Model Based Iterative Reconstructions represent a paradigm shift - Imaging with almost no noise

Transcription

1 Model Based Iterative Reconstructions represent a paradigm shift - Imaging with almost no noise Jonas Rydberg, M.D. Professor of Radiology Indiana University School of Medicine Indianapolis, Indiana Medical Director Radiology IU Health Methodist Hospital

2 Disclaimer: IU Health Methodist Hospital collaborates with Philips Healthcare on CT imaging. idose Iterative Reconstructions (IR) IMR Model Based Iterative Reconstructions (MBIR) The experience shared in this presentation is based on work with idose (IR) and IMR (MBIR) both Philips Healthcare. But, only the terms IR and MBIR will be used to name the two techniques.

3 Conclusions from body imaging and CTA: Iterative reconstructions (IR) - Radiation dose reduction 50% or more Model Based Iterative Reconstructions (MBIR) - More radiation dose reduction - Improve the diagnostic image quality

4 Iterative reconstructions (IR): Same patient scanned using IR on different occasions with different reference mas. Significant dose reductions possible. 300 mas 98 mas 71 mas CTDIvol (mgy) 17.5 CTDIvol (mgy) 5.74 CTDIvol (mgy) 4.81

5 Radiation dose reduction Abdomen-Pelvis with IR FBP: 120 kvp Ref mas mgy IR: 120 kvp Ref mas mgy Radiation dose reduction = 40 % (More recently we have moved the Reference mas to 154 = 50% radiation dose reduction.)

6 Radiation dose reduction Abd/Pel with MBIR FBP: 120 kvp Ref mas mgy IR: 120 kvp Ref mas mgy MBIR: 100 kvp Ref mas mgy Radiation dose reduction = 66 %

7 There are no extreme cases of radiation dose reduction presented here. All cases in this presentation were scanned with our standard IR (idose level 4) kvp = 120 kvp Reference mas = 173 or higher The purpose was to explore if MBIR could add diagnostic quality instead of just reducing the radiation dose.

8 MBIR A challenge for radiologists FBP / IR MBIR Hard to accept this look

9 Even tougher challenge with non-enhanced scans FBP / IR MBIR Radiologist reactions: Bland Not right Water colors Waxy Monet effect Texture missing

10 Can you handle the truth? FBP / IR MBIR The difference is reduced noise!

11 Texture?! Is there such a thing as texture in CT images?

12 Texture Random page from Radiographics 2013 Please, take a few seconds to review the texture of mets and liver parenchyma.

13 Texture Random page from Radiographics 2013 For the past 30 years I have been convinced that I have viewed the texture in both liver parenchyma and metastasis. Am I wrong?

14 Texture of liver metastasis Filtered back projection Iterative reconstructions Model Based Iterative reconstructions FBP IR MBIR Somewhat strangely the texture decreases with MBIR. Likewise strangely the texture of the metastasis and the liver parenchyma seems to be the same on both FBP and IR.

15 Does fluid in urinary bladder have texture? Which image is closest to the truth? IR MBIR

16 What shall fluid in the stomach look like? Like IR or MBIR? The same question then regarding liver! IR MBIR

17 What shall kidneys look like? IR MBIR Maybe MBIR depiction is closest to truth?

18 How about the air around the patient? FBP IR MBIR

19 How about the air around the patient? FBP IR MBIR

20 Texture? There is no such thing as texture as we have always believed. - It is just noise to varying degrees!

21 Texture of a renal cancer IR MBIR Cancer + noise There are multiple black and white noise pixels projecting over the tumor that we take for texture of the mass. Cancer CT

22 Black noise and White noise Reducing image noise has been one of the major goals for the CT development during the past 35 years. We are so used of seeing noise that we barely are aware of it. It has gone so far that the noise reduction with MBIR can creates confusion radiologists. Noise destroys the diagnostic quality of the images. To better understand the negative effect of noise in FBP and IR I think in terms of Black noise and White noise. That will be discussed in the following cases. Please, note that the black and white noise is an expression that I have adopted based on my own observations. You may see it expressed in other terms in the literature. The noise is typically quantified and expressed in standard deviations (SD).

23 MBIR improves diagnostic quality. The following examples intend to show how MBIR improves the diagnostic quality over FBP and IR.

24 8 mm cyst in liver Depiction superior with MBIR IR MBIR 8 mm Unenhanced scan with slice thickness = 4 mm

25 IR - White noise degrades depiction of both the center and the border of the cyst IR MBIR Compare Avg HU and SD

26 If slice thickness is decreased to 1 mm noise hurts the IR technique IR MBIR MBIR is noise resilient even at 1 mm slice thickness

27 4 mm slice thickness depicts cyst better than 1 mm but there is a price to pay with increased partial volume averaging. IR MBIR

28 FBP IR MBIR

29 Which technique depicts cysts the truest? FBP IR MBIR

30 FBP IR MBIR

31 FBP IR MBIR

32 FBP IR MBIR

33 FBP IR MBIR

34 IR MBIR

35 IR MBIR

36 IR MBIR 4 mm 1 mm 4 mm 1 mm

37 Between yellow arrows clear border delineating higher and lower density tissues.

38

39

40