocks à Cell behave like a leaky submarine cable! Weiss (1996)!

31 Biophysical model of a neuron 1 Basic neuroscience building blocks Variable Na+ and K+ conductances Weiss (1996)!

32 Hodgkin-Huxley equations 1 Basic neuroscience building blocks

33 1 Basic neuroscience building blocks

34 Putting the pieces together... 1 Basic neuroscience building blocks Weiss (1996)!

35 Summary (re neurons) 1 Basic neuroscience building blocks Weiss (1996)!

36 Big Picture Theme How do our sensory systems encode information about the world around us? What are the basic building blocks that make up these circuits? Human brain contains ~10 11 (100 billion) neurons! (with 100 trillion+ connections inbetween)

37 Human brain contains ~10 11 (100 billion) neurons! (with 100 trillion+ connections inbetween) à This is a pretty hard problem!

38 Epstein & Kanwisher (1998)

39 1 Basic neuroscience building blocks Palmer (1999)

40 Pop Quiz What is the difference between these two different types of spikes? Fox, Pandit-Taskar, Strauss (2013)

41 Big Picture Theme How do our sensory systems encode information about the world around us? Two broad topics to cover What are the basic building blocks that make up these circuits? What are some basic (signal processing) considerations about transforming information?

42 2 Transforming information Transduction 1. the transfer of genetic material from one organism (as a bacterium) to another by a genetic vector and especially a bacteriophage 2. the action or process of converting something and especially energy or a message into another form Merriam-Webste!

43 Peripheral sensory transduction Chemical Transduction 2 Transforming information! 9%*&+'#), +1%%1& </4(. 0/+/6 9%*&+'#), 6(.6#),.(/)#. 7/88#)'1.3 +(%% Mechanical Transduction 9%*&+'#),.(/)#.$&:#.6 "#$#%*&+'#), ;/%; 5&6&% +(%% 5&6(4(.' 4(4;)&.( Visual Transduction -%#4()/%/6 2()13%#4()/%&)$+(%% 01')&%$+(%% -)&./%( +(%% "#$%&'()&%$#%*&+'#),$')&+' Firestein (2001) Rodieck (1998)

44 Question: How similar/different is the input versus the output? 2 Transforming information

45 Somehow, the information is transformed, encoded into some other language... 2 Transforming information

46 Aside: Images as numbers (i.e., a bitmap )

47 Many ways to encode something... 2 Transforming information

48 Many ways to encode something... 2 Transforming information Bitmap version Vector version zoom-in about corner of eye

49 Many ways to encode something... 2 Transforming information Bitmap version Vector version à Same image, two very different representations Wikipedia

50 Let s focus for a bit on how the ear encodes incoming sounds 2 Transforming information

51 Spectral Decomposition 2 Transforming information An Acoustic Prism High frequencies Mid frequencies Low frequencies Zweig et al. (1976)

52 Aside: Fourier analysis EXspectrogram.m 2 Transforming information A E I O U frequency time amplitude [s]

53 Aside: Acoustic phonetics 2 Transforming information Human vocal tract cross-section

54 Aside: Speech chain 2 Transforming EXspectrogram.m information Pulkki & Karjalainen (2015)

55 Aside: Fourier analysis 2 Transforming information à Time and frequency are separated but are two sides of the same coin

56 Ear is a Fourier analyzer (Tonotopicity) 2 Transforming information Basilar membrane (BM) higher frequencies lower frequencies! Geisler (1996)!

57 Hair cell = Mechano-electro transducer Mammalian OAE generation overview! 2 Transforming information Traveling Waves!

58 2 Transforming information An Acoustic Prism High frequencies Mid frequencies Ø Ear acts as a hydrodynamic spectrum analyzer (spatial location ß à frequency) Low frequencies Ø Spectral decomposition serves as an underlying basis for auditory neural code

59 Neural coding of speech 2 Transforming information Delgutte (1997)!

60 Aside: Imaging & Fourier Analysis 2 Transforming information Medical/Biological/Neural Imaging (e.g., MRI, CT, OCT, x-ray crystallography, spectroscopy, microscopy, interferometry,...) Ø Fourier transform is a key foundation in imaging (e.g., k-space in MRI) Ø Also the backbone of modern signal processing Wikipedia

61 Aside: Fourier analysis (REVISITED) 2 Transforming information Intuitive connection back to Taylor series: Taylor series à Expand as a (infinite) sum of polynomials Different Idea: Fourier series à Expand as a (infinite) sum of sinusoids

62 Aside: Fourier analysis (REVISITED) 2 Transforming information The exponential function e x (in blue), and the sum of the first n+1 terms of its Taylor series at 0 (in red). wikipedia (Taylor

63 Aside: Fourier analysis (REVISITED) 2 Transforming information wikipedia (square

64 2 Transforming information Animation of the additive synthesis of a square wave with an increasing number of harmonics. The six arrows represent the first six terms of the Fourier series of a square wave. The two circles at the bottom represent the exact square wave (blue) and its Fourier-series approximation (purple). wikipedia (square

65 2 Transforming information But what about vision? What is the basis for the underlying neural code? Harder questions I don t know the answers to but spectral analysis is a useful starting point! Palmer (1999)

66 2 Transforming information Consider: What makes an edge and edge? wikipedia (square

67 2-D Fourier Transforms EXfourier2D.m (Richard Murray, CVR) Spatial domain Frequency domain Note: Only ½ of the information is shown on the right (amplitude only; phase not shown)

68 EXfourier2D.m (Richard Murray, CVR) à Low-pass filtered version of the image

69

70 2 Transforming information

71 Find Edges 2 Transforming information

72 Stained Glass 2 Transforming information

73 Motion Blur 2 Transforming information

74 2 Transforming information How does Photoshop work? (at least in very basic terms)

75 2-D Convolutions: Images & Filtering 2 Transforming information Ø Two basic ingredient: image and kernel (or filter) Ø Kernel is tied back to an impulse response Ø Convolution is an operation that ties the two together (surprisingly universal consideration throughout science) Pawley (ch.25)

76 Image Processing: Filtering 2 Transforming information Convolution à Blurring Blurring kernel Pawley (ch.25)

77 Image Processing: Filtering 2 Transforming information Convolution à Sharpening Basic idea is that a convolution is a numerical operation between image and kernel Wikipedia [see page for Kernel (image processing)]

78 Connection to Fourier Transforms Ø Consider the Fourier transform of the (1-D) convolution: Ø Making use of the shifting property, the term in the square brackets is: Q(ω) is the Fourier transform of q(t) Convolution theorem P(ω) is the Fourier transform of p(t) Devries (1994)

79 Convolution Theorem Ø Simple but powerful idea: Convolution in the time domain is simply a multiplication in the spectral domain Ø Door swings both ways: From the output, if we know the impulse response, we can deconvolve (i.e., divide in spectral domain) to get the original input! Short version: Numerically this is easy to do! Devries (1994)

80 What are some basic (signal processing) considerations about transforming information? Ø Does Photoshop and the visual system operate in the same way? à No. But there are likely basic signal processing facets (e.g., spectral decomposition) universally at work

81 Question: How do our sensory systems encode information about the world around us? Consider how you process this picture...

82 Ø Transducers and neurons as the basic building blocks Ø Transformation into a neural code involves abstract-ish signal processing considerations (e.g., spectral analysis, convolution) Palmer (1999)

83 Which of the six boxes below cannot be made from this unfolded box? (There may be more than ffi c National Geographic!

84 YORK Ø Slides available for download: Ø Questions? cberge@yorku.ca Ø Interested in grad school?

85

86 2 Transforming information The Eye (Al-Hasan ibn al-haitham 1083)