Leveraging Client-side Computation for Biomedical Machine Learning Inference with Open Neural Network Exchange

Carlos Vega, Beatriz Garcia Santa Cruz, Frank Hertel, Venkata Satagopam

Keywords: Data privacy, deep learning, healthcare, translational medicine, biomedicine

TL;DR: Client side use of ML models to protect data privacy

Abstract: Biomedical settings often involve large datasets and sensitive information confined within institutional boundaries that limit outbound data transfers. Low-resource areas with limited biomedical resources and unreliable internet access also struggle to forward their data for computation. This paper advocates for executing machine learning models locally on conventional computers leveraging new technologies such as Open Neural Network Exchange. The proposed design enables secure, client-side execution without data transfer, addressing privacy and bandwidth constraints while enhancing the portability of biomedical models execution.

Submission Number: 24