Online Healthcare Community With AI Based Disease Prediction

ABSTRACT

Rise in the demands of globally accessible healthcare services and resources has been a salient issue from last few decades. Lack of medical expertise in remote settlements further increases the problem. A potential solution to this issue is to set up online interaction between the consumers and the service providers. This project is aimed at development of an online healthcare community to provide people with affordable and easily accessible healthcare solutions. Setting up an interactive environment between the patients and medical experts along with AI models will allows users to seek help and guidance for their issues. Advancement in artificial intelligence has given systems the capabilities of predicting a probable disease. The system uses machine learning models to analyze and predict the potential health issues of its users such as heart illness and diabetes.

INTRODUCTION

There has been a dramatic rise in internet users globally which has increased the social interactions of people all over the world by a great extent. Social media can be effectively used to provide medical expertise and resources to remote locations all over the world. This project is based on the idea of creating a social relationship on a web platform to connect the patients, medical experts and service providers globally, where the users can seek information about their concerns and learn methods for the prevention and cure of certain diseases. Diverse sources of relevant information located in social media, ICQ, chats, health forums, etc. are then searched for, increasing the chances of finding requested knowledge, facts and evidence, and providing social support and even possible recovery for those in a vulnerable health situation.

This is because social media helps to achieve a better perspective about health problems and eliminate sources of concern. Posting, sharing, and commenting on health-related issues, joining or developing online health communities, and exchanging information about health issues are social reinforcements that enable increased access to information and social advancement and prompt a “health empowerment” process. The second aspect of the project is based on the use of AI to let the users know the probability of any health issue to them. Machine learning techniques are implemented in the system to predict the possibility of certain diseases such as diabetes and Heart issues. The users will get an aid in tracing their health status and shaping their health concerns for the future. The proposed system is still useful for certain kinds of diseases which will be extended in future.

LITERATURE REVIEW

Social media platforms such as Wikipedia, Facebook, LinkedIn, YouTube and Twitter have played a vital role in bringing new possibilities of co-creation and communication between people all over the world. This advancement can benefit the Health sector in numerous ways such as providing online platforms for users to discuss healthcare, diseases, prevention and cures, sharing experiences and knowledge about different cases. Web platform can help people get indulged in their own care and exchange information with other users as well as gaining support from the medical experts. Online communications between patient and doctors are advantageous as it opens gate for people at the extreme ends of the world to get with each other for the purpose of consultation. Online Health Communities (OHCs) are an Internet based platforms that unites groups of individuals with a shared goal or similar interest regardless of their geographical positions. These groups include patients with particular conditions, group of professionals and service providers. Members may get connected on the basis of their requirements and preferences without being concerned with physical distances or inabilities.

OHCs provide various ways for communications which include messages, Discussion forums one to one or one to many talks all through the internet. PatientsLikeMe is such an OHC with life changing conditions for them who share their experiences and self-report their diseases symptoms on a regular basis. These data provide the professionals with new variations in symptoms severity and understanding about the particular disease. Merging the OHCs with AI further increases the efficiency and availability of medical services. Several diseases such as Diabetes, Heart illness, Lung Cancer etc can be predicted using AI models trained on respective datasets. Use of Convolution Neural Networks and multilayer perceptron for prediction of Lung cancer, Diabetes using Diabetic retinopathy and Heart failures has given satisfactory results over different types of dataset and can be used effectively with the OHC platforms.

DESIGN AND PROCEDURE

SOCIAL NETWORK

The system is based on an online application where the users will interact in real time environment. Users can signup in the system as patients or doctors. The three different models of interactions are as follows:

• DOCTOR – PATIENT: This will be the traditional interaction between the information seekers and the service providers. Patient can seek guidance by the registered doctors on the network through direct chats and discussion forums.
• PATIENT-PATIENT This will be a peer group interaction between the patients for the purpose of sharing of information about a case or a disease. This will help the users know about the history of their issues and curing methods.
• PATIENT – AI This interaction is between the patient and the trained AI model. The model predicts the possibilities and severness of an health issue to the user according the data entered by him or her.
• AI BASED DISEASE PREDICTION Artificial Neural Networks are being used to make the predictions about two particular cases that are Heart related diseases and Diabetes. a. MLP for Heart Disease prediction b. Convolutional Neural Networks for Diabetic Retinopathy

METHODOLOGY

Heart Disease Prediction using Multilayer Perceptron Multilayer Perceptron Back Propagation Algorithm of Neural Networks produces an effective ANN training in conjunction with some optimization techniques like gradient descent. The method computes the depth of the loss function in the input data with respects to all the weights in the network. The gradient techniques are then applied to the optimization methods to adjust the weights to minimize the loss function in the network. Hence the algorithms require a known and a desired output for all inputs in order to compute the gradient of loss function. Architecture of the Neural Network: MLP has the same architecture of Feed-Forward back propogation for supervised training. In general, MLP network contains an input, one or more hidden layers and an output layer.

Algorithm MLPBPA

Step 1: Input the Heart Diseaes dataset into MLPBPA

Step 2: Set the class attribute (num) as target value and pass onto the MLPBPA

Step 3: Call trainbr, trainlm, traingdx, trainscg with Learngdm, MSE and purelin function to train, adpt train, fine tune performance and to transfer input/output respectively

Step 4: Set the number of default epochs and goal as 10 and 0

Step 5: train the network until the target reached to desired output

Step 6: If (target! =output) reinitialize the network and train network

Step 6. 2: Increase the number of neurons

Step 6. 3: Increase epochs, goal, number of hidden layers, transfer function and training algorithm

Step 7: else stop the execution.

Diabetic Retinopathy

Diabetic retinopathy is a compilation of diabetes that effect the eyes, PAGE 7 it is caused by damage to the blood vessels in the tissue at the back of the eye(retina), about 1 million cases are diagnosed in India and a lot more are left unchecked because of the lack of professionals and infrastructure. Convolutional Neural Networks are used for classification of different stages of diabetes. Convnet Architecture: RESULT: loss function graph: Accuracy vs Loss PAGE 8 We can clearly see that the accuracy is increasing and loss is decreasing. There is a noise in the graph it due to the fact that we are doing mini-batching. Final accuracy on train-set(after training on 1of10 dataset): 85~90%

RESULTS

The AI models were trained successfully and test predictions were made on custom data. Also the social network was implemented using Django framework. Implementations are shown below: SOCIAL NETWORK.

CONCLUSION

An online interaction platform will prove to be a potential solution over the growing demands of healthcare services over the globe. Users can get involved into discussions with fellow users and medical experts to seek information about their cases and get help by the millions of service providers over the world. Addition of AI support has further improved the system by giving the users an ability to get to know about the possible future issues and thus take respective precautions. Artificial Neural Networks has provided an accuracy of about 90% in both the considered cases and thus the models are acceptable with little improvement.