Overview / Usage

The Acute Lymphoblastic Leukemia (ALL) Detection System for Magic Leap 1 combines disruptive technologies to provide a proof of concept showing how Artificial Intelligence, the Internet of Things and Mixed Reality/Spatial Computing can be used together to improve medical diagnostics.

The motivation for this project was the interest in combining Deep Learning using Intel Technologies with Magic Leap 1 to create a unique way of carrying out diagnostics tests using Computer Vision and Mixed Reality.

Methodology / Approach

• The interface is developed in C# using the Unity3d Game Engine and deployed to the Magic Leap 1 hardware.
• The classification model used in this project is the Acute Lymphoblastic Leukemia oneAPI Classifer which was trained on Acute Lymphoblastic Leukemia images of peripheral blood samples.
• The dataset used to train the model Acute Lymphoblastic Leukemia Image Database for Image Processing dataset by Fabio Scotti, Professor at the University of Milan.
• The classifier was trained using Intel® Distribution for Python* and Intel® Optimization for TensorFlow*.
• Once the model is trained it is frozen which allows the model to be converted to an Deep Learning Network Intermediate Representation using Intel® Distribution of OpenVINO™ Toolkit and then deployed to a Raspberry Pi 4.
• Attached to the Raspberry Pi 4 is an Intel® Neural Compute Stick 2 (Intel® NCS2) which utilized the Intermediate Representation to provide accelerated, high performance inference on the edge.

Technologies Used

• Magic Leap 1
• Unity3d Game Engine
• Intel® Distribution for Python*
• Intel® Optimization for TensorFlow*
• Deep Learning Network Intermediate Representation
• Intel® Distribution of OpenVINO™ Toolkit
• Raspberry Pi 4
• Intel® Neural Compute Stick 2 (Intel® NCS2)

Repository

https://github.com/aiial/all-detection-system-for-magic-leap-1

Other links

• Official Project Page