DETECTION OF BREAST CANCER USING VARIOUS

SEGMENTATION ALGORITHM

A. Anjaline Jayapraba,
Assistant Professor, Saveetha Engineering College, Chennai
Students, Saveetha Engineering College, Chennai

ABSTRACT
Breast cancer ranks as the most prevalent form of invasive cancer and stands as
the second leading cause of cancer-related fatalities among women. Early detection can
effectively reduce the danger and threads. The Mammogram is widely used in early screening
methods but it is relatively low considering to other new techniques., Computer-aided
diagnosis (CAD) is a system employed for the purpose of both detection and subsequent
processing,there are many algorithms is there to detect they are CNN,ANN,R-CNN,DBN,
Fuzzy algorithm, decision tree algorithm and many more. They are reviewed and finally we
use the most efficient and give high accuracy in the output. The high precision level
algorithms which defines the output state will be verified in this review paper.
INTRODUCTION
Breast cancer is a form of cancer that originates in the cells of the breast, breast cancer is
second common cancer after skin cancer. Mammogram is the screening method to detect
cancer this diagnostic method to detect calcium deposit in breast it is the general screening
method. It indicate lumph, nipple discharge or changes in breast size and shape. This signs
are benign condition, mammogram are two different methods one screening and another one
diagnostic method. Diagnostic mammogram takes more time when compare to screening
mammogram because dose of radiation is higher to take more x-ray images are needed to
take image in various angle it helpful to doctor identify the accurate place where breast
cancer is palced. Clinical trial shows the motality rate reduced in screening method from
women ages from 30 to 70 years. When abnormal mammogram image shown means another
technique radiologist, ultrasound or biopsy taken whether the cancer is present. FPR is come
for women who have dense breast, family history and women who taking estrogen. More
than 50% of women in united stated have FPR in past ten years. Based on lump nodes, tumor
size and metastasis cancer stages from 0 to 4th stages are classified distinguish from American
Joint Committee.
Stage 1: The diameter of the tumor is 2 cm, indicating a small size.
Stage 2: The diameter of the tumor is 2-5 cm, indicating a small size.
Stage 3: The tumor measures over 5 cm in diameter
Stage 4: The tumor is any size they develop in lump nodes and chest wall
In 2005 the ACRIN society they compare film mammogram to digital mammogram for
women with dense breast digital mammogram better and women younger than 50 age group.
In mammogram they catogories into ABCD when ‘A’ represent fat tissue and ‘D’ represent
glandular tissue, in that x-ray image white indicate glandular tissue and dark represent fat
tissue.
LITERATURE REVIEW
Convolution Neural Networks for Breast cancer diagnosis using diffusion Optical
Tomography
This paper addresses the detection of breast cancer through the application of diffuse optical
tomography (DOT). Among fuzzy logic Bayesian networks , decision tree and k means
clustering . they use DOT with CAD methods since it has screening dense breast with safety,
cost efficiency etc.For classification support vector machine is used. CNN is used in this
technique and It will be modified to leverage the 2D structure of the input images. The
acquired 3D images are grayscale images sized at 91x91x31 .then it is given as a input
for CNN with training and testing,75%and 25%. The trained dataset with dense breast with
images itself will radiate certain amount , then will be finding out the tumorous tissues .
pooling layers conduct local averaging and subsampling, resulting in a decrease in feature map
resolution and reduced sensitivity of the output to shifts and distortions. Subsequently, a
straightforward logistic classifier is employed to detect using binary bits, represented as (1,0) and
(0,1). it will detect benign and malignant feature. Feature
Extraction,accuracy93.3%,specification-0.88%,sensitivity-0.96.and AUC is 0.94.
A novel deep learning based framework for the detection and classification of breast cancer
using transfer learning.
This paper explores the application of Transfer learning in Breast cancer detection, utilizing
two approaches: mammograms and biopsies. While CNNs are effective on large datasets,
Transfer learning is favored for cost reduction and higher accuracy. During preprocessing,
tissue images are cleaned to eliminate various forms of noise. In the segmentation phase,
preprocessed datasets from Resnet and GoogleNet are combined. Initially, distinct CNN
architectures handle feature extraction, which is later consolidated into a fully connected
layer for classification tasks. VGGnet, a model resembling AlexNet but with added
convolution layers, is implemented, featuring 13 convolution layers, rectification, pooling
layers, and 3 fully connected layers. After processing all data with the specified features,
feature extraction achieves an accuracy of 97.52% and a precision of 0.95.
Deep learning application to breast cancer detection by magnetic resonance imaging.
This paper is dedicated to the utilization of Magnetic Resonance Imaging (MRI) for breast
cancer detection. MRI is known for its heightened sensitivity and ability to provide more
detailed pathophysiological information, although it is less cost-effective when compared to
mammography for population-based screening. Within the paper, the input layer represents
the breast cancer image, serving as the input to the Convolutional Neural Network (CNN).
The design involves stacking multiple convolutional layers on top of the input layer, where
each convolutional layer employs filters or kernels to extract specific features from the input
image. Subsequent to some of the convolutional layers, pooling layers are incorporated to
downsample the spatial dimensions of the feature maps, with common pooling techniques
including max-pooling and average pooling. In the classification phase, the study involves a
comprehensive comparison of different deep learning algorithms, analysis methods, types of
ground truths, sample sizes, the quantity of benign and malignant lesions, MRI image types,
and performance metrics, among other factors. The feature extraction process ultimately
yields an accuracy of 0.86, a specificity of 0.85, and an AUC of 0.87.
Deep learning to improve breast cancer detection on screening mammography
This journal is focused on the detection of breast cancer through the use of Screening
Mammography. During the preprocessing stage, a full-field digital mammography (FFDM)
image typically has dimensions of 4000x3000 pixels, while a potentially cancerous region of
interest (ROI) can be as small as 100x100 pixels. In their approach, they propose an end-to-
end methodology in which a model for classifying local image patches is pre-trained using a
fully annotated dataset that includes ROI information.A modern Convolutional Neural
Network (CNN) is typically built by layering convolutional layers on top of the input,
followed by one or more fully connected (FC) layers to connect to the classification output.
Max-pooling layers are often incorporated with convolutional layers to enhance translational
invariance and reduce feature map size. Whole image classification from patch classifiers
involves flattening the heatmap and connecting it to the image's classification output using
FC layers.Regarding feature extraction, the achieved results include an accuracy of 0.83 with
25% of the epochs completed.
Deep learning model for fully automated breast cancer detection system from thermograms.
This journal delves into the detection of breast cancer using thermograms, a radiation-
inclusive, low-cost, non-intrusive, and non-invasive technique. Therefore, it can effectively
identify early-stage breast cancer in young women and individuals with dense breasts.
Thermography is based on the concept that all living bodies emit infrared (IR) radiation
above absolute zero.
In the preprocessing phase, images are captured from different angles to detect the emitted
infrared rays in the body. Segmentation is a crucial technique for isolating the breast region
from other parts of the body in thermal images, ensuring that the extracted region includes all
breast tissues, ducts, lobules, and lymph nodes.In the classification process, the RELU layer
is favored in CNN architectures due to its capacity to expedite the training process. Max-
pooling layers are employed to reduce parameter size and control overfitting. Neurons in the
fully connected layer function similarly to a regular neural network. Dropout layers are
utilized to prevent overfitting. Feature extraction achieves remarkable results with an
accuracy of 99.33%, sensitivity of 100%, and specificity of 98.67%
Multi-input deep learning approach for breast cancer screening using thermal infrared
imaging and clinical data.
This paper addresses the detection of breast cancer through the utilization of thermal infrared
radiations. In the preprocessing phase, images are acquired at diverse angles employing a
FLIR SC-620 thermal infrared camera. These images are subsequently transformed into
grayscale utilizing the DIT and SLT protocol.then considering three types of factors Risk
factor ,complementary features ,protocol features. Segmentation of images to get Range of
Interest. and then using CAD software with the extraction features, images are denoising
through anisotropic diffusion. Edges were identified through the use of the Canny edge
detector, followed by the recognition of points along the infra-mammary line in the lower
breast region using an image acquisition protocol. The classification approach employed a
Hybrid Deep Neural Network (DNN). Pretrained layers Alex Net and Google Net. simple
neural networks is used for the clinical data. then finding correlation between different
thermograms. images has curves edges etc..will be down sampled and reduced their
dimensionality. SGDM as optimizer. the feature extractions like accuracy 97%,Sensitivity
83%,Specifications 83.33.with their respected angles.
Breast cancer classification from histopathological images using patch-based deep learning
modelling.
This paper is used to detect Breast cancer using Patch Based deep learning modelling.in
preprocessing cropping the Images to focus on Rande of interest the convert the images into
RGB coloured images or convert to grey scale images in which the values ranges from 40 to
215 is a cancerous region then smoothening the images using Gaussian filter.In Segmentation
for creating patch, Overlapping patch size 32*32.then randomly shuffle the patch data with
the ratio of 1:1.convertconvert to vector of size 1024is in the input for PA-DBN-BC model.
This process involves three layers and in that three two are hidden layers.by NN the labeled
datasets are trained and output is taken. classification process the patch images created and
trained then it is classified into positive term or negative term. feature extraction accuracy-
86%,error ratio47%,sensitivity-87.9%,specificatiions-84%.
A Novel deep learning model for automatic detection and classification of breast cancer using
the transfer learning technique
This paper is used to detect Breast cancer using transfer learning technique. In preprocessing
a 2D median filter of 3*3 is used for image cropping.Removal of Digitalization noise from
mammogram image is the primary process. Making Histogram equal which makes contrast
level of the original image. convert into greyscale image then automatic patch extraction and
convert to RGB to match the input size. Data augmentations algorithm used for increases the
size of the dataset.VGG19,VGG16,InceptionV2 is used for technique detection and it is
trained by image net dataset. classification process pre-trained network is used. trained and
pre-trained datas combined to CNN.The extracted features are subsequently utilized to train a
Support Vector Machine (SVM) and SVM classifier. In terms of feature extraction, this paper
achieved an accuracy of 98.96%, a sensitivity of 97.83%, a specificity of 99.13%, and a
precision of 97.35.
Deep Neural Networks improves Radiologistics performance in breast cancer screening
This paper is used to detect the breast cancer Radiologistics techniques using deep neural
networks. The preprocessing work is to separate the images with the Range of
interest(ROI).and those images are adjusted to very hight resolution images ResNet-22
Networks then use a wise model to check the status then without down sampling forward
pass and gradient within GPU memory after this training the model with Adam optimization
algorithm.In segmentation Full resolution images as input 256*256 patch pixels then these
patch level converts into breast level and then creates heat maps for segmentation R-CNN is
used.classificationBI-RADS which is a radiologistic detection is used for detecting.it goes
with three types BI-RADS[0],BI-RADS[1]and BIRADS[3],here the third will be considered
as benign. Then the first two will be again checked using NN for reasonable level
performance .the final output will be accuracy-0.867% and PRAUC-0.318.
A Systematic Review of Nreast cancer detection using thermography and neural networks.
This paper is used to detect breast cancer using various properties. In pre processing,with
FLIR SC-620 thermal infrared camera images are taken in different angle.those infrared
caneras has 640*480 pixels and it creats heat map and matrix with 640*480.each patient
undergoes thermal stress.in Segmentation process ANN is reccomanted for usage it has two
types BPI and RBFN . here RBFN is used, varioscan 3021-ST is used for monitoring the
different temperatures then MRI is used to confirm the cancer.In classification the methods
are more but here DT(digital tree and fuzzy used to get high accuracy.feature extraction for
this journel will be accuracy-90.48,specification89.73,sensitivity-87.
Breast cancer detection using extreme learning machine based on feature fusion with CNN
deep features. his paper employs Convolutional Neural Networks (CNNs) for image-based
tasks, such as breast cancer detection, owing to their capability to capture intricate patterns.
The feature extraction process entails the utilization of pre-trained CNN models to extract
deep features from breast cancer images.Feature fusion combines deep features with other
relevant data, such as patient information, to enhance the model' s performance.Data
preprocessing involves cleaning and standardizing the breast cancer images to ensure
consistency , Data augmentation techniques like rotation, scaling, and flipping can increase
the diversity of the dataset, improving model generalization. To assess the model's
performance, metrics like accuracy, precision, recall, F1-score, and ROC-AUC are employed.
It is crucial to ensure compliance with healthcare regulations, including HIPAA, when
managing patient-related data. Continuous monitoring and maintenance of the model are
necessary to adapt to evolving conditions and data , Collaboration with healthcare
professionals is crucial to ensure that the model aligns with clinical practices and patient
care.Ensemble techniques, such as combining multiple models, can improve overall
performance and reliability, Explainable AI (XAI) techniques can provide insights into model
decisions, increasing trust among healthcare providers , healthcare should follow strict
protocols and ensure the security of patient data. Regular updates and retraining of the model
with new data are essential to maintain its effectiveness in breast cancer detection.
Design guidelines for mammogram-based computer-aided systems using deep learning
techniques.
This paper high-quality, well-annotated mammogram datasets are essential for training
accurate models.Ensure diversity in the dataset to account for variations in age, breast
density, and imaging equipment. Collaborate closely with radiologists to define clinical
requirements and validate model outputs , Implement strategies to reduce false positives, as
they can lead to unnecessary patient stress , maximize sensitivity to detect potential
abnormalities, even at early stages.Ensure seamless integration with existing radiology
workflows and picture archiving and communication systems (PACS), real-time or near-real-
time analysis to streamline the diagnostic process,Optimize the model for deployment on
specific hardware, such as GPUs or specialized accelerators. Adhere to regulatory
requirements, which may include FDA clearance or CE marking for medical devices ,handle
a growing volume of mammograms as healthcare facilities expand.
Breast mass classification using eLFA algorithm based on CRNN deep learning model.
Is using AI method design a CRNN shown promise in automating this classification , the
eLFA (enhanced Learning from Anatomical priors) algorithm enhances the performance of
the model by leveraging anatomical information. Preprocessing steps include resizing images,
normalizing pixel values, and augmenting the dataset with transformations like rotation and
flipping.Feature extraction is a critical step performed by convolutional layers in the CRNN
model, which identify relevant patterns in the images , recurrent layers (e.g., LSTM or GRU)
capture sequential dependencies in the data, which is important for analyzing medical images.
The eLFA algorithm incorporates anatomical priors to improve the model' s understanding of
breast anatomy and pathology , CRNN model involves optimizing a loss function using
techniques like stochastic gradient descent (SGD). The assessment of the model's
effectiveness involves the use of evaluation metrics like accuracy, precision, recall, F1-score,
and ROC AUC. It is imperative to adhere to healthcare regulations, including HIPAA, when
dealing with patient-related data. Hyperparameter tuning involves optimizing learning rates,
batch sizes, and model architecture to achieve the best results.Model explainability
techniques can help clinicians understand the model' s decisions and build trust in AI-assisted
diagnosis. Ethical considerations, patient privacy, and regulatory compliance are paramount
when working with medical data.
Detection of breast cancer from whole slide histopathological images using deep multiple
instance CNN.
In Multiple Instance Learning (MIL), an image is segmented into patches, with each patch
considered as an individual instance. MIL is employed to address the bag-level labeling of
whole slide images, particularly in cancer detection. Deep learning techniques, notably
convolutional neural networks (CNNs), have demonstrated significant potential in automating
cancer detection.MIL models are trained to classify whole slide images as either positive
(cancerous) or negative based on the presence of cancerous patches. To ensure image
consistency and quality, data preprocessing is a critical step. Deep multiple instance CNNs
are purpose-built to effectively handle MIL tasks. These networks incorporate convolutional
layers for feature extraction and MIL layers for bag-level classification.The use of transfer
learning, by leveraging pre-trained CNN models, can expedite training and enhance
performance. Model performance is evaluated using key metrics such as sensitivity,
specificity, accuracy, and AUC. Interpretability tools, like heatmaps, facilitate pathologists in
comprehending model predictions. Automating cancer detection has the potential to improve
diagnostic accuracy and alleviate the workload of healthcare professionals. Challenges in this
endeavor include addressing variations in tissue appearance, stain variability, and the
computational resources required. The availability of large annotated datasets is vital for
training robust models. Continuous research and development in the realm of deep learning
for histopathological image analysis are advancing the field and enhancing early cancer
detection.
DETECTING BREAST CANCER USING ARTIFICAL INTELLIGENCE
Cancer can be consider most deadly disease and yet not standard cure has been evolved. On
the report of, National Breast Cancer Foundation declare that ,in 2020 alone ,more than
27000 of invasive cases and 48000 non-invasive which has to be diagnosed in united
states .Usually ,Cancer can be detected using subunits of artificial intelligence like machine
learning and deep learning .Machine learning involves in detection and treatment of several
wild disease in intial stage of disease which preventing people lives from threatening disease.
In contrast with deep learning detecting highly critical diseases. Breast cancer can be detected
by two ways ,gene and Histopathological imaging .where, detecting level in gene is quite
expensive so that this methods is used in lab .rather than histopathological is most common
approach in detecting breast cancer .according to American Cancer Society ,declaring that
more than 41760 women and 500 men were died due to breast cancer in recent days. Breast
cancer has four types,1.normal(common level of cancer) 2.benign(It occur in mirror portion
of breast structure and it doesn’t harmful other parts),3.insitu carcinoma(It may occur in
mammary duct lobule and low level harmness and by diagnosed it will be curable),4.invasive
carcinoma(The most critical stage of cancer and quite difficult to solute). Histopathological
imaging at tissue level and analyse underlying molecular tissue and genetic biomarkers wher
geneotic process involves expensive in datset and high computational power require as the
result geneotic is quite expensive so that its limited ilab use thus technique.Several method
including in detect breast cancer ,X-ray mammography ultrasound (US),computed
Tomographhy(CT),Portion Tomography(PET),Magnetic Resonance Imaging (MRI) and
breast temperation measurement.Deep CNN’s was used to detect and diagnosis ,predict non –
communicable and NCD was noisy data and irrevalent to rebust against the NCD some of the
technique where used like KNN,SVM,NN.Natural inspired computing(NIC) algorithms has
been used in detecting diabetes and cancer .It’s consider has the high level performance in
detecting most common classify of cancers(breasr,lung,prostate,ovarian).To be specific in
detecting the breast cancerans effective methodology od detecting diabetes and leukemia
(cancer found in blood forming tissuse and bone marrow).Concluded that NIC with other
algotherims produce mpre precise and clear outcomes.NN’s used tp classify the initial stage
of cancer disease and detect tumour cells.Finally,several research papers of Artifical
intelligence and data mimimg algorithms are imcluded few papers used in genetics and
algorithms in predicting brest cancer.
Algotherims:NIC,ANN’s,and PNN’s Accuracy:98.7
Deep Learning Based Methods for Breast Cancer Diagnosis: A Systematic Review and
Future Direction.
In the realm of breast cancer, the most aggressive form is invasive carcinoma, known for its
capability to metastasize to various organs. Its detection primarily relies on multiple methods,
including mammography, X-rays, ultrasound (US), Positron Emission Tomography (PET),
Computer Tomography, temperature measurement, and Magnetic Resonance Imaging (MRI).
The researchers have explored various deep learning (DL) and traditional machine learning
(ML) techniques. Deep learning methods, particularly Convolutional Neural Networks
(CNNs) and Recurrent Neural Networks (RNNs), have displayed significant potential in
enhancing the accuracy of breast cancer diagnosis through image analysis.. Deep learning
models excel in extracting intricate features from complex medical images. Large and diverse
datasets are essential for training effective deep learning models. Transfer learning allows
fine-tuning pretrained models for breast cancer diagnostics. Deep learning can automate and
expedite the diagnostic process, benefiting both patients and healthcare providers. Ethical
concerns include bias mitigation, model interpretability, and patient data privacy.
Collaboration between machine learning experts and medical professionals is crucial for
practical implementation . Regulatory approval hurdles need to be navigated before
deploying deep learning in medical diagnosis .Interdisciplinary efforts can bridge the gap
between technical advancements and clinical needs. Model interpretability is a challenge for
ensuring trust and decision transparency.Combining multiple data modalities can enhance the
accuracy of breast cancer diagnosis. Lack of standardized datasets poses challenges in
benchmarking deep learning models. Addressing issues of dataset imbalance is vital for
unbiased model performance. Deep learning methods can uncover subtle patterns in medical
images for early detection. Clinical validation is crucial to ensure deep learning models align
with real-world medical scenarios. Deep learning aids in classifying benign vs. malignant
breast tumours from histopathological images. Integration of deep learning tools into clinical
workflows requires careful consideration. Multi-institutional collaboration can lead to more
robust and generalizable models. Research directions include improving explain ability ,
handling noisy data, and regulatory compliance.
Breast cancer detection using deep learning
Deep learning harnesses extensive medical image datasets, such as mammograms, to enhance
the precision of breast cancer detection. The initial steps involve cleaning and preprocessing
the images, encompassing tasks like resizing, normalization, and augmentation, to bolster the
model's generalization capacity. Deep neural networks have the capability to autonomously
acquire intricate features. Utilizing pre-trained models trained on extensive datasets can
expedite the process of fine-tuning for specific breast cancer detection tasks, conserving both
time and resources. Artificial intelligence can function as a valuable tool to augment the
diagnostic capabilities of radiologists. To establish an effective deep neural network
architecture, typically utilizing convolutional neural networks (CNNs), is favored for image
analysis, given their proficiency in capturing spatial features. Training the model entails using
a pre-processed dataset with labeled examples, involving both forward and backward passes,
as the model iteratively adjusts its parameters to minimize the disparity between predicted
and actual results. Hyperparameters like the learning rate, batch size, and network depth
should be fine-tuned to optimize the model's performance. To evaluate the model's efficacy,
metrics such as accuracy, precision, recall, and F1-score are employed. Post-processing
techniques, like thresholding or filtering, are subsequently applied to refine the model's
predictions.
DiagnosisofBreastCancerusingMachineLearningTechniques-ASurvey.
Machine learning has risen to prominence as a potent instrument in enhancing the precision
and effectiveness of breast cancer diagnosis. Support Vector Machines (SVM) are frequently
employed for classification purposes, while deep learning approaches, such as Convolutional
Neural Networks (CNNs), exhibit potential in the realm of image-based breast cancer
diagnosis. Ensemble methods are adopted to amalgamate multiple models, amplifying
predictive accuracy. In breast cancer research, the Wisconsin Breast Cancer Dataset (WBCD)
and DDSM hold a well-established reputation as widely-used datasets. The availability of
high-quality, annotated datasets is of utmost importance in the training and evaluation of
machine learning models designed for breast cancer diagnosis. Dimensionality reduction
techniques like PCA reduce data complexity while preserving important information.Cross-
validation is crucial to ensure that models generalize well to unseen data.Imbalanced datasets
can lead to biased models, especially when the occurrence of cancer is rare, Interpreting
machine learning models in a clinical context is challenging but essential for trust and
adoption.Ethical considerations include patient privacy and ensuring that models do not
reinforce biases. State of the art models often leverage deep learning and large-scale data,
involve developing decision support systems for healthcare professionals.Future research
areas include improving model interpretability and addressing real-time diagnosis challenges.
BreastCancerDetectionandClassificationusingDeepLearningXceptionAlgorithm.
The data, including resizing, normalization, and augmentation techniques to enhance model
robustness.Xception is a deep convolutional neural network (CNN) model known for its
ability to capture intricate patterns in images. It' s based on the Inception architecture with
depth-wise separable convolutions.Utilize pre-trained Xception models on a large image
dataset like ImageNet,hyperparameters such as learning rate, batch size, and regularization
techniques to optimize model performance. Evaluate the model's effectiveness by measuring
key metrics such as accuracy, precision, recall, F1-score, and ROCAUC on the testing
dataset. The interpretability of the Xception model' s decisions, methods like Grad-CAM or
saliency maps are explained.Patient data privacy and compliance with regulations such as
HIPAA,Continuously monitor the model' s performance in real-world clinical settings its
reliability. It might discuss the specific layers or components of the Xception model that are
particularly relevant for this task, providing insights into why the architecture is effective.
Breast cancer detection using artificial intelligence techniques: A systematic literature
review.
AI techniques employed in breast cancer detection e.g deep learning , machine learning ,
image analysis.The systematic approach used to identify relevant studies, including search
terms and databases, used to include studies, such as publication date range, study design, and
types of AI techniques considered.Atasets, including size, modality (mammography,
ultrasound, MRI), and availability, AI model architectures used, including CNNs, SVMs,
decision trees, etc… Indicate whether transfer learning techniques were applied, and if so,
which pre-trained models were used.This could include medical databases, publicly available
datasets, and clinical records, the types of AI model architectures used in the studies. This
could include details about convolutional neural networks (CNNs), support vector machines
(SVMs), and other models.Pre-trained models served as a starting point and how they were
adapted for breast cancer detection.The metrics employed to evaluate the AI models '
performance are outlined. These metrics might encompass accuracy, sensitivity, specificity,
ROC-AUC, and more.
MICROWAVE IMAGING FOR EARLY BREAST CANC ER DETECTION
Breast cancer is the most common leading disease which cause women death rate high in
every year worldwide. Breast cancer can be detected in many method like Artificial
intelligence subunits (Deep learning and machine learning). Recently, Microwave Imaging
(MWI)is boomed to detect the breast area which gives promising prediction over the invasive
cancer. When compared to other techniques microwave imaging is quite inexpensive, easy on
hands, and safe detecting tool. This paper carry out on principle, operation , future
development and present invention of Microwave Imaging. Initially, paper talks about
overview of MWI technique used in detecting the breast cancer and their types. Later, the
paper ended with recent studies on MWI technique and detailed challenge faced by MWI
while detecting the breast tissue. On result of American cancer society(ACS) overall 1.9
million new cases were found and 600000 death occur in United states. There’s various
technique to detect the cell mutation and simultaneously detect(substitution, duplication,
insertion, deletion). PCR, Sanger dideoxy sequencing, pyrosequencing, multiplex ligation-
dependent probe amplification (MLPA), mass spectrometry (MS), and fluorescence in situ
hybridization (FISH) are all techniques used to investigate genetic mutations within a single
gene. Various methods are available for the detection and screening of breast cancer,
including clinical breast examination, X-ray mammography, ultrasonography, magnetic
resonance imaging (MRI), and positron emission tomography (PET). Mammography plays a
pivotal role in early cancer prediction, followed closely by Sono mammography.
Which is mostly used to suspect the breast tissue. Mostly, MWI is best technique for post-
chemotherapy imaging and have sensible to silicon breast implants. United states uses the
MRI more because of having better experience and less working. There are two types in
MWI, Microwave tomography(MWT), and radar -based technique. Dielectric proprieties of
breast tissue are used in MWI to differentiate the healthy and affected tissue of breast. MWI
have classified into two groups: passive and active . Passive,(natural electromagnetic
radiation released by healthy tissue) active, (electromagnetic reflected signal are measured).
Microwave tomography can be best treatment in 2D dimensional or object of interest in
dielectric properties. Advanced in computational methods and microwave hardware design
reason behind in using MWI by researchers. Microwave imagining is included frequency in
different applications where resolution directly proportional to frequency wavelength and
enhance the operation .Overall medical field uses different technique in detecting and
screening breast cancer like machine learning ,deep learning, thermography and microwave
imaging which optimise the predicting outcomes these technique used to prevent the death
case in worldwide survival.
MACHINE LEARNING ALGORITHM FOR BREAST CANCER
Breast cancer is a deadly disease, claiming a high number of lives each year and ranking as
the most prevalent cancer in women worldwide. Methods like classification and data mining
are highly effective for breast cancer detection and data analysis, particularly in the medical
field, where they play a crucial role in disease diagnosis and decision-making. In comparison
to other techniques, machine learning methods such as Support Vector Machine (SVM),
decision trees, Naive Bayes, and K-Nearest Neighbors (K-NN) have shown their potential.
The primary objective is to classify data and accurately analyze the stage of breast tissue,
achieving a high level of accuracy. Breast cancer is the second leading cause of death after
lung disease, making it a significant health concern. Informative and Communicative
Technologies (ICT) have played a vital role in cancer research, not only in handling large
datasets but also in extracting valuable insights. These advancements have improved outcome
predictions and enhanced patient survival rates and the quality of healthcare. Classification
stands out as a crucial task in breast cancer detection, and the utilization of data mining and
machine learning algorithms simplifies the predictive process. In their study, Vikas Chaurasia
and Saurabh Pal compared various classifiers, including Naive Bayes, SVM with an RBF
kernel, RBF neural networks, decision trees, and the simple CART method. Their research
determined the best results for detecting breast cancer. T. Joachims also explored the use of
neuron-fuzzy techniques in the analysis and detection of critical decisions using the
Wisconsin Breast Cancer original dataset.
Accuracy in naïve bayes and k-NN 95.12% and 95.28%
SVM is 97%
THERMOGRAPHY BASED BREAST CANCER DETECTION
Breast cancer is the leading cause of death among women and is commonly detected through
mammography screening and diagnostic modalities. This paper aims to explore the role of
Thermography in the treatment of breast cancer. Thermography is a noninvasive imaging
technique that captures thermal patterns using an infrared camera (IR). It is considered a
promising tool for the feasibility of thermal imaging in breast cancer detection. Various
methods exist for sensing breast cancer, including optical (light), ultrasound (sound),
thermography (heat), magnetic, microwave, X-rays, nuclear, electrical impedance, and
computer-aided diagnosis. Mammography, for instance, is employed to identify abnormal
breast masses, involving the compression and flattening of breast tissue to enhance image
quality and prevent image blurriness. Where tissue density also affect results of
mammography imaging. Magnetic Resonance Imaging (MRI) used to detect early stage of
breast cancer focusing on pre-menopausal women. They detect and measure the healthy
tissue and cancerous breast and metabolic activity ,vascular circulation in pre cancerous
tissue and surrounding area which has high chance to affect that can be easily detected by
infrared imaging. Breast cancer detection methods include mammograms, ultrasound, and
MRI. In a study conducted by Gautheric et al., patients initially classified as healthy but
showing abnormal thermograms were tracked over a 12-year period, revealing that 44% of
these patients developed cancer. The study's conclusion indicated a high cancer risk
associated with abnormal thermograms. Several case studies have also supported this finding,
underscoring the significance of breast temperature symmetry for thermogram accuracy. n the
present day, the use of infrared (IR) techniques has become widespread for the detection of
various diseases, including the analysis of malignant tumors. This approach can be
categorized into four main methods: statistical, model-based, geometrical, and signal
processing methods. Notably, automatic diagnosis systems have been developed to detect
breast cancer using thermograms, texture analysis, and a Support Vector Machine (SVM)
classifier, offering a promising tool for physicians in breast cancer detection with an accuracy
rate of 88.10%.
Sensitivity and specificity 85.71% and 90.48%.
BREAST CANCER DETECTION USING DEEP LEARNING TECHNIQUE
Breast cancer is the major death cause for women lives. Sorting at early stage by detecting
and diagnosis effective to prevent the lives against cancer. The most recommended imaging
method is mammography which is frequently used to detect breast cancer. Convolution
neural network(CNN) is one of the classification of deep learning and Mobile net and
Inception V3 is consider to be the model of CNN which helps in classification of
mammogram and to assist the radiologist diagnosis into normal and abnormal. Further
discuss about these two models of CNN. Breast cancer treated has the most common form of
cancer occurring in women which cause 246000 death cases in 2016 and the crucial level of
29% among other kinds of cancers. Breast Cancer consider has the second highest count of
14% than other cancer classification.Early detection is much needed to prevent the women
lives. Detecting the breast masses and correctly diagnosis (i.e normal and abnormal) and
computer-aided detection and diagnosis (CAD) are which is a second chance to physician aid.
Deep learning the currently used for detection and analysis the breats area which is subunits
of Artifical Intelligence and Machine Learning . To make the process easier Deep learning
with CNN used in image classification and processing . The convolution process which
reduces the millions of pixel into set of small accurate map. Here, Mobile net and Inception
V3 to classify mammogram which is designated by google researchers and it’s streamlined
architecture to build up the deep wise neural network. The reason to use Mobile Net insteadof
CNN’s traditional method was to detect 3*3 convolution layer followed by bath norm and
ReLU. The inception conv was created by szegedy et al which act has “multi level feature
extractor “by computing different convolutions. Concluded that,Convolutional neural
network are model architecture in Artifical Intelligence that solve the image classification
problems .the process involve in classification are, convolution(Mathematical classification
which used in filtering signal and find pattern),pooling layer (which is used after convolution
process to detect spatial size like width and height it doesn’t include depth),Flattering(),Fully
connected layer( matrix multiplication of output neurons which receive input neuron).Finally
this paper concluded that computer-aided diagnosis(CAD) used in mammograms tht detect
normal and abnormal levels at early stage and advice the patient to treat then with help of
inception v3 to detect Breast cancer.
Accuracy :83% for inception v3 58% mobile net.
 Chart Title

MRI, IR

Navive bayes,KNN

MWI

CNN

MIL, CNN

CRNN

PACS

CNN

DT,CNN

CNN

SVM,CNN

CNN

MRI,CNN

NN,CNN

CNN using convolution layers

MRI,Max-Pooling

CNN

CNN

RBFN

BI-RADS

Data augmentation, CNN

NN Algorithm

Hybrid DNN

0 20 40 60 80 100 120

SPECIFICITY SENSITIVITY ACCURACY

SNO TITLE ALGORITHM ACCURAC SENSITIVITY SPECIFICITY
Y
1. Multi-input deep
learning approach for
breast cancer screening
using thermal infrared Hybrid DNN 97% 83% 83.33%
imaging and clinical
data.
2. Breast cancer 86% 87% 84%
classification from
histopathological images
using patch-based deep
learning modelling. NN Algorithm

3. A Novel deep learning Data 98.96% 97.83% 99.13%

model for automatic augmentation,
detection and CNN
classification of breast
cancer using the transfer
learning technique
4. A Novel deep learning BI-RADS 86.67% 88.23% 84.45%
model for automatic
detection and
classification of breast
cancer using the transfer
learning technique
5. A Systematic Review of RBFN 90.48% 89.73% 87%
Breast cancer detection
using thermography and
neural networks.
6. Convolution Neural CNN 93.3% 88.23% 96.20%
Networks for Breast
cancer diagnosis using
diffusion Optical
Tomography.
7. A novel deep learning CNN 97.52% 92.80% 95.33%
based framework for the
detection and
classification of breast
cancer using transfer
learning.
8. .Deep learning MRI,Max- 86.98% 85.45% 87.45%
application to breast Pooling
cancer detection by
magnetic resonance
imaging.
9. Deep learning to CNN using 83.67% 88.79% 87%
improve breast cancer convolution
detection on screening layers.
mammography
10. Deep learning model for NN,CNN 99.33% 100% 98.67%
fully automated breast
cancer detection system
from thermograms.
11. Deep Learning Based 89% 85% 81%
Methods for Breast MRI,CNN
Cancer Diagnosis: A
CONCULATION

Conclusion:
This paper has provided a systematic review of research works dealing with AI models in
conjunction with thermography for early breast cancer detection. The integration of
thermographic imaging with neural network-based analysis holds potential for improving
early breast cancer detection. However, further research and refinement of techniques are
needed to address limitations and enhance accuracy. Furthermore, the study indicated that
ANN and deep learning models could achieve an acceptable level of detection accuracy but
they can be improved. CNN, Data augmentation ,hybrid DNN, NN, DT, PACS, CRNN, MIL,
Navive bayes, KNN, this all are highly values in accuracy it can used mostly 93%, but ANN
in backpropagation network(BPN) only we are analysing to decision tree(DT) and fuzzy
surgeon is take high accuracy to analyse the classification in ANN,SVM the accuracy in
96.48%. That’s why we are taking ANN method to verified.
REFERENCES:

1) Hasan Nasir Khan , Ahmad Raza Shahid ,Basit Raza , and Hani Alquhayz , ”Multi-
View Feature Fusion Based Four Views Model for Mammogram Classification Using
Convoluation Neural Network”,2019.
2) Maged Nasser and Umi Kalsom Yusof, ”Deep learning Based Method for Breast
Cancer Diagnosis: A Systematic Review and Future Direction” ,2023
3) Rahual Kumar Yadav, Pardeep Singh*, and Poonam Kashtriya, ”Diagnosis of Breast
Cancer using Machine Learning Techniques-A Survey”,2023.
4) Bhavi Gami ,Khushi Chauhan, and Brijesh Kumar Y.Panchal,”Breast Cancer Detection
Using Deep Learning”,2023.
5) Ali Bou Nassif*,Manar Abu ,Qassim Nasir ,Yaman Afadar ,Omar Elgendy, ”Breast
Cancer detection using Artificial Intelligence techniques: A systematic literature
review” ,2022
6) Basem S. Abunasser ,Mohammed Rasheed J.AL-Hiealy ,Ihab S. Zaqout ,Samy S. Abu-
Naser, ”Breast Cancer Detection and Classification using Deep Learning Xception
Algorithm” ,2022.
7) Zhiqiong wang, Mo li ,Huaxia Wang, Hanyu Jiang ,Yudong Yao, Hao Zhang, and
Junchang Xin, “Breast Cancer Detection using Extreme Learning Machine Based on
Feature Fusion with CNN Deep Features” ,2019.
8) Farnoosh Azour and Azzedine Boukerche, “Design Guidelines for Mammogram-
Based Computer-Aided System Using Deep Learning Techniques” ,2022.
9) Chang-Min Kim ,Roy C.Park ,and Ellen J.Hong, “Breast Mass Classification Using Elfa
Algorithm Based on CRNN Deep Learning Model” ,2020.
10) Moi Hoon Yap, Gerard Pons, Joan Marti ,Sergi Ganau ,Melcior Sentis, Reyer
Zwiggelaar Adrian K. Davison , and Robert Marti, “Automated Breast Ultrasound
Lesion Detection Using Convolutional Neural Networks”, 2018.
11) Kausik Das, Sailesh Conjeti ,Jyotirmoy Chatterjee, and Debdoot Sheet ,”Detection of
Breast Cancer From Whole Slide Histopathological Images Using Deep Multiple
Instance CNN”,2020.
12) Irum Hirra, Mubashir Ahmad ,Ayaz Hussain ,M. Usman Ashraf, Iftikhar Ahmed Saeed
,Syed Furqan Qadri ,Ahmed M. Alghamdi, and Ahmed S. Alfakeeh, “Breast Cancer
Classification From Histopathological Images Using Patch-Based Deep Learning
Modeling”,2021.
13) Abeer Saber, Mohamed Sakr ,Osama M. Abo-Seida ,Arabi Keshik, and Huiling Chen,
“A Novel Deep-Learning Model for Automatic Detection and Classification of Breast
Cancer Using the Transfer-Learning Techniques” ,2021.
14) Nan Wu , Jason Phang , Jungkyu Park, Yiqiu Shen , Zhe Huang, Masha Zorin,
Stanisław Jastrze˛bski, Thibault Févry, Joe Katsnelson, Eric Kim , Stacey Wolfson,
Ujas Parikh, Sushma Gaddam, Leng Leng Young Lin, Kara Ho, Joshua D. Weinstein,
Beatriu Reig, Yiming Gao, Hildegard Toth, Kristine Pysarenko, Alana Lewin, Jiyon
Lee, Krystal Airola, Eralda Mema, Stephanie Chung, Esther Hwang, Naziya Samreen,
S. Gene Kim, Laura Heacock , Linda Moy, Kyunghyun Cho, and Krzysztof J. Geras,
“Deep Neural Networks Improve Radiologists Performance in Breast Cancer
Screening” ,2020.
15) Mohammed Abdulla Salim Al Husini, Mohammed Hadi Habaebi ,Shihab
A.Hameed ,MD.Rafiqul Islam and Teddy Surya Gunawan, “A Systematic Review of
Breast Cancer Detection” ,2020.