Template: Instructions On Data Annotation (Machine Learning / AI)

Sven Piechottka
Updated April 26, 2024
0 comments

Template Download

This is a free template, provided by OpenRegulatory.

If you are a user of Formwork, our eQMS software, you can save a lot of time by choosing “QMS” on the top menu and “OpenRegulatory Templates” on the left menu, and then opening the relevant folder to find this template ready to load into Formwork.

If, for some mysterious reason, you’re using a different QMS Software, you can also simply download this template – specifically, as Word (.docx), PDF, Google Docs or Markdown file. Scroll down for a preview!

The template license applies (don’t remove the copyright at the bottom, don’t re-use this for commercial purposes).

Lost In Regulation? Book a Free 30-Minute Consulting Call.

Unsure how to get started and how to get your EU MDR medical device certified?
We’ve already helped hundreds of companies with their MDR compliance.
Book a free 30-minute consulting call and let’s discuss how you can get your compliance done efficienty.

Book a free 30-minute consulting call
Template preview

Intended Purpose:

The Instructions on Data Annotation specify the way in which datasets shall be annotated by medical experts to apply labelling required for the training of the machine learning model. It provides a standard annotation style, including the accuracy and form of annotation. Annotation requirements shall be specific enough to enable any given third party to evaluate the quality of annotation. The guidelines shall als provide a reasoning for the selection of labels and the methodology used to derive the ground truth for model development.

Relevant other documentation:

  • SOP Machine Learning Model Development
  • Intended Use
  • Instructions on Data Acquisition

Note: This is an example of content that could be considered for a specific use case. All content has to be replaced and customized to your specific product use case. You could also consider documenting this in format of a slides presentation in order to easily include helpful images.

General Information

Annotation experts shall draw segmentations around the following findings:

  • Tumors
  • Other suspicious findings
  • (…)

Annotation experts must make sure to segment all visible findings as missing a possible findings will lead to an algorithmic classification as unsuspicious and decreased model performance.

Annotation experts shall NOT segment the following findings:

  • Benign finding XYZ
  • (…)

Tumors

Annotation instructions:

  • Ensure that the segmentation captures the entire tumor
  • (…)

Document the following information:

  • Tumor classification
  • Biopsy information
  • Confidence level
  • (…)

Other Suspicious Findings

Annotation instructions:

  • Ensure that the segmentation captures the entire finding
  • (…)

Document the following information:

  • Type of suspicious finding
  • Biopsy information
  • Confidence level
  • (…)

Template Copyright openregulatory.com. See template license.

Please don’t remove this notice even if you’ve modified contents of this template.

Template preview

Comments

Leave the first comment