Gigabyte-scale image files avoid bottlenecks with JPEG2000

March 17, 2008

Radiology is not the only medical discipline struggling to handle massive digital imaging studies. Pathology and cytology may face even more severe problems.

Radiology is not the only medical discipline struggling to handle massive digital imaging studies. Pathology and cytology may face even more severe problems. Image files produced by microscope scanners are typically on the gigabyte scale. The sheer size of the data is the main bottleneck in the development of digital image applications in clinical pathology and cytology. "The computational demands of virtual microscopy images differ to a large extent from those of radiology and other medical imaging because of the larger size of the files, which can exceed tens of gigabytes," said Dr. Jorma Isola, a professor of cancer biology at University of Tampere in Finland. Virtual microscopy is the viewing of entire microscope specimens on a workstation.Isola coauthored a paper that describes the use of JPEG2000 file format to manage images 100 to 1000 times larger than typical x-ray, CT, or MR studies (J Digit Imaging 2007 Nov 13[Epub ahead of print]). "Although JPEG2000 has theoretical advantages for application in virtual microscopy, no efficient application software has until now been described," Isola said.

The JPEG2000 software package Isola developed enables compressing, viewing, and serving large image files produced by modern microscope slide scanners. The system consists of a main JPIP (JPEG2000 Interactive Protocol) server with two network interfaces:

  • a 100-Mb link for client access via Internet or intranet
  • a 1-Gb local area network connection for data traffic between the main server and subservers

The software is available free for noncommercial purposes.

Lossy compression is mandatory in virtual microscopy, due to the large size of the files. JPEG2000 provides excellent image quality preservation compared with compressed image file size, Isola said."We found compression ratios ranging from 25:1 to 30:1 in areas containing cells or tissue sections produced virtual slides without disturbing visible image compression artifacts at the original image resolution," he said.Isola said of two alternative wavelet filters -- reversible and irreversible -- the irreversible filter yielded somewhat better visual quality, but at the expense of compression speed."We estimate that a JPEG2000 compression application using a quadcore processor is capable of compressing over 100 GB of image data per hour," he said. In practice, an 8-GB uncompressed 50,000 x 50,000-pixel virtual slide can be readily compressed into a 300-MB JPEG2000 image. Isola is currently concentrating on linking JPEG2000 file technology with DICOM.