NillionHealthcare

Nillion Health is a program that demonstrates the power of blind computation in a healthcare setting. Making use of Nillion's multiparty computation, the program provides a breast cancer image classification test performed over multiple health providers. The objective is to make use of modern cryptography to prevent barriers in the healthcare industry and ensure the world's most important data type, medical imaging, can be shared in a way that is fair and secure.

There are two primary sections, a test computed over a full dataset and a test computed on Nillion's protocol on combining multiple datasets.

Full Dataset Test 550+ image data instances 30 parameters and 1 target value (Diagnosis) 80/20 Split, (80) Training data, (20) Testing 3 Figures of Plot Distributions - Full, Small Subset, Large Subset Simple classification model use for computing thetas and test predictions Single randomly selected test instance used, not a full test evaluation. Do not use program for predictions. Purpose of testing is to compare calculated values.

Multi Party Nillion Test Small Subset (25%) and Large Subset (75%) 30 parameters and 1 target value (Diagnosis) Weighted average method used when combining thetas from subsets A scaling factor was applied to satisfy integer requirements in Nillion's Nada program