Detection of Malaria by Multispectral Microscopy using Statistical Classification Methods
(2013) FMS820 20132Mathematical Statistics
- Abstract (Swedish)
- There has been much work on classification of malaria infected blood; in resent time, a method using
LED-based microscopy has been developed with the goal of reducing time and cost. The education level
needed to make such decisions is also reduced using this microscope. This is mainly done to help the
developing countries in the fight against malaria and develop these countries competence in the field of
multispectral analysis. The LED-microscope used, was constructed during a workshop including scientists
from Lund and 6 developing countries in Africa, so there is identical equipment in the field in these
countries. This could be a useful complement to the pathologists in the field. The LED- microscope uses
13 different wavelengths... (More) - There has been much work on classification of malaria infected blood; in resent time, a method using
LED-based microscopy has been developed with the goal of reducing time and cost. The education level
needed to make such decisions is also reduced using this microscope. This is mainly done to help the
developing countries in the fight against malaria and develop these countries competence in the field of
multispectral analysis. The LED-microscope used, was constructed during a workshop including scientists
from Lund and 6 developing countries in Africa, so there is identical equipment in the field in these
countries. This could be a useful complement to the pathologists in the field. The LED- microscope uses
13 different wavelengths and 3 ways to illuminate the sample (reflecting, scattering and transmitting).
Each combination is used, all in all 39 different pictures of the sample. An automatic process based on
this would be a great help to simplify the detection of malaria.
The goal of this thesis is to analyze different methods of classifying malaria using data from the
LED-microscope. The data was collected by already existing, and under development, methods using
the LED-microscope. The main two statistical methods used in this thesis to do the classification are:
First Fisher’s Linear discriminant to reduce the dimensions with minimal information loss. Second Ellipsoidal
constraints to formulate an optimization problem that is then rewritten as a convex optimization
problem, which is then solved. Then the classification is analyzed to conclude if the samples contain
malaria, and to find suitable thresholds for the classifiers, every analyzed method is evaluated. Even
when this is mainly software optimization it can impact the work needed to construct new microscopes
to make it more efficient. Also contained in this thesis are some examples from analysis of both fresh and
old malaria-infected blood samples to see the difference between the different methods. The methods is
quite general and can with little extra work be applied to different data sets, when the microscope is used
to gather data from other sources than blood samples. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/4238046
- author
- Arnström, Oskar
- supervisor
- organization
- course
- FMS820 20132
- year
- 2013
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 4238046
- date added to LUP
- 2014-01-09 10:51:30
- date last changed
- 2014-01-09 10:51:30
@misc{4238046,
abstract = {{There has been much work on classification of malaria infected blood; in resent time, a method using
LED-based microscopy has been developed with the goal of reducing time and cost. The education level
needed to make such decisions is also reduced using this microscope. This is mainly done to help the
developing countries in the fight against malaria and develop these countries competence in the field of
multispectral analysis. The LED-microscope used, was constructed during a workshop including scientists
from Lund and 6 developing countries in Africa, so there is identical equipment in the field in these
countries. This could be a useful complement to the pathologists in the field. The LED- microscope uses
13 different wavelengths and 3 ways to illuminate the sample (reflecting, scattering and transmitting).
Each combination is used, all in all 39 different pictures of the sample. An automatic process based on
this would be a great help to simplify the detection of malaria.
The goal of this thesis is to analyze different methods of classifying malaria using data from the
LED-microscope. The data was collected by already existing, and under development, methods using
the LED-microscope. The main two statistical methods used in this thesis to do the classification are:
First Fisher’s Linear discriminant to reduce the dimensions with minimal information loss. Second Ellipsoidal
constraints to formulate an optimization problem that is then rewritten as a convex optimization
problem, which is then solved. Then the classification is analyzed to conclude if the samples contain
malaria, and to find suitable thresholds for the classifiers, every analyzed method is evaluated. Even
when this is mainly software optimization it can impact the work needed to construct new microscopes
to make it more efficient. Also contained in this thesis are some examples from analysis of both fresh and
old malaria-infected blood samples to see the difference between the different methods. The methods is
quite general and can with little extra work be applied to different data sets, when the microscope is used
to gather data from other sources than blood samples.}},
author = {{Arnström, Oskar}},
language = {{eng}},
note = {{Student Paper}},
title = {{Detection of Malaria by Multispectral Microscopy using Statistical Classification Methods}},
year = {{2013}},
}