Compression assessment based on medical image quality concepts using computer-generated test images

Area of Science:

• Medical Imaging
• Image Compression
• Radiology

Background:

• Lossy compression algorithms are crucial for medical imaging efficiency.
• Assessing 'visually lossless' thresholds and developing time-efficient methods are vital for clinical adoption.
• Current observer performance studies are data-intensive and time-consuming.

Purpose of the Study:

• To propose quantitative measurements for assessing medical image compression.
• To determine the visually lossless threshold for the Joint Photographic Expert Group (JPEG) lossy algorithm.
• To offer a cost- and time-efficient alternative to observer performance studies.

Main Methods:

• Utilized region of interest (ROI) operations on computer-generated test images mimicking radiographic characteristics.
• Assessed the lossy Joint Photographic Expert Group (JPEG) algorithm within a healthcare telematics application.
• Measured image quality parameters including contrast discrimination, noise levels, line-pair amplitude, and gray level differences.

Main Results:

• A compression ratio of 15 was identified as the visually lossless threshold for JPEG lossy compression.
• Low contrast discrimination remained unaffected up to this threshold.
• Image noise decreased, high contrast line-pair amplitude decreased by <3%, and gray level differences were <1%.

Conclusions:

• Quantitative measurements provide a cost- and time-efficient method for compression assessment in medical imaging.
• This approach offers insights into the type of loss incurred during compression.
• Test image adaptation allows for tailored assessments across different imaging modalities and clinical needs.