High Throughput JPEG 2000 (HTJ2K): A Practical Path to Faster, Scalable Medical Imaging

As medical imaging datasets grow in size, resolution, and complexity, performance has become a primary constraint across radiology and digital pathology environments. High-resolution CT, MR, and PET studies routinely contain thousands of slices, while digital pathology whole slide images (WSI) often exceed multiple gigapixels per specimen. These workloads require rapid encoding, decoding, and streaming to support clinical workflows, cloud migration, and AI-driven analysis.

High Throughput JPEG 2000 (HTJ2K), defined in the International Organization for Standardization standard ISO/IEC 15444‑15, was developed to address these performance constraints. By modernizing the most computationally intensive stage of the JPEG 2000 pipeline, HTJ2K delivers substantial throughput improvements while maintaining full compatibility with the existing JPEG 2000 ecosystem.

With the Dicom Systems Unifier platform, healthcare organizations can archive, route, and convert existing imaging datasets to HTJ2K at scale and deploy faster imaging workflows across radiology, digital pathology, enterprise archives, and cloud infrastructure.

Why Performance Now Matters More Than Compression Ratios

Traditional JPEG 2000 remains a high-quality, standards-based compression format widely used across medical imaging. It supports lossless compression, progressive decoding, and multi-resolution access, making it ideal for diagnostic workflows. However, its original design prioritized compression efficiency rather than computational throughput.

This limitation becomes increasingly visible in modern environments:

Radiology performance constraints

• Viewer responsiveness is affected by decode latency when loading multi-series studies.
• Large CT and MR datasets require significant CPU resources to decompress
• Cloud migration pipelines experience bottlenecks during bulk data transfers
• AI pipelines spend substantial time decoding image data before analysis

Digital pathology performance constraints

• Whole slide images often exceed 500 MB to 4 GB per slide
• Viewing requires the rapid decoding of huge pyramidal image structures
• Network and compute latency directly impacts pathologist productivity
• Real-time collaboration and remote review demand high-throughput streaming

In these environments, throughput, latency, and scalability have become more critical operational metrics than marginal improvements in compression ratio.

HTJ2K: Designed for Modern CPU Architectures

HTJ2K (ISO/IEC 15444-15) introduces a high-throughput block transform that replaces the traditional JPEG 2000 EBCOT block coder with a new high-throughput block coder (FBCOT). The wavelet transform (DWT) is the same in both J2K and HTJ2K; HTJ2K accelerates the block-coding stage, not the DWT, optimized for:

• SIMD vector processing on modern CPUs
• Multi-core parallel execution
• Predictable memory access patterns
• Efficient scaling across cloud compute environments

Unlike alternative codecs, HTJ2K maintains full compatibility with existing JPEG 2000 standards and infrastructure. The codestream remains compliant with the Digital Imaging and Communications in Medicine (DICOM) standard maintained by the National Electrical Manufacturers Association (NEMA).

Ensuring:

• Bit-exact lossless compression
• Standards-compliant DICOM integration
• Compatibility with existing viewers and archives
• No loss of diagnostic fidelity

HTJ2K is an evolution of JPEG 2000 rather than a replacement, allowing organizations to modernize performance without disrupting established workflows.

Additional technical background is available from the JPEG Committee and the Digital Pathology Association.

Benchmark testing demonstrates that HTJ2K performance improvements scale with image size and complexity, with the most dramatic gains occurring in digital pathology workloads.

Radiology Imaging Performance Improvements

Radiology images with HTJ2K typically range from tens to hundreds of megabytes per study.

Measured improvements include:

• Decode speedups measured at ~2.0× to 6.8× across common radiology sizes
• 2× to 6× faster encoding performance
• Reduced latency when loading multi-series studies
• Improved scalability across multi-core processing systems. Performance varies by modality, CPU features, codec build flags (SIMD), and threading configuration

These improvements directly enhance radiologist productivity and viewer responsiveness.

Digital pathology is among the most demanding imaging workloads in healthcare. Pathology information can encompass various data types, including histopathology, cytopathology, clinical pathology, and molecular pathology. Whole slide images commonly contain:

• 1 to 4 gigapixels per image
• 500 MB to 4 GB compressed file sizes
• Multi-resolution pyramidal image structures
• High bit-depth color channels

Given these massive data demands, HTJ2K delivers game-changing performance gains for digital pathology workflows.

Measured performance improvements with HTJ2K include:

• 8K region: ~7.8× faster decode and 15.8× faster encode (HTJ2K vs J2K in the benchmark environment)
• 16K region: 18.1× faster decode and 22.7× faster encode
• Summary: pathology images demonstrate ~8–23× speedups at scale (vs 2–7× typical for radiology), depending on image size and workload composition
• Near-linear scaling with CPU core count

These improvements enable real-time viewing, faster collaboration, and scalable cloud-based digital pathology deployments.

The transition to digital pathology is accelerating globally, supported by initiatives from organizations such as the College of American Pathologists and the U.S. Food and Drug Administration.

Digital pathology enables pathologists to view and analyze tissue samples digitally rather than through physical microscopes. When combined with cloud computing, this approach allows for remote diagnosis, collaboration, and AI-assisted analysis.

However, the extreme size of whole slide images introduces significant performance challenges.

Key digital pathology characteristics include:

• Extremely large image dimensions
• Multi-resolution storage structures
• High data throughput requirements
• Continuous streaming during navigation

Cloud-based viewers must rapidly decode and stream image tiles to maintain a responsive diagnostic experience.

HTJ2K addresses this challenge by significantly reducing decode latency and improving streaming performance without sacrificing image quality.

Beyond raw throughput gains, HTJ2K optimizes overall CPU efficiency, delivering substantial cost and scalability advantages in imaging workflows.

Key efficiency improvements include:

• Fewer CPU cycles required per pixel
• Lower peak CPU utilization during batch conversion
• Improved performance-per-core in cloud environments
• Reduced infrastructure requirements for image processing

These improvements translate into measurable operational benefits:

• Lower cloud compute costs
• Faster dataset migration
• Improved scalability of AI pipelines
• More predictable performance under load. In cloud environments, compute costs often exceed storage costs, making throughput optimization a critical factor.

HTJ2K achieves its performance improvements with a modest reduction in compression efficiency. Typical file size increases range from approximately 5 to 7 percent larger than traditional JPEG 2000.

In practice, this trade-off is acceptable because:

• Storage costs continue to decline
• Compute and latency costs dominate operational expenses
• Lossless image quality remains unchanged
• Workflow efficiency improvements provide greater overall value

HTJ2K shifts optimization from storage-centric compression toward end-to-end workflow performance.

HTJ2K provides clear benefits across multiple medical imaging domains.

Radiology benefits

• Faster viewer load times
• Improved PACS performance
• Faster cloud migration pipelines
• Improved AI inference throughput

Digital pathology benefits

• Faster whole slide image viewing
• Reduced latency during navigation and zoom
• Improved performance for remote collaboration
• Scalable cloud-based pathology archives

These gains in radiology and pathology speed up clinical workflows while improving infrastructure efficiency to meet growing imaging needs.

The updated benchmarks highlight an important finding: on dense diagnostic tissue, JPEG-LS and HTJ2K achieve similar lossless compression (~2.6–2.7×). The key difference is throughput and streaming: HTJ2K-encoded is ~11× faster, and HTJ2K-decoded is ~5× faster than JPEG-LS in the test environment, and HTJ2K supports progressive streaming (when using RPCL and viewer/server support).

• HTJ2K (especially RPCL) is ideal for interactive viewing and high-throughput ingestion/migration.
• JPEG-LS is a strong option for a cold archive where streaming is not required.

HTJ2K/J2K can enable progressive viewing and region-of-interest workflows, but this requires system-level support (e.g., RPCL progression order + appropriate viewer/server partial delivery, such as byte-range requests or JPIP). Simply encoding HTJ2K does not automatically guarantee progressive UX.

For streaming viewers, the recommended HTJ2K transfer syntax is 1.2.840.10008.1.2.4.202 (HTJ2K Lossless RPCL)

Adopting HTJ2K requires a scalable, standards-compliant conversion layer that can handle large imaging datasets without disrupting clinical operations.

The Unifier Enterprise Imaging Platform provides a vendor-neutral workflow engine that enables high-throughput conversion and integration of HTJ2K into existing imaging environments.

Unifier enables healthcare organizations to:

Convert existing datasets at scale

• Automated batch conversion of radiology and pathology datasets
• High-throughput processing optimized for multi-core infrastructure
• On-demand conversion workflows for legacy archives

Preserve clinical and regulatory integrity

• Full preservation of DICOM metadata and unique identifiers
• Standards-compliant conversion workflows
• No impact on diagnostic image fidelity

Support radiology and digital pathology workflows

Radiology integration:

• PACS integration for CT, MR, CR, DR, PET, and ultrasound
• VNA integration for enterprise imaging archives
• Viewer optimization for faster image loading

Digital pathology integration:

• Support for whole slide imaging formats and pyramidal structures
• Conversion and routing of digital pathology DICOM objects
• Integration with LIS, pathology viewers, and AI platforms

Deploy across a hybrid infrastructure

• On-premises deployment
• Cloud deployment
• Hybrid imaging environments

Unifier integrates with PACS, VNAs, cloud storage, and AI pipelines, enabling healthcare organizations to modernize imaging infrastructure without replacing existing clinical systems.

Dicom Systems Unifier enables healthcare organizations to convert existing imaging datasets to HTJ2K and deploy faster, scalable workflows across radiology and digital pathology. Unifier automates high-throughput conversion while preserving DICOM metadata, diagnostic image fidelity, and full compliance with standards. Organizations can accelerate image access, modernize imaging infrastructure, reduce operational costs, and support AI-driven diagnostics without disrupting clinical systems.

With Unifier, healthcare organizations can:

• Convert legacy JPEG 2000 datasets to HTJ2K using automated, high-throughput workflows
• Improve viewer responsiveness and reduce image load latency across radiology and digital pathology
• Accelerate cloud migration, AI pipelines, and enterprise archive performance
• Preserve DICOM metadata, image fidelity, and full standards compliance
• Deploy across on-premises, cloud, and hybrid environments without workflow disruption

Unifier provides the workflow orchestration layer required to operationalize HTJ2K and deliver measurable performance improvements across enterprise imaging environments.

See how Dicom Systems Unifier accelerates enterprise imaging with HTJ2K.