Overview / Usage

Nephron**AI **- early detection and etiology of chronic kidney disease using deep learning. Overview

Chronic kidney disease, or CKD, causes more deaths than breast cancer or prostate cancer. It is the under-recognized public health crisis. It affects an estimated 37 million people in the U.S. (15% of the adult population; more than 1 in 7 adults) and approximately 90% of those with CKD don’t even know they have it. 1 in 3 American adults (approximately 80 million people) is at risk for CKD. CKD is more common in women (15%) than men (12%). CKD is the 9th leading cause of death in the U.S. In 2016, over 500,000 patients received dialysis treatment, and over 200,000 lived with a kidney transplant.

For long, existing research works involving artificial intelligence have focused solely on the detection of a disease. Understanding the cause of the disease is equally and probably more important. However, to tackle a disease on such a large scale a more comprehensive work must be done. Therefore, I came up with the idea to tackle chronic kidney disease using a three-step approach – early detection, knowing the underlying cause, and recommending a lifestyle approach to minimize risks – using one of the most advanced technology in computational intelligence, ‘deep learning’. The work has been implemented to make a web-based solution – NephronAI, which is accessible to anybody for free with the ease of a web browser.

Methodology / Approach

​

NephronAI works in three steps: Predict Disease

Using AI Algorithm built on Microsoft's Azure Machine Learning Architecture, our AI systems can predict whether a patient is affected by Chronic Kidney Disease or not.

• Input simple medical records
• Our AI Algorithm processes it
• You are notified if you are vulnerable

Predict the Cause

If our AI Algorithm figure out that you are vulnerable and have been affected by Chronic Kidney Disease then we shall investigate the cause also.

• Answer basic lifestyle questions
• Provide your disease history
• Another layer of AI will predict the cause

Recovery Strategy

Now, that we have told you about your diagnose is as well as it's cause, then let's fight it. We'll recommend you certain lifestyle measures that you can take to at least lessen the ill effects of the disease.

• Based on the cause predicted
• You'll get some lifestyle tips
• Follow them and keep up the fighting spirit

Technologies Used

Technologies Used

I have used the following technologies for creating the project:

• Azure Machine Learning (deployment)
• PyTorch (deep learning model)
• Matplotlib and Seaborn (visualization of medical data)
• Intel DevCloud - Jupyter Lab (development environment)

​

Future Prospects

I would wish to pursue further research on the project with the following objectives:

1. How artificial intelligence can be applied to benefit the process of ‘hemodialysis’ (the treatment to chronic kidney disease).
2. Investigate into other parameters which may help us in more accurate and feasible diagnosis of chronic kidney disease.
3. How to fully integrate this project into real-world healthcare systems to increase adoption and accessibility.

Documents and Presentations

Repository

https://github.com/ayushanand18/nephron-ai

Other links

• Research Paper
• NephronAI [Live]