Dr Laurence Sutton Calderdale and Huddersfield NHS Foundation Trust Halifax West Yorkshire

1
Dr Laurence SuttonCalderdale and
HuddersfieldNHS Foundation TrustHalifax West
Yorkshire

• Is Lossy Compression of Image Data detrimental to
  Image Interpretation?
• Will Lossy Compression
• Become standardised Practice?

2
Image Compression

• Definition.
• Do we need to compress image data?
• Compression Methods.
• Canada Health Infoway Project.
• Technical and Clinical Evaluation.
• Medico-Legal Aspects (Regulations).
• Next Steps?

3
Definition

• Image Compression is used to reduce the amount of
  Memory required to store the Image data without
  degrading the (visual) quality of the image to an
  unacceptable level

4
Image Compression

• Definition.
• Do we need to compress image data?
• Compression Methods.
• Canada Health Infoway Project.
• Technical and Clinical Evaluation.
• Medico-Legal Aspects (Regulations).
• Next Steps?

5
Image Data Explosion fuelled by clinical
advancements
18800
Number Of Slices (Millions)

• 20 million CT slices produced by one imaging
  department
• 10 Tera Bytes of data in 1 year
• Speed, Storage, and Interpretation a HUGE
  challenge

10300
6300
3600
2700
2000
1900
1100
444
8 Slice
16 Slice
Single Slice
4 Slice
VCT 64 now a reality Technology not standing still
160 slices
40 slices
640 slices
80 slices
6
Open for new vocabulary?
(exact 1024 Byte)
new vocabulary ? 1 Googol 10100
7
Storage Management
of Retrievals
10
Digital
Film
1 Year
Time (Years)
8
TCO total cost of ownership

• Management cost increases
• Data availability
• People
• Software
• Security
• Facilities
• Energy
• Others

Storage TCO increases
9
Image Compression

• Definition.
• Do we need to compress image data?
• Compression Methods.
• Canada Health Infoway Project.
• Technical and Clinical Evaluation.
• Medico-Legal Aspects (Regulations).
• Next Steps?

10
Image CompressionLossless or Lossy

• Lossless
• Regenerated is guaranteed to be identical to the
  original image
• Also known as reversible compression
• Lossy
• A form of image compression where the regenerated
  image is not guaranteed to be identical to the
  original image
• Also known as irreversible compression

11
(No Transcript)
12
Historical

• SCAR Paper 2000,Bradley J. Erickson
• Increasing evidence suggests that some forms
  of irreversible compression can be used with no
  measurable degradation in aesthetic or diagnostic
  value.
• Today however Primary Reading of Irreversibly
  compressed images is not an accepted standard of
  practice

13
Do we need to compress Image Data?

• Compression is a form of information processing.
• The latter we do all the time and throw away
  data!
• Selected fluoroscopy Images
• Selected CT slices sent to PACS
• Representative US Images
• System and Technique information loss (In Plane
  Loss)

14
In Plane Loss

• In Plane Loss
• Not saving CT raw data
• Low dose techniques (quantum noise)
• Applying and saving only one CT reconstruction
  algorithm
• Cleanup and interpolation in CR/DR
• Any form of image processing (information
  theory!)
• 8 or 10 bit display/windowing of 12-16 bit data
• Displaying any part of an image at less than 11
  zoom
• Non-linear, badly calibrated monitors
• Archiving on Film
• (lossy compression)

15
Some types of Image Compression

• Joint Photographic Experts Group JPEG (ISO/IEC).
  (OSI)
• Graphic Interchange Format GIF (LZW)
• Potable Network Graphics PNG
• WWW. Image File Formats
• GIF and PNG are lossless
• JPEG can be lossless

16
Some types of Image Compression

• GIF and PNG are lossless
• LZW (PK-zip, built-in in RAID etc.)
• lt 21
• JPEG can be lossless
• JPEG Lossless (Pulse Code Modulation compression)
• Totally unrelated to normal JPEG
• 2 varieties defined in DICOM
• About 2.51
• JPEG 2000 lossless mode
• About 31

17
JPEG2000 and JPEG

• JPEG, (Lossy) Discrete Cosine Transform Technique
  resulting in blocky artefacts at higher level of
  compression.
• JPEG2000, Wavelet based compression technique
  which allows for higher compression ratios
  without overt artefacts.
• Part of JPEG2000 is now part of an ISO standard
• File ex, .jp2 or .j2c

18
JPEG2000 Advantages

• Superior Compression Performance.
• Lossless and Lossy Compression.
• Error Resilience.
• Ideally suited to photographic type images.
• Smoothing Artefacts at high compression levels
• Regenerated Images are less Noisy and cleaner!
• JPIP compression streamlining protocol.

19
Explicit Lossy Processes
20
JPIP

• Uses the minimum bandwidth required.
• Ability to only download the requested part of
  the image.
• Quick view of a large image at low resolution.
• Ability to pull more data for higher resolution
  images of smaller parts of the image.
• Reduces computer processes at server and client
  ends.

21
Image Compression

• Definition.
• Do we need to compress image data?
• Compression Methods.
• Canada Health Infoway Project.
• Technical and Clinical Evaluation.
• Medico-Legal Aspects (Regulations).
• Next Steps?

22
Canada Infoway Project

• National EHR
• Storage and Network Impact of 35M images (1.5PB)
  annually.
• Cost of storage versus cost of ownership.
• Lossy Compression could save CDN100M/year.
• Can Lossy (Irreversible) Compression be used as a
  standard of practice?
• Commissioned series of projects to evaluate

23
Canada Infoway Project

• Research conducted so far and the evaluation of
  the effects on Image Quality..2 Independent
  Reviews.
• Legal Implications of using Irreversible
  compression..2 Independent Reviews.
• Current Regulations around the world..1 review
• The outcome of the reviews is such that the CAR
  is motivated to adopt Irreversible Compression as
  a standard and to such an end .

24
Canada Infoway Project

• Clinical Review in conjunction with Infoway to
  assess the most appropriate compression ratios
  for a variety of different body parts for each
  modality.
• Canadian Association of Radiologist endorsement
  of the resulting Guidelines.
• Results SIIM (SCAR) June 2007 Rhode Island.

25
Outcomes.

• Review of research.
• Both reviewers concluded that based on
• scientific studies, irreversible compression
• is a clinically acceptable option for the
• compression of medical images.
• Legal Review.
• Presume no visual impact on the quality of the
  Image.
• Main risk is Tort of Negligence.
• Reduce risk by universal adoption of the
  standard.
• Respectable minority of Radiologists

26
Outcomes.

• Legal Review, BUT
• Can not alter the Medical Record once it has been
  approved.
• Requirement to compress the image prior to the
  Primary Read.
• BUT, that is ok isn't it?
• In this country the report is still the legal
  document.
• Regulatory Review.
• Non of the regulatory bodies in
  Canada,USA,UK,EU,Australia have prevented or
  endorsed the use of imaging compression

27
Technical and Clinical Review.

• A Review and Comparison of Medical Image
  Compression Algorithms
• August 2004
• Kirk W. Finnis, PhD
• Atamai Inc., 100 Perth Drive, Suite 500,
• London, Ontario, Canada

28
Technical and Clinical Review.

• Research Review
• And Assessment of Compression performance.
• Comparisons of different algorithms between
• Medical Images and Standard Image sets.
• Lossy Compression of Medical Images.

29
Technical and Clinical Review.

• Effects of Lossy Compression on Diagnosis.
• Usually centred on one modality and a Pathology.
• Dental
• CT and MR
• CXR
• Digital Angiograms
• Mammograms
• Ultrasound

30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
CXR

• Slone, Muka Pilgram,
• http//radiology.rsnajnls.org

34
Mammography

• ECR 2007, German Study, Dr R Loose, Nuremberg
• 33MB down to 325KB using standard wavelet
  compression.
• No statistically significant impact on quality or
  visible differences in images for compression
  rates up to 1001
• Significant Impact on storage costs and the
  provision of Teleradiology services

35
CAD CT Lung Nodule Volume Assessment.

• Jane B Ko et al, Radiology 200523783-88
• Low dose CT technique, ROI
• No significant difference 101
• Difference observed at 201
• Reduced volume measurements.
• Need to be aware of the issues in Computer Aided
  Techniques

36
Summary
Metrics used by researchers to assess compression
algorithms included Quantitative measures - a
measure of image quality Mean Squared Error
(MSE) Peak Signal to Noise Ratio (PSNR)
Qualitative measures - a measure of diagnostic
quality Difference Image and Histogram
Receiver Operating Characteristics (ROC) analysis
37
Conclusions and Caveats

• Limits (Tolerance) Image Compression
• MR/CT 101
• CR 201
• JPEG2000 adopted as a DICOM Standard
• Rapid Adoption by the Imaging Community.
• Medico-Legal uncertainties exist.
• Evidence supports the safe adoption of Lossy
  compression for storage and compression without
  compromising diagnostic quality.

38
Next Steps.

• ? Case for adoption of Lossy Compression in the
  UK
• Royal College Approval of process to validate the
  case for Lossy Compression.
• Adopt Lossy Compression as a safe standard for
  the purpose of diagnosing from medical images.
• Consider potential issues around computer aided
  measurements.