Author: Dr. Arpana Chaturvedi (Jagannath International Management School, New Delhi, ac240871@gmail.com)
Artificial Intelligence, Machine Learning and Deep Learning now-a-days started playing its very effective and important role resulting great impact on various domains. These fields have been used in all areas as Data scientists realized that with the strength and power of rapidly growing data. The data shared by people of all ages in almost all social media handlers is of different types and in huge volume. This data consists of various kind of information related to almost all domains. Data analyst knows the power of this data and they introduced various techniques to get fruitful hidden insights from the data to benefit various organizations.
Role of machine learning in detection, prevention and treatment of cancer
1. Role of Machine Learning in Detection, Prevention and Treatment of
Cancer
Author: Dr. Arpana Chaturvedi (Jagannath International Management School, New Delhi,
ac240871@gmail.com)
9.1 Introduction: Artificial Intelligence and Machine Learning
Artificial Intelligence, Machine Learning and Deep Learning now-a-days started playing its
very effective and important role resulting great impact on various domains. These fields
have been used in all areas as Data scientists realized that with the strength and power of
rapidly growing data. The data shared by people of all ages in almost all social media
handlers is of different types and in huge volume. This data consists of various kind of
information related to almost all domains. Data analyst know the power of this data and they
introduced various techniques to get fruitful hidden insights from the data to benefit various
organizations.
Algorithms used by Artificial Intelligence (AI), Machine Learning (ML) and Deep Learning
(DL) started impacting the day-to-day life of society and have dominant influence in almost
all industries. Various disciplines in which AI and Machine Learning is playing very
important Role are Education, Medical, Agriculture, Manufacturing, Aeronautical, Big
organizations like Bank, Scientific Applications and many more.
Health Care industry get influenced with these technologies most. Medical Science is one of
the major areas where the ML and DL algorithms are used to diagnose various diseases, so
that patient can be treated well and in advance. Researchers use AI and ML in medical
applications such as EEG analysis and Cancer Detection/Analysis
Advancement in technologies especially in the area of Artificial Intelligence, Machine
Learning and Deep Learning have paved the path to Health Care Industry. There are so many
2. powerful Diagnosis Tools which uses the available data sets to detect Cancer at early stages.
The biological datasets of DNA Methylation and RNA sequencing are used now-a-days to
examine it which helps to know about the genes. It can help doctors and analysts to know
which all genes cause cancer and which all are able to suppress its expressions.
Fig 1. Mapping between AI, ML, DL and NLP
In the above picture it is clear that ML and NLP are the subset of AI, DL is the subset of ML.
Various neural network base algorithms are developed in ML. These Algorithms allows the
machine to learn in different ways and resolve the problems in the same manner as out human
brain do. Deep Learning (DL) is basically works like human brain. It has the ability to
process the data and to identify different images, various objects. DL can also process
languages and helps humans to take appropriate decisions to improve drug discovery, to
upgrade precision medicines and to improve diagnosis of various diseases.
Deep Learning uses Artificial Neural Networks to process various medical Images to
diagnose further. The Process of Artificial Neural Networks impersonates the human neural
architecture. To enhance the processing power of Machine Learning, ANN us used which is
basically consists of various input, generates output, and various hidden multi-layer networks.
3. Fig. 2. Applications of AI, ML and DL in Digital Healthcare
In this chapter, we talk about medical discipline and see the various ways to diagnose various
kind of cancers in patients. Before discussing various algorithms, which can be used to
diagnose different types of cancer, we will briefly talk about Machine Learning and AI along
with various types of algorithms used to perform analysis.
9.2 Role of AI, ML and Deep Learning in Health Industry for Cancer Detection
Mathematical Models of Artificial Intelligence have the ability to understand the human
cognitive abilities. This feature easily handles the healthcare challenges like complex
biological abnormalities exists in cancer. The role of AI Technologies has played a very
important role and this exponential growth of technologies AI assisted us to make effective
and optimal decision. The super-intelligence these technologies uses has overcome the issues
of challenging areas where human mind is limited to process huge data within fraction of
seconds. AI has helped to diagnose Cancer disease by using AI Based Algorithms to identify
ARTIFICIAL
INTELIGENCE
AI
MACHINE
LEARNING
ML
DEEP
LEARNING
DL
DIGITAL HEALTHCARE
Artificial Intelligence
Programming System to Perform task
which normally required Human
Intelligence
Machine Learning
A Subfield of Artificial Intelligence
Deep Learning
A Subfield of Machine Learning
Precision Oncology
Drug Discovery
Imaging and Digital
Pathology
Next Generation
Sequencing
Pattern Data Management
4. complex and multifaced disorder with thousands of genetic and epigenetic variations.
Different AI algorithms help to identify these complex and large number of genetic mutations
and aberrant protein interactions at a very early stage. Modern biomedical research is
working so hard to bring AI technology to the assist pathologists and physicians also. These
technologies enable them to predict any disease risk, diagnosis it minutely and suggests
effective treatments on time. Researchers are digitally collaborating in real time to diagnose
Cancer and provide treatments to heal millions of people. are continuously using Machine
Learning, AI based system approach for Clinical Applications, researchers can collaborate in
real-time and share knowledge digitally to potentially heal millions. Researchers contribute
their knowledge in this area and help oncologist so that they can give precise treatment to
every individual in future.
Classification accuracy of deep learning classifier is largely dependent on the quality and size
of the datasets. Deep learning requires massive amount of medical imaging training dataset to
give precise and successful results. The biggest challenge in the success of deep learning is
unavailability of enormous amount of dataset.
9.2.1. Deep Learning Architectures:
There are several types of Deep learning architectures, also known as artificial neural
networks of multiple nonlinear layers. Characteristics of input data and the objective of the
research work helps one and individual to decide which Deep Learning architecture is to be
used and when. Various types of emerging and popularly used Deep Learning Architectures
with various research objectives are:
Deep Neural Networks (DNNs)
5. Convolutional Neural Networks (CNNs)
Recurrent Neural Networks (RNNs).
Deep Neural Network DNN: -Various Deep Learning Architectures in DNN are designed
on the basis on building blocks of Neural Networks. These building blocks are based on
Multilayer Perceptron (MLP) which uses Perceptron‟s, Stacked Auto-Encoder (SAE) which
uses Auto-Encoders and Deep Belief Networks (DBNs) which use Restricted Boltzmann
machines (RBMs).
Convolution Neural Network CNN: CNNs architectures are consists of different layers like
convolution layers, nonlinear layers, and pooling layers. The CNN‟s are mostly successful in
the field of image recognition.
Recurrent Neural Network RNN: RNNs architectures are designed on the basis on building
blocks like perceptron‟s, Long Short-Term Memory Units (LSTMs), or Gated Recurrent
Units (GRUs). These architectures use the sequential information of input data with different
cyclic connections.
There are many other emerging and research objective based Deep Learning Algorithms in
use. Few are Deep Spatial-Temporal Neural Networks (DST-NNs), Multi-Dimensional
Recurrent Neural Networks (MD-RNNs), and Convolutional Auto-Encoders (CAEs).
9.3. Introduction to Cancer Disease:
Cell Biology: The fundamental unit of Life is basically the Smallest Structure of Body. This
smallest structure is known as Cell. These Cells are capable of performing the process in
Life. In the life cycle of Cell, Reproduction of Cells is known as Cell Division. In a
Multicellular Organism of any body, each and every cell has same DNA (Deoxyribonucleic
6. Acid). Each cell has a continuous growth process in a Human body. During the process of
growth human Cells grow and multiply rapidly to form new cells. With the time and age, the
Cell grows Old and become damaged or they die. These cells get replaced by the new cells.
Cancer is a disease takes place because of rapid growth, spread and division of uncontrolled
and abnormal cells. Mutation is a process in which the changes in the cell happened due to
mutation can viewed by the immune system. In this disease division of cell becomes
uncontrolled because of fast growth of abnormal cells. This rapid growth of large numbers of
abnormal white cells in the blood cells creates a formation of a lump known as tumor. These
cells have the ability to spread rapidly as they can move within the body and also survive in
another part of body easily.
9.3.1. Cancerous Cells:
These cells in our body exists with uncontrollable growth and further, fast reproduction of
Body Cells. Cancer Disease can start anywhere in the Human Body, wherever such cancerous
cells exist. The abnormality in the cell‟s internal regulatory mechanism gives birth to cancer
in that specific area.
Hence keeping Cancer disease in our mind, we must know the difference between Tumor and
Cancer.
Tumors are Lump or Mass of Cells formed in Body due to Abnormal Cellular Growth. There
are two types of Tumors. These are:
Benign Tumors
Malignant Tumors
Benign Tumors are actually Noncancerous Cells. These cells do not spread to other parts of
the Body. It can be easily removed by Surgery. It becomes dangerous when it presses vital
organs or block channels.
Malignant Tumor: Malignant Tumors are made of Cancerous Cells. It invades nearby
7. tissues. Metastasis – Moves into Bloodstream or Lymph nodes. This Malignant Tumors
spreads through other tissues within the Body.
As per the WHO report, Cancer is a leading cause of death worldwide and the most common
causes of cancer death are consisting of Lung Cancer (approx. 1.69 million deaths), Liver
Cancer (788 000 deaths), Colorectal Cancer (774 000 deaths), Stomach Cancer (754 000
deaths), Breast Cancer (571 000 deaths).
9.3.2. Types of Cancer
• Lung cancer
• Colon and rectal cancer
• Endometrial cancer
• Pancreatic cancer
• Kidney cancer
• Thyroid cancer
• Leukemia
• Laryngeal Cancer
9.3.3 Role of Diet in Cancer
A. In Type of Cancer
Type of Cancer Due to Intake of Food Items
Breast Cancer
Pre-Menopause
Alcoholics drinks, body fatness, sedentary
living
Colon Cancer
Rectum Cancer
Kidney Cancer
Red meat, processed meat, barbecuing meat,
high intake of fat, body fatness, abdominal
fatness, sedentary living
Lung Cancer Drinking Water Containing Arsenic.
8. Stomach Cancer
Liver Cancer
Mouth Cancer
Pahrynx Cancer
Larynx Cancer
Oesophagus Cancer
Pancreas Cancer
High intake of alcohol, body fatness
Prostate Cancer Diet in high calcium
Cervical Cancer Folate deficiency
Table 1. Role of Diet in Cancer
B. Cause due to Lack of Food Items
Lack of Type of Food Items, Minerals and
other components
Type of Cancer
Minerals-Selenium or Zinc Selenium and zinc deficiency may also
increase risk of cancer
Vitamin A or Carotenoids Low level of carotenoids in blood causes lung
cancer
Vitamin C Low dietary vitamin C cause oropharyngeal,
stomach and esophageal cancer
Vitamin E Low vitamin E cause lung, cervix and
colorectal cancer
Nitrates Cause nasopharyngeal, stomach and
colorectal cancer
Aflatoxins Cause liver cancer
Energy dense foods, sugar drink fast foods Cancer in any part of the body
Table 2. Types of Cancer due to Lack of Food Items
9. 9.4. Need of AI, ML and DL Algorithms to Diagnose Cancer:
The development of large medical imaging data is quite challenging. Medical experts need
extensive time to study in depth and need to give so much of explanation. Multiple experts‟
opinion is considered to reach to some conclusion after analyzing volume of medical images.
The high rate of cancer underlines the importance of research in Screening, diagnosis and
monitoring of WBC blood smear images. The 5-years survival rate is decreasing. The
morphology and size of CLL cells are close to the normal lymphocyte. This introduces
difficulties in identifying this disease at early stages. This signifies the importance of research
in the Segmentation and further classification of cancer affected Blood smear images. Cancer
in childhood is the most commonly used to designate cancer that affects children prior to the
age of 15, representing between 0.54% and 4.86% of all cancers. Overall prevalence rates
range from 50 and 200 per million children across the world. In India, cancer has become the
ninth prevailing element for the mortality among children in the age group of 5 to 14 years.
The percentage of such cancer related to all cancers disclosed by Indian cancer research
varies from 0.8% to 5.8% in boys, and from 0.5% to 3.4% in girls. Indian Council of Medical
Research (ICMR) began a National Cancer Registry Program NCRP in 1981. It is a network
of cancer registries listed by National Cancer Institute. NCRP‟s exhaustive study revealed a
general increase of Cancer cases in childhood with greater percentage among boys of the
Southern region. The most common causes of cancer death are because of Lung Cancer (1.69
million deaths), Liver Cancer (788 000 deaths), Colorectal Cancer (774 000 deaths), Stomach
Cancer (754 000 deaths), Breast Cancer (571 000 deaths). According to WHO - Dietary
factors have been thought to account for about 30% of cancers in Western countries. They eat
animal products, fat and sugar and have high rate of colorectum, breast and prostate Cancer.
The contribution of diet to cancer risk in developing countries has been considered to be
lower, perhaps around 20%. Developing countries typically have diets based on one or two
10. starchy staple foods, low intakes of animal products, fat and sugar. Hence the developing
countries have low rates of „Western‟ cancers, and sometimes high rates of other types of
cancer such as oesophagus, stomach and liver cancer.
Ways of Cancer Treatment: There are several ways to treat Cancer disease. These are:
• Surgery
• Radiation Therapy
• Chemotherapy
9.5 Machine Learning and Role of Model in Machine Learning
Machine Learning concept is used in developing different modelling technique that involves
Data means information such as documents, images, etc. Modelling technique analyzes the
data and finds the suitable model by itself rather than having a human interference in it. Data
used by Machine Learning modeling process are of two types, one is called “Training” data
and the second one is “Test” Data. The primary idea of Machine Learning is to achieve a
model using the training data.
Types of Machine Learning Techniques:
There are three types of Machine Learning Modelling Techniques. These are:
Supervised Learning: In this Modelling Technique each training dataset should consist of
input and correct output pairs.
Unsupervised learning: This modelling technique is used for investigating the
characteristics of the data and for preprocessing the data
Reinforcement learning: This technique employs sets of input, some output, and grade as
training data.
11. Fig. 3. Types of Machine Learning
9.5.1. Model in Machine Learning:
In Machine Learning, different techniques used to develop a Model which enables the
researcher to achieve the goal for the given problem as the final product. There are so many
modelling techniques which are used develop a model to solve different Problems. In this
developing process Machine Intelligence is involved. Some of the techniques are image
recognition, speech recognition, and natural language processing. Machine learning
Modelling techniques are used develop a model to solve those problems for which analytical
models are hardly available.
Fig. 4. Model in Machine Learning
12. Process Flow Diagram of Machine Learning: There are two Flows in which the process can
be represented
1. Vertical Flow: It Indicates Learning Process
2. Horizontal Flow: It Indicates the trained Model
Fig 5. Process Flow of Model in Machine Learning
Generalization
A model is said to be a good machine learning model if it generalizes any new input data
from the problem domain. This helps us to make predictions in the future data on any
given input data that model has never used or seen before.
The process which is used to make the model performance consistent regardless of the
training data or the input data is called generalization. The main achievement or success
of Machine Learning model depends upon its Generalization Capability.
13. Fig. 6. Generalization Capability of Machine Learning
To check the machine learning model whether it has capability to learns and generalizes
the given new data as input two methods are used. These performance checking and
analyzing methods are overfitting and underfitting. These methods are majorly
responsible for the poor performances of the machine learning algorithms.
Steps to deploy a Model
Fig. 7. Steps to design and deploy the ML Model
Steps used to design
a
Machine Learning
Model
1. Define Objective 2. Collect Data 3. Prepare Data
4. Select Suitable Algorithm
5. Train Model
3. Perform Prediction 6. Test Model
8. Deploy the Model
14. Bias and Variance also plays an important role to understand well the developed Machine
Learning Models.
Bias is an assumption made by a model to make a function easier to learn whereas the
Variance signifies the occurrence of high Error on the model. For example: If you train
your data on training data and obtain a very low error, whereas when we change the data
and then train the same previous model, we experience very high error.
Underfitting:
Underfitting of a model is a situation which arise when the algorithm is not able to fit or
capture the data accurately. It destroys the accuracy of the statistical model and usually
occurs when the amount of data is very less. It might result to lot of wrong predictions.
This situation can be avoided by using large number of datasets and reducing the features.
Usually, Machine Learning Models with Underfitting is High bias and low variance
Models.
Overfitting:
Overfitting situation occurs in statistical models because of non-parametric and non-linear
methods we use. This situation arises when we use huge amount of data because of which
it starts learning from the inaccurate entries of the data set. Hence lot of confusion arises
the model does not categorize the dataset accurately. Too many details in the form of
noise appears and the model does not categorize the data correctly. The overfitting
situation can be overcome by using linear algorithms in case of linear data or using
parameters in case of decision trees. Usually, Machine Learning Models with Overfitting
is High variance and low bias Models
Overfitting basically affects the level of performance of Machine Learning which helps us
to check whether the training model is overfitted or not. Validation is a process that
reserves a part of the training data and uses it to monitor the performance.
15. Fig. 8. Steps Used in Validation Process during designing a Model in ML
The validation set is not used for the training process. Model is overfitted when the
trained model yields a low level of performance to the reserved data input.
Train-Test-Split in Machine Learning
Evaluation process to check the performance of Machine Learning Algorithms, the
technique used is known as Train-Test-Split Method. This technique is used all the
problems related to classification and regression. It can also be used for any type of
supervised algorithms.
Step 1
Divide the training data set into two groups
Group 1: Data Set for training
Group2: Data Set for validation.
Rule of thumb
Ratio of the training data set to the validation data set = 8:2.
Step 2
Train the model with the training data set.
Step 3
Evaluate the performance of the model using the validation set.
a. If the model yields satisfactory performance, finish the training.
b. If the performance does not produce sufficient results, modify the model and
repeat the process from Step 2.
Steps of Validation Process
16. Fig. 9. Splitting the Data
This process involves the division of datasets into three subsets. These are:
Training Subset or Training Dataset: It is the first subset which is used to fit the model.
The model learns from this data. For example: in Regression Model to reduce the cost it
helps to predict the data.
Test Subset or Test Dataset: This second subset is not used to train the model whereas it
is used to evaluate the fit machine learning model. In fact, here the input element of the
dataset is provided to the model. On the basis of given input, the model makes predictions
which are then compared to the expected values. It tells us that how effectively the model
is going to predict values which makes sense. It is used to evaluate the performance of a
model.
Cross Validation or Development Dataset: It is the third dataset which is basically
referred to holdout cross-validation or development (dev) set. The sample of data used to
provide an unbiased evaluation of a model fit on the training dataset while tuning model
hyperparameters. The evaluation becomes more biased as skill on the validation dataset is
incorporated into the model configuration. If the difference between error on the training
set and error on the dev set is huge, it means the model as high variance and hence, a case
of over-fitting
17. The objective is to estimate the performance of the machine learning model on new data.
We use the developed and tested model in practice which is fit it available data with
known inputs and outputs, to make predictions on new examples in the future where we
do not have the expected output or target values.
Fig.10. Three different Dataset after Splitting Dataset
The train-test procedure is appropriate when there is a sufficiently large dataset available.
The major problem which ML/DL practitioners face is how to divide the data for training
and testing. Though it seems like a simple problem at first, its complexity can be gauged
only by diving deep into it. Poor training and testing sets can lead to unpredictable effects
on the output of the model. It may lead to overfitting or underfitting of the data and our
model may end up giving biased results.
K- fold Cross Validation
It splits the data into k folds, then trains the data on k-1 folds and test on the one-fold that
was left out. lt divides the training data into groups for the training and validation, but
18. keeps changing the datasets.
Fig. 11. K Fold Cross Validation
9.6. Relationships and high-level schematics of different disciplines.
Every discipline in various levels of schematics need to follow an early approach and is to be
programmed on the basis of required knowledge to obtain solution for the given problem. In
every step lot of difficulties are to be faced in dealing with complex real-world problems.
Artificial Intelligence, Machine Learning and Deep learning has given us the new dimension
to look at the problems. They have the learning capability and feature of improving itself with
the experience and data. These disciplines have the capability to predict, classify and to give
possible and accurate solutions to problems related to any area.
Fig.12. Relationships and high-level schematics of different disciplines
19. Machine Learning:
Machine Learning can extract patterns from data but it also has limitations to process the raw
data. This raw data is actually highly dependent on hand-designed features. To advance from
hand-designed to data-driven features, and to give solution to the issue, representation
learning and deep learning has shown great promise.
Representation Learning:
Representation learning can discover effective features as well as their mappings from data
for given tasks. Furthermore, deep learning can learn complex features by combining simpler
features learned from data.
In other words, with artificial neural networks of multiple nonlinear layers, referred to as
deep learning architectures, hierarchical representations of data can be discovered with
increasing levels of abstraction
Deep Learning:
Deep learning has advanced rapidly and played an important role in various fields. Many
applications of deep learning are performing excellent in the field of bioinformatics. It‟s
performance has gain insight from data and has emphasized in both academia and industry.
Various Algorithms used in different domains, areas and applications are briefly summarized
below with the help of a diagram This diagram gives you the complete information about
various types of algorithms and their applications with example to make you understand
better.
20. Fig. 13. Various types of Algorithms, Applications with Example where they used
9.7 Deep Learning Methods Cancer Diagnosis
There are so many Deep Learning methods as explained in above figure also, that can be used
to diagnose Cancer. To perform diagnosis, to develop a machine learning model, it is very
necessary to understand and take appropriate decisions that what can be done. Exactly what
21. kind of intelligence we try to bring into machineries and then we try to give them some tasks
of recognition to recognize a pattern or an object as being a member of a class.
We can use Classification, Recognition and Detection problem to detect or observe a pattern
or any image or in any signal. Then we need to map it with one of the associated members of
the class to which it belongs.
Classifiers in Machine Learning and its Application:
Various Machine learning techniques using which we can give solution to Classification or
Recognition or detection problems are KNN Classifiers, SVM Classifiers and many more.
The various application where these classifiers are used are:
Fig.14. Applications of Classification or Detection or Recognition Problems
In machine learning various kind of classifiers are used for detection purpose. These
classifiers are used in various application for detection purpose, so that the solution can be
provided timely for further appropriate action. Lot of research work has been done using
22. these classifiers as they can detect easily the occurrences of any event in any object. Let‟s
take example of ECG signals. Using detection techniques, we can analyze whether the heart
beat is normal or abnormal. With the help of X Ray report, we can detect whether the person
is healthy or Covid 19 effected. With the help of medical health record also called as
electronic health record, the machine can easily detect whether the patient is a healthy or
diabatic.
Similarly, Machine can detect the input in case of text and can easily convert it into speech
and in case of voice message-speech, it can convert the voice-data into text.
Optimizers used in Deep Learning Network:
Optimization is the process to optimize or train the Deep Learning Neural Networks. During
the process of optimization, the weights and learning rates are updated to reduce the CE Loss.
Different types of optimizers supported by Deep Learning Neural Network are:
Fig. 15. Different types of Optimizes used in Deep Learning Neural Network
Approaches in Deep Learning approaches:
In Deep Learning network, Hybrid method approach can also be considered. In Hybrid
methods features of CNN can be used with SVM Classifiers. In this approach features of
CNN can be taken from Fully-Connected layer and can be given to SVM to train. Various
other Approaches used for Transfer learning are AlexNet, GoogLeNet, VGG16Net,
Gradient Descent
Stochastic Gradient Descent
Momentum
Adagrad
Adadelta
Adam
23. ResNrt50, and InceptionV3.
9.8. Cancer Segmentation:
Cancer segmentation is a process used for detecting cancerous and non-cancerous cells in any
part of our body. To know this, we need to perform Image segmentation. This is a process
which partitions the images into multiple sub-regions or segments. This division helps to
detect, diagnose and analyze easily to detect the presence of cancer cells. It helps to let the
patient know the level of risk and type of disease. Each and every pixel of an Image is
associated with the cell type, whether it is malign or Benign. Each pixel is labelled
accordingly with the respective class.
Popular Deep Learning Models used for Cancer Segmentation:
There are some popularly used Deep Learning Models used for segmentation purpose in
detecting any kind of cancer. Various applications where these segmentation techniques are
used for detection are to know presence of Breast Cancer, Skin Cancer uses Transfer
Learning Approach (Alex-Net), Cervical Cancer, Multiclass Brain Abnormalities, Prostate
Cancer uses 3D-CNN, Diabetic Retinopathy uses Light Weight CNN, Chest X-Ray using
Transfer Learning, Brain Hemorrhagic and Ischemic Strokes uses CNN, BU-Net Architecture
cells. So much of research work is available to detect all the cancers mentioned above using
different approaches and techniques. Some of the techniques used to perform cancer
detection are:
Fig. 16. Deep Learning Neural Network Approaches and techniques used in Cancer detection
Here U-Net architectures in Neural Network is a Fully Convolutional Neural Network used
U-Net V-Net R-CNN RNN LSTM
24. for the purpose of biomedical Image Segmentation. V-Net are Fully Convolutional Networks
used to process Volumetric Biomedical Image segmentation. R-CNN are Region Based
Convolutional Neural Network used for different object detection and accordingly sementic
segmentation in a particular region. RNN are Recurrent Neural Networks used for modelling
sequential data for variable sized inputs. Long Short-Term Memory (LSTM) is similar to
RNN. It is used for learning order dependence in sequential prediction problems.
Conclusion:
For cancer detection, it is necessary to know the cell histology, then only can predict whether
the cell is cancerous or healthy. As we have seen that there are so many numbers of
Supervised, unsupervised and Reinforcement Algorithms which can be used for various
purposes. We have also seen various applications which can be used to Classify, detect or
recognize various diseases. Deep Learning Techniques and its algorithms are mostly used to
detect Cancer or any other disease. The successes of deep learning are built on a foundation
of significant algorithmic details. To implement it successfully, it is suggested to understand
well various types of deep learning architectures or networks.