|
Work item:
|
F.Med-AI-Colour
|
|
Subject/title:
|
Requirements on Colour Data Representation for AI Development and Utilization of Medical Imaging Data in Telehealth
|
|
Status:
|
Under study
|
|
Approval process:
|
AAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2026-10 (Medium priority)
|
|
Liaison:
|
IEC, ISO
|
|
Supporting members:
|
The University of Tokyo, OKI
|
|
Summary:
|
In the medical field, the importance of leveraging AI to improve the efficiency of healthcare professionals is steadily increasing. However, the development of AI requires a large volume of training data, and close collaboration across multiple facilities and participants is essential for generating such data.
Currently, in the use of colour for medical imaging, different devices and software adopt varying approaches to colour reproduction. As a result, even images of the same subject often display inconsistent colour representations, which in turn reduces the efficiency of AI development workflows.
To address these challenges, colour normalization is one effective approach. By using normalized colour images, AI developers can reduce the burden of data collection for each device, and healthcare professionals can also lessen the workload associated with data creation. Additionally, the accuracy of AI analysis can improve as noise from colour differences due to devices and software is minimized, enhancing reproducibility and precision. This will increase the reliability of medical image processing and display, ultimately contributing to the improvement of healthcare services for patients.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2025-11-13 14:32:17
|
|
Last update:
2026-01-06 10:10:40
|
