ABSTRACT:

A noninvasive computer-aided-theragnosis (CAT)system was developed for the

early therapeutic cancer responseassessment in patients with locally advanced breast
cancer(LABC) treated with neoadjuvant chemotherapy. The proposedCAT system
was based on multi-parametric quantitativeultrasound (QUS) spectroscopic methods
in conjunction with advancedmachine learning techniques. Specifically, a
kernel-basedmetric named maximum mean discrepancy (MMD), a techniquefor
learning from imbalanced data based on randomundersampling, and supervised
learning were investigated withresponse-monitoring data from LABC patients. The
CAT systemwas tested on 56 patients using statistical significance tests
andleave-one-subject-out classification techniques. Textural featuresusing
state-of-the-art local binary patterns (LBP), and gray-scaleintensity features were
extracted from the spectral parametricmaps in the proposed CAT system. The system
indicated significantdifferences in changes between the responding
andnon-responding patient populations as well as high accuracy,sensitivity, and
specificity in discriminating between the twopatient groups early after the start of
treatment, i.e., on weeks 1and 4 of several months of treatment. The proposed CAT
systemachieved an accuracy of 85%, 87%, and 90% on weeks 1, 4 and8, respectively.
The sensitivity and specificity of developed CATsystem for the same times was 85%,
95%, 90% and 85%, 85%,91%, respectively. The proposed CAT system thus
establishes anoninvasive framework for monitoring cancer treatment responsein
tumors using clinical ultrasound imaging in conjunction withmachine learning
techniques. Such a framework can potentiallyfacilitate the detection of refractory
responses in patients totreatment early on during a course of therapy to enable
possiblyswitching to more efficacious treatments.

EXISTING SYSTEM:

Currently, the most established methods for cell death detectionare those focussed
on the detection of apoptosis (programmedcell death). These methods could
potentially beused in a clinical setting and are based on imaging modalitiessuch as
positron emission tomography (PET), single photonemission computed tomography
(SPECT), and magneticresonance imaging (MRI). The previous application of QUS
methods in the evaluation of cell death response was mainly limited to analyses such
as statistical tests of significance, regression, and discriminant analysis in order to
demonstrate the proof of principle of theproposed QUS systems on preclinical.

DISADVANTAGES OF EXISTING SYSTEM:

 Despite recent advances in the early detection of breast cancer, it still is often not
being discovered until it is fairly large or locally advanced.

 Changes in size can take weeks to months after the completion of treatment to
become apparent, and do not alwaysoccur even when the treatment is effective.

 They are presently expensive, and requirecontrast agents to be administered in

order to enhance the contrastfrom soft tissue. The potential for side effects from
allergicreactions and the radioactivity of some agents are also limitingfactors.

PROPOSED SYSTEM:

This work represents one of the first studies to apply machine learning techniques
such as kernel-based methods and supervised learning on QUS imaging for cancer
response monitoring in clinical applications such as LABC. It has been shown that
using MMD can significantly improve the classification performance compared with
simple metrics such as norm on both feature sets, histogram of intensity and LBP, as
well as their combination. Combining the two feature sets further improved the results
of classification in the work here. Data collected from each patient during treatment
on weeks1, 4 and 8 was independently compared to baseline, which was the
“pre-treatment” data for the same patient.In addition to the performance evaluation of
the proposed CAT system to correctly classify the patients to res-ponders and
non-res-ponders,a statistical test of significance was performed between responders
and non-responders at each treatment time to assess whether the two populations
exhibit any significant differences in changes for each particular measured
QUS-based feature-distance combination.
ADVANTAGES OF PROPOSED SYSTEM:

Combining the two features sets was beneficial as it enhanced the significance of
changes between the R and NR populations using MMD as the metric. This
observation also indicated that both textural and gray-scale intensity information were
useful in the discrimination of the two populations in cancer response monitoring.

Between the two spectral parametric maps, MBF was more discrimination between
the two populations as it led to significant changes being detected between the two
populations on week 1 for Hist Int and combined feature sets with MMD as the
dissimilarity measure, whereas the intercept did not detect such significant changes on
week 1.

The proposed CAT system can progressively monitor and classify treatment response
early after the start of neo adjuvant chemotherapy adminstration in locally advanced
breast cancer patients non-invasively and with high accuracy.

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

System : Pentium Dual Core.

Hard Disk : 120 GB.

Monitor : 15’’LED

Input Devices : Keyboard, Mouse

Ram :

SOFTWARE REQUIREMENTS:

Operating system : Windows 7.

Coding Language : MATLAB

Tool : MATLAB R2013A