High Frequency Ultrasonic Characterization of Carrot Tissue

1
High Frequency Ultrasonic Characterization of
Carrot Tissue

• Christopher Vick
• Advisor Dr. Navalgund Rao
• Center for Imaging Science
• Rochester Institute of Technology

2
Overview

• Introduction
• Hypothesis
• Theory
• Experimental
• Results
• Conclusion

3
Introduction

• Ultrasound fast, nondestructive, noninvasive,
  and inexpensive.
• Long history of diagnostic use.
• Many medical applications consist of interpreting
  an image, based on gray-level and texture.

4
Introduction

• System and processing limitations make this
  ineffective in identifying small variations in
  specific tissue structure.
• Computer texture analysis models are limited in
  scope.
• Models can be aided by quantitatively examining
  the ultrasonic response of tissue.

5
Alternate Ultrasound Uses

• Ripeness measurement in banana and avocado
  animal backfat estimation examination of the
  structure of metals and wood.
• Ultrasound has been proposed for texture
  evaluation of plant tissues, but not widely
  tested.

6
Why carrots?

• Biological changes well documented.
• Homogenous structure
• Since the changing carrot biology is well
  understood, can examine how ultrasound propagates
  through various tissues.

7
Previous Research Results

• Previous research used low frequency ultrasound.
• Notice the nature of their two variables. This
  makes identifying a carrots exact texture
  difficult.

Velocity, Attenuation Vs. Cooking Time
8
Hypothesis

• High frequency ultrasound can be used to
  characterize the cell texture of cooked carrots.
  
• It is hypothesized that varied carrot tissues
  have uniquely identifiable frequency responses.

9
Ultrasound theory

• An ultrasound transducer can convert electrical
  energy to mechanical waves.
• Velocity and attenuation of this signal in a
  medium are characteristic of the mediums
  physical properties.
• The amount of scattering, absorption, and
  reflection, are a function of the medium as well.

10
(No Transcript)
11
Experimental Setup
12
Experimental

• Input Signal Selection
• Input Signal FFT

0
10
5
Frequency (MHz)
13
Transducer Response

• Measure transducer response by filling the jar
  setup with water.

- Less than 5 variation across response curve.
14
Carrot Sample Preparation

• Samples were cored
  from normal Dole
  carrots, using
  an apple corer.
• Samples to be cooked
  were placed in boiling
  water for the appropriate
  0-16 minute cooking
  times, removed,
  and cooled in distilled water.

15
Tests Same Sample

• Examine signal variation from imaging the same
  carrot sample, repeatedly.

- Align carrot/transducers - Image the
sample - Remove the sample
- Repeat process
16
Testing Different Samples

• Examine signal variation along the length of the
  carrot, as the xylem core diameter changes.
• Examine signal variation among different carrots
  of equal cooking time.

17
Testing Cooked Carrots

• Random carrot segments, boiled for between 1-16
  minutes, in 30 second intervals.

• Lastly, random carrot samples were cooked for an
  unknown length of time.
• If successful, results from the previous tests
  should allow for identification of the unknown
  samples.

18
Results Same Sample Readings
- Magnitude variation as high as 20.
- Sources Alignment, transducer coupling
19
Results Normalized
- Variance drops to below 7.
20
Results Different Segments
- Notice that Magnitude decreases as the xylem
core diameter increases.
21
Results Normalized
- After Normalization, variation drops
significantly, to less than 10
22
ResultsDifferent Carrots
- Magnitude Variation can exceed 80
- From alignment, coupling, natural
sample differences
23
Results Normalized
- Variation is significantly decreased.
- Is error too high to allow accurate
classification?
24
Results Various Cooked Carrots
- Frequency response changes can be explained by
the structural changes invoked through cooking.
25
Results Normalized Response LUT
26
Results Normalized Response LUT
Side View of Normalized Response LUT
0 5
10 Frequency (MHz)
27
Analysis Unknown Sample

• IDL Program is given the system output signal of
  a carrot of unknown cooking time.
• Program calculates the FFT, normalizes it, and
  attempts to identify the lowest error associated
  with a match from the known LUT.

28
Results Unknown Carrot Example
1) Given unknown output signal
2) Program calculates signal FFT
29
Results Unknown Analysis
3) Program normalizes FFT,
compares to known FFTs.
4) Program identifies the best match.

5) Program Predicted time 13 minutes
6) Actual Cooking time 13 Minutes
Result Match
Only 10 unknown trial conducted. 4/10 successful.
30
Conclusions

• Focused on the frequency response of carrots.
• Magnitude variation is important factor.
• By normalizing, variation among same sample, or
  different segments is lowered substantially.
• Large signal variation among different carrots.

31
Conclusion

• IDL analysis needs further attention not all
  carrots can be identified.
• Combining analysis with the previously studies
  variables of Velocity and Attenuation would
  likely provide a more robust tissue
  identification model.

32
Special Thanks to
Dr. Navalgund Rao Maria Helguera Brad Miller