Image Compression in Morphometry Studies Requiring 21 CFR Part 11 Compliance: Procedure Is Key with TIFFs and Various JPEG Compression Strengths

Abstract

This study aims to compare the integrity and reproducibility of measurements created from uncompressed and compressed digital images in order to implement compliance with 21 CFR Part 11 for image analysis studies executed using 21 CFR Part 58 compliant capture systems. Images of a 400-mesh electron microscope grid and H&E stained rat liver tissue were captured on an upright microscope with digital camera using commercially available analysis software. Digital images were stored as either uncompressed TIFFs or in one of five different levels of JPEG compression. The grid images were analyzed with automatic detection of bright objects while the liver images were segmented using color cube-based morphometry techniques, respectively, using commercially-available image analysis software. When comparing the feature-extracted measurements from the TIFF uncompressed to the JPEG compressed images, the data suggest that JPEG compression does not alter the accuracy or reliability to reproduce individual data point measurements in all but the highest compression levels. There is, however, discordance if the initial measure was obtained with a TIFF format and subsequently saved as one of the JPEG levels, suggesting that the use of compression must precede feature extraction. It is a common practice in software packages to work with TIFF uncompressed images. However, this study suggests that the use of JPEG compression as part of the analysis work flow was an acceptable practice for these images and features. Investigators applying image file compression to other organ images will need to validate the utility of image compression in their work flow. A procedure to digitally acquire and JPEG compress images prior to image analysis has the potential to reduce file archiving demands without compromising reproducibility of data.

Adaptative, signal-preserving compression of microscopic images using noise modeling in the wavelet domain and JPEG2000 coding

“…Hence, perceptual image quality may not be used with confidence to validate compression algorithms of microscope images, if data integrity is to be preserved. The problem of validation of these algorithms to meet requirements of 21 CFR part 11 for automated image analysis has been addressed in several publications [7,19]. A common approach has been to demonstrate that irreversible compression did not affect results of specified measurements which used the images as input data.…”

Digital Images Are Data: And Should Be Treated as Such

“…The original data may be needed to satisfy regulatory requirements (e.g., CFR 21 part 11 (72, 73), HIPPA (74), Forensics (75, 76)) for maintaining the data.…”

Avoiding Twisted Pixels: Ethical Guidelines for the Appropriate Use and Manipulation of Scientific Digital Images

“…Research that falls under the United States Food and Drug Administration’s “Final Rule on Electronic Records and Electronic Signatures” (21 CFR part 11) is subject to specific requirements for maintaining the integrity of the original image (Horn and Sterchi 2005; Tengowski 2004). The nature of the work done by forensic scientists (Scientific Working Group Imaging Technology 2004) and in clinical/diagnostic labs (Pritt et al 2003) requires particular attention to protecting the integrity of the original image.…”