Melanoma and skin cancers vs Image Processing

1
Melanoma and skin cancers vs Image Processing
2
Skin cancer and melanoma

• Skin cancer most common of all cancers

3

1. According to the latest statistics available from
   the National Cancer Institute, skin cancer is the
   most common of all cancers in the United states.
2. More than 1 million cases of skin cancer are
   diagnosed in the US each year.
3. Whats shown here are some examples of skin
   lesion images.
4. The four images shown on the left are various
   form of skin lesions, cancerous or non-cancerous.
   
5. The two on the right are a specific form of skin
   cancer melanoma.

4
What is Melanoma?

1. A type of skin cancer that starts from
   melanocytes
2. 6th leading cause of cancer death in the US
3. No single etiology
4. Some risk factors include
5. Sun exposure -depleting ozone layer
6. Presence of many or unusual moles
7. Skin types
8. Genetics predisposition

5
benign
skin
malignant
6
Skin cancer and melanoma

• Skin cancer most common of all cancers

Image courtesy of An Atlas of Surface
Microscopy of Pigmented Skin Lesions Dermoscopy

7
Use of color to distinguish malignant and benign
tumors

• Skin tumors can be either malignant or benign
• Clinically difficult to differentiate the early
  stage of malignant melanoma and benign tumors due
  to the similarity in appearance
• Proper identification and classification of
  malignant melanoma is considered as the top
  priority because of cost function
• Classification of skin tumors using computer
  imaging and pattern recognition
• Previous texture feature algorithms successfully
  differentiate the deadly melanoma and benign
  tumor seborrhea kurtosis
• Relative color feature algorithm is explored in
  this research for differentiate melanoma and
  benign tumors, dysplastic nevi and nevus
• Successfully classify 86 of malignant melanoma
  using relative color features, compared to the
  clinical accuracy by dermatologists in detection
  of melanoma of approximately 75

8
Types of Melanoma

• Superficial Spreading Melanoma
• 70, neck, legs, pelvis
• Nodular Melanoma
• 15, dome-shaped nodule
• Acral-Lentiginous Melanoma
• 8 , Common in dark-skin
• Lentigo Maligna Melanoma
• 5 , sun-exposed area, mistaken for age spot
• Amelanotic Melanoma
• 0.3, non-pigmented
• Desmoplastic
• 1.7, ½ amelanotic

9
Benign vs Malignant
9
10
10
11
Automated Melanoma Recognition UsingImaging
Techniques

• Melanoma is one of the most aggressive cancers,
  but it can be healed by surgical excision
  successfully only if it is recognized in the
  early stage.
• Since the melanoma emerges as a tiny dot in the
  topmost skin layer, it can be examined during
  routine medical check up.
• Although the lesions are accessible, in many
  cases, it is a difficult task to make decisions
  whether nevi are benign or malignant.
• Further, frequent use of biopsy is also not
  encouraged.
• Hence, to assist dermatologist's diagnosis, it is
  useful to develop an automated imaging-based
  melanoma recognition system.

12

• Uncontrolled growth of melanocytes give rise to
  dark and elevated appearance of melanoma.
• Neoplasm- growth of tissue, tumor
• Melanoma is a type of malignant skin cancer that
  starts from melanocytes. Its caused by
  uncontrolled growth of melanocytes that gives
  rise to tumor.

13

1. Nonetheless there are risks factors that highly
   attributed to its incidence. Some of the them
   are
2. amount of sun exposure the more cumulative
   exposure the higher
3. presence of many of unusual mole people with
   many moles in the body
4. Fitzpatricks Skin Type I and II have higher risk
   1975 Thomas Fitzpatrick, Harvard skin typing
   system based on skin complexion and response to
   sun exposure
5. genetic predisposition if there history of
   melanoma that runs in the family
6. According to a study ,compared to general
   population, people who with 2 risk factors have
   3.5 times risk of developing MM and 20 times
   those who have 3 or more risk factors.

14

1. These are the types of melanona
2. As you see, SSM is the most prevalent one that
   makes up 70 of most diagnosed melanoma
3. In this work, images of superficial spreading
   melanoma were only explored.
4. The reason being, and the problem that this work
   is trying to solve, Dysplastic Nevi ( a benign
   mole) has properties that are highly similar to
   this SSM melanoma, which makes the diagnosis of
   melanoma difficult.

15
Melanoma Incidence
NCHS national center for health
statistics Bureau of Health Statistics
Incidence highest in Caucasian skin
Graph one- Caucasian has the highest incidence of
MM. Having fair complexion is one of the risk
factors. Researches attribute this to low level
of melanin that absorbs harmful UV radiation in
fair skin, thus UV penetrates much deeper layer
affects the surrounding cells.
16

1. Graph two men shows higher incidence than
   women.
2. A study of in Germany linked this trend to
   mutation of genes called BRAF 4 and CDKN2A 1.

17
Melanoma Incidence
Graph thee Incidence increases with age. Link
to cumulative sun exposure Some studies
suggested that people who had significant
exposure to UV at younger age have higher risk in
later age when UV exposure decreases.
Incidence increases with age
18

• Age-adjusted- distribution of age by percentage
• Its a way of data normalization so that you can
  compare two different countries, cities and so
  forth
• Need standard population distribution
• Who use it
• NCHS national center for health statistics
• Bureau of Health Statistics
• What to say
• So these are three graphs that show melanoma
  incidence in different dimensions based on race,
  gender, and age.
• Here, its evident that Melanoma has its
  favorites, so to speak.

19
Melanoma Incidence
It is estimated that 62,480 men and women (34,950
men and 27,530 women) will be diagnosed with and
8,420 men and women will die of melanoma of the
skin in 2008 (SEER)
20

• Surveillance Epidemiology and End Results
• What to say
• This is the combination of all of the data from
  the previous slides.
• Average of 4.2 percent increase per year

21
Survival Rate by Stage
The American Joint Committee on Cancer (AJCC) TNM
System
http//www.cancer.org
22

1. The imaging is performed by a special CCD camera
   combined with an epiluminescence microscope in
   order to produce digitalized ELM images of the
   skin lesions.
2. Once the images are captured, the lesion has to
   be segmented from the background and useful
   information should be extracted from the lesion
   region.
3. Based on the extracted features, decisions have
   to be made about the nature of the skin lesion.
4. The decisions should be supported by descriptive
   justifications so that dermatologist can
   understand the decision making process.

23

• Contact person Assoc. Prof. PonnuthuraiNagaratnam
  Suganthan, email epnsugan_at_ntu.edu.sgTel
  6790-5404
• Collaborators Prof. C L Goh, MD, National Skin
  Centre, Singapore Dr. H Kittler, University of
  Vienna
• This is an on-going project. We have implemented
  the segmentation, feature extraction and
  clasifcation modules satisfactorily, although
  further improvements are desirable. The module to
  provide explanations supporting the classifcation
  decisons is yet to be developed. siii

24
Skin cancer and melanoma

• Skin cancer most common of all cancers
• Melanoma leading cause of mortality (75)

1. Although represent only 4 percent of all skin
   cancers in the US, melanoma is the leading cause
   of mortality.
2. They account for more than 75 percent of all skin
   cancer deaths.

Image courtesy of An Atlas of Surface
Microscopy of Pigmented Skin Lesions Dermoscopy

25
Skin cancer and melanoma

• The time line shown here is the 10 year survival
  rate of melanoma.
• If caught in its early stage, as seen here,
  melanoma can often be cured with a simple
  excision, so the patient have a high chance to
  recover. Hence, early detection of malignant
  melanoma significantly reduces mortality.

• Skin cancer most common of all cancers
• Melanoma leading cause of mortality (75)
• Early detection significantly reduces mortality

Image courtesy of An Atlas of Surface
Microscopy of Pigmented Skin Lesions Dermoscopy

26
Clinical View
Dermoscopy view
Image courtesy of An Atlas of Surface
Microscopy of Pigmented Skin Lesions Dermoscopy

27
Dermoscopy

1. Dermoscopy is a noninvasive imaging technique,
   and it is just the right technique for this task.
2. It has been shown effective for early detection
   of melanoma.
3. The procedure involves using an incident light
   magnification system, i.e. a dermatoscope, to
   examine skin lesions.
4. Often oil is applied at the skin-microscope
   interface.
5. This allows the incident light to penetrate the
   top layer of the skin tissue and reveal the
   pigmented structures beyond what would be visible
   by naked eyes.

28
Dermoscopy

• Dermoscopy improves diagnostic accuracy by 30 in
  the hands of trained physicians
• May require as much as 5 year experience to have
  the necessary training
• Motivation for Computer-aided diagnosis (CAD) of
  pigmented skin lesion from these dermoscopy
  images.

29

• In the future, with the development of new
  algorithms and techniques, these computer
  procedures may aid the dermatologists to bring
  medical break through in early detection of
  melanoma.

30

• 40,000 people between 1988-2001
• Cancer stage is categorized into TNM level
• T tumor ( localized)
• N regional lymph-nodes
• M -Metastasis
• The key point is the earlier the better of
  survival
• 5- and 10- year survival mean percentage of
  people who live at least 5 and 10 years
  respectively after being diagnosed

31
Diagnosis- ABCDE System
32

1. E evolution/elevation
2. What to say
3. ABCDE system is the tool for detecting melanoma.
   This is a list of criteria that can be used for
   distinguishing between benign and malignant
   melanocytic skin lesions.
4. A- if you draw a line across the center of MM,
   youll see that is not symmetric compared to
   regular mole
5. B- the border is uneven or ragged is a sign of
   melanoma
6. C-if there are multiple shades of pigment is
   presence
7. D- diameter gt 6mm
8. Dermatologist adds E for either evolution if
   lesion changes upon observation or E for
   elevation.
9. Suspicious lesion is followed by histological
   confirmation.

33
Where the problems lie

• Atypical nevi acquire several properties similar
  to melanoma, their recognition posed high
  difficulties even to experts. The classical ABCD
  guidance is not reliable therefore cannot be used
  as sole indicator for detection of melanoma for
  both clinical and public examination.
• In clinical setting, recognition and
  discrimination are highly subjective with rate of
  success based on experts years of experience. As
  was found, inexperienced dermatologists showed
  decrease sensitivity in the detection of melanoma
  in both live and photo examinations.
• General practitioner 62 sensitivity and 63
  specificity
• Dermatologist 80 sensitivity and 60
  specificity

34

• OK, so we have the ABCD diagnosis tool plus the
  experts.
• So anyone with sort of skin lesion can step in a
  clinic get the ABCD tool and experts examination
  undertaken then there you have the results.
• You either have benign mole or malignant melanoma
  at the end of the consultation. Everything just
  goes as plan.
• Unfortunately it is not always the case.
• Sensitivity TP/TPFN
• Specificity TN/TNFP
• Read the bullet
• The objective of the this work is to address
  these problems

35
MM and DN
Here you have some samples of MM on the top row
and DN on the bottom row Atypical Nevi (mole)
shares some sometimes all characteristics of
MM. This actually what makes melanoma detection
difficult.

• ABCD Rules

Malignant Melanoma
Dysplastic Nevi
36
Objectives

• To construct an automated, image-based system for
  classification of Malignant Melanoma and
  Dysplastic Nevi using solely the visual texture
  information of the lesion. The system will be
  based on methodologies that emanate and/or
  correlated with human vision therefore will
  closely emulates human experts only with greater
  extent of accuracy, reliability and
  reproducibility
• Investigate new segmentation methods that will be
  effective on both lesions
• Extract most relevant texture information from
  the image
• Construct a classification system of the lesion

37

• Ultimate goal is the construction of the
  classification system
• The uniqueness of the system is the fact that
• only texture information is used robust in
  color variability
• Methodologies used through out the whole process
  emanate from the human vision thus emulate human
  expert

38
Systems, Materials and Tools

• Image database
• Original tumor images
• 512x512 24-bit color images digitized from 35mm
  color photographic slides and photographs
• 160 melanoma, 42 dysplastic, and 80 nevus skin
  tumor images
• Border images
• Binary images drawn manually and reviewed by the
  dermatologist for accuracy
• Software
• CVIPtools
• Computer vision and image processing tools
  developed at our research lab
• Partek
• Statistical analysis tools

39
CVIPtools
40
(No Transcript)
41
Other approach - Texture
42
System for Melanoma Detection
43
Outline of the Process

1. Here you have the outline of the process
2. Each of the subsequent step is dependent of the
   of the preceding steps. In other terms, the
   results of subsequent step is only as good as the
   results of preceding steps.
3. Therefore, since segmentation is the top most of
   the hierarchy, its important to make sure the
   method is robust.

44
Hypotheses

• Due to observable pattern disruption in the skin
  tissue driven by the MM, It is hypothesize that
  measuring magnitude of pattern disruption
  provides discriminative features for diagnosing
  MM.
• Since visual texture is highly length-scale
  dependent, It is hypothesized that the detection
  and analysis methods that explore texture at
  different scales such as the wavelet is the most
  appropriate approach.
• It is hypothesized that texture descriptors that
  emanate from and highly correlated with human
  vision system provide the utmost representation,
  and thus yield a more contextual systema system
  that closely emulate human expert

45

• Item one skin has distinct uniform pattern
  (glyphic pattern).
• MM disrupts texture.
• Quantifying texture differences between MM and NV
  is more reliable method than color-based (
  color-based in prone to variability in imaging
  system)
• Item two texture come in different sizes.
• Detection method that explore texture image at
  different possible scale is more sensitive than
  methods that are using one scale.
• Example of this snake-based ( gradient-based),
  Normalized Cut, histogram threshold

46

• Item three there are many texture descriptors
  that are purely algorithmic that may not
  necessary correlate with human vision.
• One example is first-order statistics of texture
  ( variance ,mean), structure-based approach,
  laplacain of Gaussian.
• Texture classifiers that emanate from or highly
  correlated with human visual system provides a
  closer approximation of experts perception of
  texture.

47
Visual Texture
48
Texture
Technical Definition

• Texture is regarded as what constitutes a
  macroscopic region. Its structure is simply
  attributed to the repetitive patterns in which
  elements or primitives are arranged according to
  a placement rule(Tamura et al, 1978).
• Texture is both the number and types of its
  (tonal) primitive and their spatial arrangement
  (Haralick ,1979).
• The term texture generally refers to repetition
  of basic texture elements called texels. The
  texel contains several pixels, whose placement
  could be periodic, quasi-periodic, or random.
  Natural textures are generally random, whereas
  artificial textures are often deterministic or
  periodic. Texture may be course, fine, smooth,
  granulated, rippled, regular, irregular, or
  linear (Jain, 1989).
• Texture is intuitively viewed as descriptor in
  providing a measure of properties such as
  smoothness, coarseness, and regularity (Gonzales
  and Woods, 1990).
• Texture is an attribute representing the spatial
  arrangement of the gray levels of the pixels in a
  region (IEEE, 1990).
• Texture is both grey level of a single pixel and
  its surrounding pixels, which was coined as a
  unit texture, texels. These texels conformed
  repetitive patterns that dictated the effective
  texture analysis approach (Karu et al, 1996).
• Patterns which characterize objects are called
  texture in image processing (Jähne, 2005).

49

1. Texture has no single definition.
2. Definitions from previous literature dedicated in
   studying texture
3. The first three definitions, tells us texture is
   composed of a building block that is spatially
   arranged based on the placement rule (periodic,
   quasi periodic, or random) like a brick a single
   brick is the building block, the arrangement of
   the bricks that gives rise to a texture of a
   brick wall
4. Texture is descriptors for smoothness,
   coarseness, and regularity
5. In computer vision
6. Spatial arrangement of gray levels of the pixel
7. Pattern

50
Texture and Human Vision System

• Pre-attentive visual system-1962-1981
• Dr. Julesz
• Neuroscientist
• Texture perception
• Statistical approach
• Disproved conjecture that second-order is
  processed in the vision system
• Textons
• Contrast
• Terminator-end of lines, corners
• Elongated blobs of different sizes - granularity

51

1. As one of the hypothesis. Texture
   characterization emanate from visual system
   closely emulates experts
2. Neuroscientist, studied perception of texture
3. Before disproving, he conjectured that
   second-order statistics is processed in the
   vision system, and He claimed that two textures
   with similar second-order statistic is not
   pre-attentively recognizable.
4. In other words without close inspections, two
   different texture with same sec stat would seem
   to look similar.
5. After series of experiments, he finally suggested
   that textons are the major player for texture
   discrimination.
6. And the textons are contrast, terminators.
   granularity

52
Texture discrimination
Textons instead of second-order statistics that
cause the texture discrimination
Textons
Second-order statistics
53
The image on left is an example of two different
textures with the same SO that is not
pre-attentively detectable.The right image is
two different textures with the same SO but
pre-attentively detectable. Among others this
leads to the final statement texture
discrimination is made possible through the
textons. Here in this one is the difference
termination of the two texture elements .In
this work, the second-order statistics CoM and
contrast of edge elements will be explored for
extracting visual texture properties of skin
lesion.
54
Texture and Human Vision System

• Frequency and Orientation
• Multi-frequency and orientation analysis
• decomposition (1968) Campbell and Robson
• Simple cells of the visual cortex respond to
  narrow ranges of frequency and orientation, cells
  act as 2D spatial filter-(1982) De valois et al.
• Orientation-based texture segregation involves
  the generation of a neural representation of the
  surface boundary whose strength is nearly
  independent of the magnitude of orientation
  contrast - Motoyoshi and Nishida (2001)

55

• More studies had been conducted in part to
  understand human vision.
• This Campbell and Robson found that when signal
  received by the eye is decomposed into multiple
  frequencies and orientation
• Another work in the subsequent year that further
  support the previous finding that simple cells
  are highly selective/tuned to narrow frequency
  and orientation.
• Another work found that neural representation of
  texture boundary is formed that is independent of
  magnitude and orientation of the contrast
• In this work in wavelet analysis will be used for
  segmentation. Frequency and contrast

56
(No Transcript)
57
Texture
58
(No Transcript)
59
(No Transcript)
60
(No Transcript)
61
Method Design

• Creation of relative color images
• Segmentation and morphological filtering
• Relative color feature extraction
• Design of tumor feature space and object feature
  space
• Establishing statistical models from relative
  color features

62

• COLOR

63
Create Relative Color Skin Tumor Images

• Purpose
• to equalize any variations caused by lighting,
  photography/printing or digitization process
• to equalize variations in normal skin color
  between individuals
• the human visual system works on a relative color
  system
• Algorithm
• Mask out non-skin part in the image to calculate
  the normal skin color
• Separate tumor from the image
• Remove the skin color from the tumor to get a
  relative color skin tumor image
• CVIPtools functions were used to create relative
  color skin tumor images

64
Calculate Skin Color
65
Tumor Image
66
Relative Color Tumor Image
67
Segmentation and Morphological Filtering

• Image segmentation was used to find regions that
  represent objects or meaningful parts of objects
• Morphological filtering was used to reduce the
  number of objects in the segmented image
• Easy to use CVIPtools for experimenting and
  analysis

68
Feature Extraction
69
Relative Color Feature Extraction

• Necessary to simplify the raw image data into
  higher level, meaningful information
• Feature vectors are a standard technique for
  classifying objects, where each object is defined
  by a set of attributes in a feature space.
• Totally 17 color features and binary features
  were extracted using CVIPtools
• The three largest objects, based on the binary
  feature area, were used in feature extraction
• Histogram features, that is, color features, were
  extracted in each color band from relative color
  image objects

70
17 Features

• Histogram features in R, G, B bands
• Mean
• Standard deviation
• Skewness
• Energy
• Entropy

• Binary features
• Area
• Thinness

71
17 Features (Cont.)
72
Design Two Feature Spaces

• Tumor feature space
• consists of 277 feature vectors correspond to 277
  skin tumor images.
• each feature vector has 51 feature elements,
  which are the total of 17 features of each three
  largest objects within the same tumor.
• Object feature space
• had 842 feature vectors corresponding to 842
  image objects
• each feature vector has 17 feature elements,
  which were the binary features and color features
  stated as above

73
Establishing Statistical Models

• Two feature spaces serve as two data models in
  order to maximize the possibility of success
• Two classification models, Discriminant Analysis
  and Multi-layer Perceptron, were developed for
  both data models
• The training and test paradigm is used in
  statistical analysis to report unbiased results
  of a particular algorithm
• due to small size of data set, 282 images, we
  used the leave x out method, with both one and
  ten for x
• Partek software was used
• to analyze the data representing the features
• to develop a model or rules for classifying the
  tumors

74
Quadratic Discriminant Analysis

1. A statistical pattern recognition technique based
   on Bayesian theory, which classifies data based
   on the distribution of measurement data into
   predefined classes
2. Normalization the feature data as preprocessing
3. performed to maximize the potential of the
   features to separate classes and satisfy the
   requirement of the modeling tool such as
   Quadratic discriminant analysis for a Bayesian
   distribution of the input data
4. Variable selection was used to choose dominant
   features.

75
Multi-Layer Perceptron

• A feed forward neural network
• neural networks modeled after the nervous system
  in biological systems, based on the processing
  element the neuron
• widely used for pattern classification, since
  they learn how to transform a given data into a
  desired output.
• Principal Component Analysis (PCA) as
  preprocessing
• a popular multivariate technique, is to reduce
  dimensionality by extracting the smallest number
  components that account for most of the variation
  in the original multivariate data and to
  summarize the data with little loss of
  information
• the dispersion matrix selected for PCA in this
  project is correlation

76
Multi-Layer Perceptron (Cont.)

• Creation, training and testing of neural
  networks
• Creation a neural network involves selection of
  hidden and output neuron types and a random
  number generation.
• Four output neuron types Softmax, Gaussian,
  Linear and sigmoid
• Three hidden neuron types Sigmoid, Gaussian and
  Linear
• Scaled Conjugate Gradient algorithm is used for
  learning in this project.
• Automated and independent of user parameters
• Avoids time consuming
• Stopping criteria, sum-squared error, is selected
  to determine after how many iterations the
  training should be stopped
• The trained data is then tested on itself first
  to examine how far the neural network is able to
  classify the objects correctly.
• Leave x partition out method is used for testing
  the algorithm

77
Experiments and Analysis in Object Feature Space

• Discriminant Analysis
• 8, 9, 11 and 12 significant features were
  selected respectively for leave one out method

Number of Histogram Features Area Mean Mean Mean STD STD STD Skewness Skewness Skewness Energy Energy Energy Entropy Entropy Entropy
Number of Histogram Features Area R G B R G B R G B R G B R G B
8 X X X X X X X X
9 X X X X X X X X X
11 X X X X X X X X X X X
12 X X X X X X X X X X X X
78
Experiments and Analysis in Tumor Feature Space

• Discriminant Analysis
• 24 features selected for leave ten out method

Histogram Features Mean Mean Mean STD STD STD Skewness Skewness Skewness Energy Energy Energy Entropy Entropy Entropy
Histogram Features R G B R G B R G B R G B R G B
Object 1 X X X X X X X X
Object 2 X X X X X X X
Object 3 X X X X X X X X

• 10 features selected for leave one out method

Histogram Features Mean Mean Mean STD STD STD Skewness Skewness Skewness Energy Energy Energy Entropy Entropy Entropy
Histogram Features R G B R G B R G B R G B R G B
Object1 X X X
Object 2 X X X X
Object 3 X X X
79
Experiments and Analysis in Tumor Feature Space
(Cont.)

• Discriminant Analysis (Cont.)

80
Experiments and Analysis in Tumor Feature Space
(Cont.)

• Multi-layer Perceptron

• Best features, being in the first three
  components of the PCA projection data, were used
• Success percentages of melanoma as high as 77
  and nevus is as high as 68

81
Experiments and Analysis in Object Feature Space
(Cont.)

• Discriminant Analysis (Cont.)

• Yield consistent results in classifying melanoma
  from other skin tumor with above 80 success rate

82
Experiments and Analysis inObject Feature Space
(Cont.)

• Multi-layer Perceptron (MLP)

• 5 out of 12 hidden-output layer neuron
  combinations gave better classification results
• Leave one out method
• Yield success percentage as high as 86 for
  classifying melanoma.
• MLP is more consistent in classifying melanoma as
  well as nevus

83
Conclusion

• Multi-Layer perceptron (MLP) with feature data
  preprocessed by Principal Component Analysis
  (PCA) gave better classification results for
  melonoma than Discriminant Analysis (DA)
• The best overall successful rate of 78, of which
  percentage correct of melanoma is 86, nevus is
  62 and dysplastic is 56.
• The best classification results are achieved with
  sigmoid used as the hidden and output layer
  neuron type for the MLP with PCA on Object
  Feature Space.
• The three largest tumor objects are
  representative for the whole skin tumor.

84
Conclusion (Cont.)

• However the small percentage of melanoma
  misclassification as well as the relatively low
  success rate for nevus and dysplastic nevi
  suggests that we may not have the complete data
  set for the experiments.
• In order to achieve better classification
  results, future experiments
• Needs more complete skin tumor image database.
• Should combine texture and color methods to get
  better results
• Will include dermoscopy images

85
(No Transcript)
86
Acknowledgement

• Dr. Scott E Umbaugh, SIUE
• Mr. Ragavendar Swamisai
• Ms. Subhashini K. Srinivasan
• Ms. Saritha Teegala
• Dr. William V. Stoecker, Dermatologist, UMR

87
Thank You!
Yue (Iris) Cheng Graduate Student _at_ Computer
Vision and Image Processing Research
Lab Electrical and Computer Engineering
Department Southern Illinois University
Edwardsville E-mail cheng_at_westar.com https//www.
ee.siue.edu/CVIPtools
88
(No Transcript)
89
CLASSIFICATION OF MALIGNANT MELANOMA AND
DYSPLASTIC NEVI USING IMAGE ANALYSIS A VISUAL
TEXTURE APPROACH

• Dr. Dinesh Mital

University of Medicine and Dentistry of New
Jersey School of Health Related
Profession Biomedical Informatics March 2009
90
Color-based Diagnosis Clinical Images

• Research Project Funded In Part by NIH

Yue (Iris) Cheng, Dr. Scott E Umbaugh _at_ Computer
Vision and Image Processing Research
Lab Electrical and Computer Engineering
Department Southern Illinois University
Edwardsville E-mail cheng_at_westar.com https//www.
ee.siue.edu/CVIPtools
91
Spatially Constrained Segmentation of Dermoscopy
Images

• Howard Zhou1, Mei Chen2, Le Zou2, Richard Gass2,
• Laura Ferris3, Laura Drogowski3, James M. Rehg1

1School of Interactive Computing, Georgia
Tech 2Intel Research Pittsburgh 3Department of
Dermatology, University of Pittsburgh