Artificial Intelligence, and in our case, machine learning enable tremendous improvements in many areas, especially in the health sector as we will see in this project dedicated to breast cancer.



The following project is based on supervised learning, as a matter of fact, we first have to train our model to recognize key characteristic of breast cancer.Then, once our model will be trained, it will be able to identify (with a certain degree of precision) breast cancer.

One may also use a pre-trained model but the aim of this project is to learn how to do it from scratch.



If you want to have fun with the “almighty” google’s API, feel free to click on the logo below

Concerning the technical part of our project, we're going to rely on three entities

ml5.js

A JS library based on TensorFlow (Open source ML library)

MobileNet

Powerful and fast Convolutional neural network (CNN)

p5.js

A JS library which makes programming accessible

As aforementioned, we need data to train our model, please find-below two websites hosting datasets (images) of breast cancer:

CIA - KAGGLE

Step by step process to create our image classification tool

Create a file entitled index.html and enter the following code

Create a file entitled sketch.js and enter the following code

To obtain the comprehensive documentation about image classification with ml5.js click the logo below

We are using mobile net which is a pre-trained model, as mentioned above we want to “create” our own model, more precisely, train MobileNet with our datasets.


In our case, we want to “re-feed” the model with images of patient with breast cancer and images of patient without breast cancer. In order to do so, we must teach our A.I to differentiate positive and negative results regarding breast cancer it can be done by using pre-labeled datasets.

Example of labelled images

To learn everything about retraining MobileNet lick the logo below

If you’re interested in mastering image classification with ml5.js and mobile net feel free to take a look at the video below


To sum-up

This project could be very useful in helping doctor during their diagnosis. Needless to say that this kind of technology needs to be monitored by humans to avoid misinterpretation and false-positive.

I see this project as a way to enhance the human-computer relationship and gather both intelligences to achieve better results (i.e greater precision in breast cancer detection).

A.I for common good® : Detect Breast Cancer

A project by Roméo LÉON