Artificial Intelligence In Healthcare | Examples Of AI In Healthcare | Edureka



despite the fact that artificial intelligence invokes fear and most of us it is benefiting us in numerous ways AI in healthcare is revolutionizing the medical industry by providing a helping hand hi everyone I'm lekha from arica and I welcome you to this session on AI in healthcare in this session we'll discuss the different ways in which artificial intelligence is impacting the healthcare domain so let me begin by discussing the agenda with you we're going to begin the session by understanding what exactly artificial intelligence is we're going to move on and understand how artificial intelligence is used in healthcare and we're going to look at a couple of use cases where we will understand how AI is implemented in real world examples after that we look at what exactly machine learning is and what deep learning is and we'll end the session by performing a demo in Python will be using deep learning concepts like neural networks in order to solve a real-world problem so let's move on and take a look at our first topic before I get started with what exactly artificial intelligence is make sure that you subscribe to Ed you rake our YouTube channel in order to stay updated about the latest and the most trending technologies so what exactly is artificial intelligence now artificial intelligence is the development of computer systems that are capable for performing tasks that normally require human intelligence tasks such as decision making object detection solving complex problems and so on the four main benefits of artificial intelligence is that it gives us predictions with an increased level of accuracy it helps us in decision making processes it has to solve complex problems and it performs high-level computations that will take days for a human to solve artificial intelligence is something that makes our lives easier by performing high-level computations and solving complex problem this is a simple definition of what exactly artificial intelligence is now let's move on and discuss artificial intelligence in the healthcare domain as the introduction of artificial intelligence in the 1950s it has been impacting various domains including marketing finance the gaming industry and even the musical arts however the largest impact of artificial intelligence is been in the field of health care now are going to the latest report of PwC the artificial intelligence will contribute an additional fifteen point seven trillion to the world economy by 2030 and the greatest impact will be in the field of healthcare so we know that healthcare is getting more importance or is using artificial intelligence in a more advanced manner now what has led to the sudden importance of AI in the healthcare industry what do you think is a reason behind the sudden growth of artificial intelligence in the healthcare industry let me narrow down for you there are two major points of which have made AI so impactful in the field of healthcare the first reason is high availability of medical data now all of us have tons and tons of medical data in the form of medical history whenever we go to any hospitals our history is written down in accounts so basically with the availability of data implementing artificial intelligence becomes much easier the AI is based on technologies such as deep learning and machine learning which require tons and tons of theta so with the availability of data it became easier to implement or it became easier to use artificial intelligence in the healthcare industry another important reason that led to the development of AI in healthcare is the introduction of complex algorithms now what happens in machine learning is machine learning is not capable of handling high dimensional data and particularly the medical data that we have or healthcare data that we have is of very high dimension in character the data is very vast there are thousands and thousands of attributes now for us in order to process and analyze data of this dimension is hard to do with machine learning but as soon as deep learning and neural networks was introduced this became much easier because deep learning and neural networks basically focus on solving complex problems that involve high dimensional data so the development of deep learning and neural networks also played a major role in the impact of AI in healthcare so I hope all of you understood why artificial intelligence is impacting the health care industry in such a huge man now let's move on and look at a couple of use cases of artificial intelligence in healthcare so AI is actually benefiting healthcare organizations by implementing cognitive technology in order to unwind a huge amount of medical records and in order to perform any power diagnosis take for example neons nons is a production service provider that uses artificial intelligence and machine learning in order to present or in order to predict the intent of a particular user by implementing neurons in an organization's system or in an organization's workflow you can develop a personalized user experience that allows a company to make better decisions and better actions that enhances the customers experience and overall it will just benefit the organization so neurons basically helps in storing collecting and reformatting data in order to provide faster and a more consistent access to all the data so that any further analysis or any diagnosis can be performed now let me tell you a few key features of our neurons so first is service acceleration onion suggests the best next step that needs to be taken so that the customers needs are met it basically derives useful insights so that you can improve customer retention it accelerates all your services another feature is called deflection so it minimizes the volume of inbound calls and lowers the expenses by anticipating the customers intent and diverting the customer to other online engagements churn reduction so basically by using machine learning and natural language processing it can predict the behavior of leads that may be close to in validating or canceling their service based on their history or their searches sentiments and so on so it basically understands the customers that are trying to cancel out their service and it takes appropriate actions in order to avoid any cancellations basically studies of patterns and the behavioral trends of each and every customer it also automate tedious tasks now of course it helps in revenue generation right it increases the amount of revenue that a company or an organization can get by predicting the behavior of patterns and therefore taking relevant actions it is also very assistive and it automates all the tedious tasks so basically it's rid of the monotonous task of calling customers and it basically implements automated systems that will send notifications via SMS or email and uses AI base chat BOTS that make things much simpler so that was all about how neons uses artificial intelligence in managing medical data and any healthcare history records next use case is artificial intelligence in medical diagnosis now medical imaging and diagnosis powered by a I should Brittney's more than 40% growth to surpass 2.5 billion u.s. dollars by 2024 now this was something that was found in the global market insights so with the help of neural networks and deep learning models artificial intelligence is actually revolutionising the image diagnosis field in medicine one major application of AI in medical diagnosis is the MRI scans AI has taken over the complex analysis of MRI scans and it has made it a much simpler process so MRI scans are actually the most difficult to analyze because of the amount of information that they contain normal MRI analysis will take about several arts it'll take about four five hours it can take between two to seven hours also and any researcher that is trying to formulate an outcome or is trying to get the results from these large datasets they have to wait four hours for a computer to generate these scans now the solution here is clearly deep learning so like I mentioned earlier large and complex datasets can be analyzed with the help of neural networks and this is exactly what a team of researchers implemented in MIT right they developed a neural network called voxel morph that was trained on a dataset of approximately 7,000 MRI scans so how a noodle network functions is it functions by inputting the data at one end so you just give an input at one end of the neural network and this input will undergo transformation throughout the network until the final desired output is formed so neural networks work on the principle of weights and bias now if you want to learn more about neural networks and deep learning I'll leave a link in the description section y'all can check out that content on deep learning so the end result of this was that walks will morph succeeded in beating conventional MRI analysis methods the noodle network took seconds to perform MRI analysis the same analysis that will take about hours for a conventional MRI program so you can clearly see there's a huge difference when you implement AI based technologies like deep learning and neural networks so that's how artificial intelligence is used in medical diagnosis now let's understand how artificial intelligence has helped in detecting diseases at an earlier stage so actually artificial intelligence has played a very important role in the early predictions of medical condition such as heart attacks there are many AI B's wearable health trackers that are being developed to monitor the health of a person and display any warnings when the device collides something unusual or something unlikely examples of such variables include Fitbit Apple watch and many others now like this a precaution is always better than cure I think this was a motto behind the latest release of the Apple watch Apple used artificial intelligence in order to build a watch that monitors an individual's health this watch will basically collect data like the person's heart rate sleep cycle breathing rate activity level blood pressure and so on and it keeps a record of all of these measures 24 by 7 so all you've to do is you have to wear the watch and all of this data is automatically collected by the watch so now that you've collected data you know the next step in machine learning is processing analyzing and making predictions from data so this collected data is processed and it's analyzed using machine learning and deep learning algorithms so that you can build a model that predicts the risk of a heart attack so with the help of data you're going to predict whether a person has chances of getting a heart attack or not now before I move on to the next slide I want to mention a story about how Apple watch actually saved a person's life now there was this person known as Scott Gillan and he suddenly had a small heart attack and his life was saved by this Apple watch because it gave repetitive warnings to him regarding his blood pressure I think his blood pressure or his heart rate had increased and he was about to get a heart attack but his life was saved because he received an immediate notification from his Apple watch which showed that his heart rate was increasing at a drastically pace so I stated this example because a lot of us believe that artificial intelligence is a threat to humankind now it is only a threat if you use it as a weapon but if you use it as a solution it can help and save so many lives and millions of dollars so now let's move on and understand how artificial intelligence is used in medical assistants now as an engine for medical assistants has grown the development of artificial intelligence based virtual nurses has increased according to a recent survey virtual Nursing Assistants corresponds to the maximum near-term value of 20 billion u.s. dollars by 2027 and that's a huge amount of money right 20 billion dollars by the end of 2027 a virtual nurse named sense lee is implementing natural language processing speech recognition machine learning and wireless integration with medical devices such as blood pressure cuffs in order to provide medical assistance to patients basically a nurse but it's not physically present it's just what we present feel so sense lebay secretion self-care it provides clinical advice to you whether you should eat this medicine or not does also helps in scheduling appointments for you so essentially it helps in self-care it basically as a virtual nurse for you and it helps you with your medication it tells you when exactly you should take your Medicaid and whether or not you should eat a certain food it gives you a lot of clinical advice so with such revolutions in the field of health care it is clear that despite all the risk and all the so-called threats artificial intelligence is actually benefiting us in many ways I'm a Tory believer that AI will help and improve our situation in each and every domain so we just need to know how to use AI in the correct manner now let's move on and discuss artificial intelligence in decision-making now AI has played a very important role in decision making not only in the field of health care but AI has also improved businesses by studying customer needs and evaluating any potential risk that a business might face a powerful use case of artificial intelligence in decision-making is the use of surgical robots that can minimize errors and any variations and will eventually help in increasing the efficiency of your surgeons one such surgical robot is the dementia I think it's quite aptly named basically this device allows professional surgeons to implement complex surgeries with better flexibility and control than any other invention approaches so the key features of da Vinci it helps in aiding the surgeons with an advanced set of instruments it is used in translating the surgeons hand moments at the console in real time it produces clear and magnified images of the surgical area now guys don't get confused about such surgical robots Dmitri is not a robot that performs the surgery rather it provides a set of instruments that will help in performing the surgery we still haven't developed AI robots or AI based systems that are well capable of performing surgeries on their own there are a couple of robots but they require human intervention and human assistance so dementia is basically a instrument that helps in performing surgeries it provides a set of instruments that can be used in performing complex surgeries so not only do these robots assist in decision making processes they also improve the overall performance by increasing the act you see and the efficiency of the work that has been done so guys those were a couple of real-world applications about artificial intelligence in healthcare now let's move on and understand two of the most important fields I'd say in artificial intelligence which is machine learning and deep learning I've been mentioning machine learning and deep learning throughout the video so for those of you who don't know what machine learning and deep learning is and try to understand what exactly they mean machine learning is a subset of artificial intelligence which provides machines the ability to learn automatically and improve from experience without being explicitly programmed to do so in simple terms add value machine learning is the process of failing machines lots and lots of data so that they can interpret process and analyzes data in order to produce any actionable insights that benefit an organization so the main thing in machine learning is you're going to take lots and lots of data and feed it to your machine and your machine is going to try and understand and interpret this data by using machine learning models and machine learning algorithms and it's finally going to solve a problem or predict an outcome based on that data so that's what machine learning is and if any of you are interested in learning more about machine learning I'll leave a couple of links in the description box y'all can understand in depth about machine learning now let's understand what exactly deep learning is because deep learning is a more advanced concept of machine learning so basically deep learning is a more advanced field of machine learning that uses the concept of neural networks in order to solve more complex problems that require high dimensional data and automated feature extraction so you can see that deep learning is used to solve more complex and higher dimensional data problems and the main thing about deep learning is it makes use of neural networks in order to map your input your output so basically in order to solve a problem in deep learning you make use of neural networks or artificial neural networks to be more precise the neural networks are basically what are the human brain is formed of so scientists basically replicated our brain in the form of artificial neural networks or also deep learning so basically artificial neural networks function exactly how the human brain functions so you've given a lot of input data and this data is then transformed and you apply weights and bias and make the data go through a couple of transformations in order to result in precise output if you want to learn more about the planning I'll leave a link in the description box y'all can check that out as well now let's move on and look at the last topic for today I'll be discussing how deep learning can be used to predict the stage of breast cancer so basically a problem statement is to study a breast cancer data set and to model a neural network classifier that predicts the stage of breast cancer as either malignant or benign malignant means cancerous cells and benign is non cancerous or normal cells so basically we'll be feeding our neural network data set which contains observations and samples of both malignant and benign cells and so basically both cancerous and non cancerous cells are present in this data set and you're going to use artificial neural networks in order to classify them into two separate classes one will continue in cells that are non cancerous and the other will contain cells which are cancerous now guys before I get on with the demo I want to tell you that I'll be using Python in order to solve this problem so for those of you who don't know Python I'll leave a couple of links in the description box and we have a lot of content over Python machine learning and deep learning so you all can go through that content and maybe then come back to this video so enough of theory now let's look at how you can solve a real-world problem by using artificial intelligence concepts such as deep learning so this is what our code looks like so I'll be going through this entire code so that you can understand what exactly is happening over here now you always begin a project by importing the necessary libraries or packages in Python all right so we're going to import the numpy and the panda's library going to import match plot library as well and we'll be importing the sq learn and the sub process and repeat so we so basically here what we're doing is we're importing all the necessary packages that are required for this demo we are going to read our data set our dataset is stored in this data or CSV file and we are going to input this data set into a variable called data after that we'll display the first few observations in our dataset let me show you what that looks like so this is what our dataset looks like we have variables such as ID diagnosis fractural dimensional worst unnamed and so on now there are around 32 or 33 variables if I'm not wrong and all these variables are known as predictor variables the variables that I used to predict the outcome unknown as predictor variables now if you look at this variable diagnosis it has a value called M it also has values called B which are not displayed here because I'm just showing the first few observations in my dataset now this diagnosis variable is our output or our target variable that we are supposed to protect so if the value of diagnosis is M it means that the cancer is malignant or it is a cancerous cell but if the value of diagnosis is B it means that it is a benign tumor or a non cancerous tumor I hope the difference is clear next we will perform data cleaning data cleaning is one of the most important steps in deep learning and in machine learning so we need to get rid of any inconsistent values or any values that are missing or any duplicate values so that our output is as precise as possible so first of all we are dropping our ID variable which is not required and we don't need the ID of a particular cell in order to predict whether a cell is cancerous or not so veges performing that sort of cleaning by getting rid of any inconsistent data or any unnecessary variables that are not needed the step is followed by mapping benign to zero and malignant to one meaning that we're changing the value of this diagnosis Meeropol from M and B to zero and one so if the value is benign and then we are going to set it to zero and if the cancer is malignant then you're going to get an output of 1 that's exactly what we're doing which is changing this into 1 and 0 after that we are going to scale our data set now scaling our data set is very important in order to avoid any biasness in the values so there might be some parameters which are in two digits and three digits whereas there are some parameters like these the value of these parameters are in decimal points they're in zero point zero zero one zero point zero zero seven so when there is a different range of values in your input data set then there is a high chance that your output is going to be very biased that's exactly why you perform ski so we're scaling the data set in this code snippet after that we are creating a feature snippet X so this X feature variable of feature space will contain our data set without the diagnosis variable and we're getting rid of the diagnosis variable which is basically a target variable so that we can predict this variable we don't want this as our input variable so which is getting rid of that after that we are going to create a feed-forward neural network now these are the necessary packages that are needed we are using the Kira's package in order to build the neural network now a feed-forward neural network is basically a neural network in which all the inputs are connected to all the hidden layers now before I continue with what exactly this part of the code means let me just swiftly explain a little bit more about deep learning now in a multi-layer neural network there can be one input layer one output layer and in between you can have a number of hidden layers so first you have your input layer then you have a number of hidden layers and then you have your output layer now input layer is where you feed the input and the outer layer is where you get your output from now in between these input and output layers you have hidden layers in these hidden layers all the processes are take place I will help you map your input to your output so in the hidden layers you basically assign each input a particular weight age so some weight age is assigned to each input and this beta age basically denotes the importance of a particular input variable so more the weight age of a variable the more significant that we will is in predicting the output apart from assigning VT is you also perform summation in the hidden layer summation is basically the process of multiplying your respective inputs with their weight age and adding all of these products and after you perform all of this summation and weight age you then put your input to the activation function now an activation function will basically help you map your input data to your output data so that's a little bit about neural networks so because I'm now going to explain neural networks in depth because that's not the focus of our session today however I will leave a couple of links in the description box you can go through those video so what we're doing is we're importing Kira's and Kira's is one of the most important libraries in python for deep learning we're going to create a sequential model now sequential is a function defined in the Kira's library itself so we're just going to call out that function a sequential model is nothing but a model wherein the input is passed from one layer to the other one after the other it's basically a feed-forward neural network after that we are going to add your input layer we are going to use the rel you activation function so similarly we'll be assigning of the activation function for our other hidden layers and we'll have our output layer which will basically contain only one node because our output variable is a single variable that will have a value of either 0 or 1 also dropout is basically optimization technique in a deep learning that will help you better predict the output in this technique what you do is you randomly drop out a couple of nodes in your neural networks in order to avoid overfitting of your data so guys that was us building the model after that what we'll do is we'll split our dataset into two parts this process is also known as data splicing and what you do here is you split your input data into two parts one is for training and the other is for testing so here we have assigned the validation split of 0.33 which means that 33 percent of the data will be used for validation or testing and the remaining 67 percent of the data will be used for model training after that we are implementing the cross validation technique and we're going to use of k-fold cross-validation technique so guys again for this I'll leave a couple of links in the description box so that you can understand what exactly I'm doing here now once you've implemented these techniques you need to finally run the model this model outfit will help you fit your data into your neural network so we're basically passing our training data set into our neural network and we're assigning a box of 10 and bath size of 10 a pox and bad sides would basically help you split your data into smaller batches so that you know accuracy can be calculated on each of these batches if box are basically the number of iterations that you'll run on your model once you run this you will get an output like this so we basically signed 10 a box meaning that 10 iterations this is basically training your neural network now you can see that in the first box we have an accuracy of 95% approximately 95% then we have in our second Epoque 96% similarly if you keep going down we have 97 and then in between again it drops down to 96 97 98 and 96-97 and so on this happens because of the batches we're using different batches on each of these boxes so guys a point to note here is if you want to train your model in a better way then you can also increase your number of iterations or you can also increase or decrease your batch size and understand if that helps in any way apart from this we have dropout techniques which basically a technique in order to avoid overfitting of your data if your data over fits into the model it means that you cannot predict your outcome very precisely we're finally printing our final accuracy over here and so our accuracy is around ninety eight point six percent for two fold cross validation so as 98 percent is actually a pretty good value if you wish to improve your model any further you can perform parameter tuning and optimization techniques such as drop of method with a more efficient drop value so guys that was our entire demo so we basically built a artificial neural network that helped in classifying a tumor as malignant or benign we built a model with an accuracy rate of 98% which is a really good value for a startup and also told you how to improve the efficiency of the model so for now I think this was good enough for our artificial intelligence and healthcare demo got any more detailed demos on such topics please let us know in the comment section also if you have doubts and leave them in the comment section so guys that's all for today and I hope or if you enjoy the video until next time happy learning I hope you have enjoyed listening to this video please be kind enough to like it and you can comment any of your doubts and queries and we will reply them at the earliest do look out for more videos in our playlist and subscribe to any rekha channel to learn more happy learning

1 Comment

  1. Got a question on the topic? Please share it in the comment section below and our experts will answer it for you. For Edureka AI & ML Engineer Masters Program Course curriculum, Visit our Website: http://bit.ly/2wziDSd

Leave a Reply

Your email address will not be published. Required fields are marked *