Work Item
ASTM WK77015

New Practice for Machine Learning Training Dataset Structuring and Revision

1. Scope

The Working Group aims to develop a Good Documentation Practice (GxP) for machine learning and AI in the healthcare industry. Specifically, the Group will focus on: i) how software that leverages neural networks (or other machine learning techniques) with supervised or unsupervised learning should be created, ii) secure an efficient pathway for FDA clearance, and iii) create platforms that can accept evolving datasets with embedded regulatory safeguards. We will investigate and create a standard that outlines how training and test-sets can be dynamic, change-notice type documents instead of an integral part of the code that utilizes them. FDA has released a guidance on the need for AI/ML based systems to be regulated with increasing use in medical devise technologies. The Group will proactively create a GxP that outlines how to develop, use, and update data structures in software as a medical device.

Keywords

AI; ML; GxP

Rationale

FDA has released a guidance on the need for AI/ML based systems to be regulated with increasing use of the technologies in medical devices. We would like to proactively create a GxP type document to outline how to bring, use, and update data structures in software as a medical device.

The title and scope are in draft form and are under development within this ASTM Committee.

Details

Developed by Subcommittee: F04.38

Committee: F04

Staff Manager: Kathleen Chalfin

Work Item Status

Date Initiated: 05-21-2021

Technical Contact: Gokce Yildirim

Item: 000

Ballot:

Status: