Friday, July 29, 2016

Entropy-based feature extraction and decision tree induction for breast cancer diagnosis with standardized thermograph images

Entropy-based feature extraction and decision tree induction for breast cancer diagnosis with standardized thermograph images

by
Ming-Yih Leea,
Chang Gung University, Graduate Institute of Medical Mechatronics

and

Chi-Shih Yanga,
Lee-Ming Institute of Technology, Department of Mechanical Engineering

a b s t r a c t

In this study, a computer-assisted entropy-based feature extraction and decision tree induction
protocol for breast cancer diagnosis using thermograph images was proposed. First,
Beier–Neely field morphing and linear affine transformationwere applied in geometric standardization
for whole body and partial region respectively. Gray levels of pixel population
at the same anatomical position were statistically analyzed for abnormal region classification.
Morphological closing and opening operations were used to identify unified abnormal
regions. Three types of 25 feature parameters (i.e. 10 geometric, 7 topological and 8 thermal)
were extracted for parametric factor analysis. Positive and negative abnormal regions
were further reclassified by decision trees to induce the case-based diagnostic rules. Finally,
anatomical organ matching was utilized to identify the corresponding organ with the positive
abnormal regions. To verify the validity of the proposed cased-based diagnostic protocol,
71 and 131 female patients with and without breast cancer were analyzed. Experimental
results indicated that 1750 abnormal regions (703 positive and 1047 negative) were detected
and 822 branches were broken down into the decision space. Fourteen branches were found
to have more than 4 positive abnormal regions. These critical diagnostic paths with less
than 10% of positive abnormal regions (61/703 = 8.6%) can effectively classify more than half
of the cancer patients (42/71 = 59.2%) in the abovementioned 14 branches.

1. Introduction

Thermographs were designed to capture the infrared ray from
objects if its temperature were higher than the absolute zero
degree. This special characteristic was utilized to display the
metabolic heat that was dissipated from the skin surface
of human body, especially for the medical thermographs of
breast cancer. The pathological mechanisms of breast cancer
are that cancer cells produce nitric oxide (NO) at proliferative
stage [1]. This chemical material will interfere with the
neuronal control of blood vessel flow and lead up to regional
vasodilatation in the early stage of cancerous cell growth. This
angiogenesis facilitates a local temperature rise about several
years earlier than the tumor forming. And even deep
breast lesions seem to have the ability to induce changes
in skin temperature [2]. The abovementioned characteristics
of breast cancer tissue specifically imply some meaningful
graphic patterns (i.e., geometric size and location, topological
shape or thermal features) in thermograph images. Hence
thermographs are better suited than mammography for early
breast cancer prediction even when the tumor is taking shape.

Medical thermographs become a suitable tool for early warning.
This key point of this paper emphasizes the usage of
thermal image to diagnose breast cancer. The algorithm with
high concentration rate of cancer patients in a few rule paths
was proposed. The effect of early diagnosiswas not in the field
of discussion.

Besides, there exists highly false positive prediction due
to the little difference of thermographs between the normal
subject (negative case) and cancer patient (positive case) as
shown in Fig. 1. Due to the lack of effective image processing
methods to support the use of thermographs, the credibility
and sensitivity of thermographs are still in question in
medical applications. Medical thermograph has been quiescent
for a long time due to the limitation of hardware and
software. The hardware limitation was broken recently by the
advancement of the uncooled focal plane array (UFPA) photo
detector and personal computer, and the software limitation
was also overcame by the progress of analyzing algorithms. All
the signs indicate that the infrared thermal images will be successful
in medical usage [3]. Recently, large scale of case-based
studies showed that thermographs had an average sensitivity
and specificity of 90% as the significant biological risk marker
for the existence of breast tumor [4]. But all these analytical
techniques still come to a standstill with the traditional statistics
and manual evaluation. The parametric analysis on ROI
(region of interest) of hot spots and cold spots in the abnormal
regions still remain unsolved.

Different medical applications with thermographs were
received attention from various research groups. To name a
few, female breast cancer [4,5], pain management [6] or diabetic
foot [7]. Also various digitally analytical methods for
image processing algorithms were proposed, such as abnormal
statistics method [8], the thermal asymmetric method
[9,10] and the image standardization and anatomical matching
methods [11], etc. Although these image processing
techniques are important in medical thermograph analysis,
lots of these methods still lack delicate algorithms or cannot
totally be a part of the computer-assisted diagnosis solution.

The aim of this paper is to establish a computer-aided
cased-based diagnostic tool to automatically interpret the
thermal pattern by a bio-statistical technique [4] for investigating
the parametric connection between female breast
cancer and thermal physiology quantitatively. The proposed
algorithm consists of five steps, i.e. geometric lofting
standardization, abnormal region statistics, parametric factors
analysis, decision tree induction and anatomical organ
matching. This methodology is not intended to replace the
traditionally golden method on mammography and also not
to compare the minimal tumor diameter missed by these two
instruments. This paper is intended to explore the potential
benefits of thermograph for early detection of breast
cancer without comparing its sensitivities or specificities
with that of mammography. In addition, a parametric analytical
algorithm was proposed to extract the governing
rules from large scale dataset by decision tree induction
for breast cancer. Finally, the clinical applicability of
the proposed computer-assisted diagnosis tools was analyzed
by 71 and 131 cases with and without breast cancer
respectively.

No comments:

Post a Comment

Warning !!!

=> Please leave a comment polite and friendly,
=> We reserve the right to delete comment spam, comments containing links, or comments that are not obscene,
Thanks for your comments courtesy :)