TwinTree Insert

16-04 Clinical Examples

he following paragraphs present the fundamentals and the development of the techniques of three clinical examples. Lesion characterization often re­quires a dynamic contrast series.

However, procedures and indications continue to be a subject for debate.

Contrast-enhanced studies may show pathologies better than plain images or as­sist in diagnosis. Few will be truly decisive. But they facilitate treatment and men­tal comfort for patients and radiologists alike. The only clear and un­dis­put­ed in­di­ca­tion is breast MRI. Without a contrast agent, this examination is useless. With con­trast en­han­ce­ment, it is the best mammography technique we have.

16-04-01 Breast Imaging

As described by Harms [⇒ Harms 2001], the use of MRI in the role of breast cancer staging has the following minimum requirements:

spaceholder darkblue(1) It must have high-resolution (approximately 1 mm resolution in all three di­rec­tions) in order to consistently detect cancers missed by mam­mo­gra­phy;

spaceholder darkblue(2) it should employ fat suppression in order to separate enhancing tumors from fat; and

spaceholder darkblue(3) it should be a rapid acquisition (preferably about 5 min) in order to se­pa­ra­te enhancing tumors from ductal tissue that has been shown to have de­lay­ed en­han­ce­ment. Dynamic imaging reveals that tumors and ductal tis­sue in­ten­si­ties merge and become isointense between 5 and 10 min post-injection.

Dynamic imaging of the breast became the first major application of dynamic MR imaging with gradient-echo pulse sequences being used. The combination of rapid imag­ing and contrast agent application increased both sensitivity and spe­ci­fi­ci­ty of breast MR imag­ing (MRM = magnetic resonance mammography) and allowed the dif­fe­ren­ti­a­tion between benign and malignant lesions.

Signal intensity-versus-time uptake curves showed that malignant lesions ta­ke up con­trast agent faster than benign lesions, although there remained a cer­tain over­lap. Since these measurements were done manually, it was difficult to find the pixel or region of highest uptake of contrast agent within the breast, in particular on gray-scale images. In these cases, postprocessing became very va­lu­able.

Originally, subtraction images were used; however, this approach highlights all pi­xels with contrast enhancement without any differentiation between fast and slow en­han­ce­ment over time.

Then, mathematical approaches to en­han­ce­ment curves were introduced to cre­ate parametric images, calculating a pa­ra­me­ter value for each pixel in a slice and plot­ting the values as an image, typically a gray-scale image with the in­ten­si­ties pro­por­ti­o­nal to the pa­ra­me­ter value. Pixel-by-pixel calculation of en­han­ce­ment in­ten­si­ty and speed or slope led to pa­ra­met­ric images which can be color-coded in a way that regions of fast and high en­han­ce­ment are highlighted in a specific color.

For instance, enhancement of more than 90% in less then 90 seconds on T1- weight­ed images, or signal intensity loss of > 20% during the first 30 seconds after con­trast ma­te­rial injection on T2*-weighted images, is considered typical for ma­lig­nant breast lesions, although not all malignant lesions follow this pat­tern (Figure 16-08) [⇒ Boetes 1994, ⇒ Flickinger 1993, ⇒ Gribbestad 1994, ⇒ Heywang-Kö­brun­ner 1995, ⇒ Kaiser 1990, ⇒ Kvistad 2000, ⇒ Torheim 1997].

Figure 16-08:
Magnetic resonance mammography (MRM). (a) Image from a data set of 44 dynamic slices. T1-weighted RF-spoiled gradient-echo sequence. The ROI is positioned in the tumor.
(b) Same patient: Parametric map based on the T1-weighted image time series: ma­xi­mum en­­han­­ce­­ment image.
(c) Dynamic uptake curve of an ECF con­trast agent in the breast lesion depicted in (a) and (b).

Processing of dynamic imaging with color coding can also visualize the en­han­ce­ment pat­tern over time. When large enough, fibroadenomas usually de­mon­stra­te initial peak enhancement in the center of the tumor, whereas carcinomas tend to enhance in the periphery. However, since the enhancement pattern de­pends on the vas­cu­la­ri­ty of the lesion, no direct histological tumor-typing by dy­na­mic MRI is possible.

Figure 16-09 shows the dynamic uptake patterns of a number of breast le­sions. The curves are created by the averaged intensities in regions-of-interest in frames. A frame is an image series along the time axis (time series).

Figure 16-09:
Dynamic uptake pattern of a Gd-based ECF-space agent in breast le­sions. Enhancement of more than 90% in less than 90s (reddish area) after bolus injection occurs most likely in invasive duc­tal car­ci­no­mas only. Thus, such tumors can be identified in parametric images, where all pixels with en­han­ce­ment >90% at time ≤90 s can be color-coded. The carcinoma is this picture appears in bright red.

spaceholder redMeanwhile, a shortened breast MRI protocol comprising an abridged dynamic se­ries of one pre- and one post-contrast acquisition has been proposed. The re­sult­ing sub­trac­ted images are processed by maximum intensity projection (MIP).

spaceholder redThere are a large number of partly contradictory publications on dynamic MR imag­ing of the breast. Besides the above mentioned detailed overview by Harms an­other excellent review was published by Kuhl [⇒ Kuhl 2015].

16-04-02 Brain Imaging

Dynamic (or perfusion) imaging of the brain must not be confused with func­tio­nal imag­ing of the brain (fMRI) although similar pulse sequences and parallel imaging tech­ni­ques are applied for these examinations [⇒ Petrella 2000, ⇒ Torheim 1997].

Echo planar techniques and gra­dient echo three-dimensional magnetic resonance imag­ing techniques, e.g., PRESTO (principles of echo shifting with a train of ob­ser­va­tions) [⇒ van Gelderen 2012] are applied for for bolus tracking. PRESTO allowes for higher temporal resolution and has less susceptibility artifacts. Regional cerebral blood volume (rCBV) can be estimated by fitting first-pass transit curves to the pixel in­ten­si­ties of a series of images.

spaceholder redPatients with acute stroke make up the group of most interest for dynamic imag­ing of the brain [⇒ Orrison 1995]. In clinical practice, perfusion imaging has pro­ven to be an early and reliable predictor of prognosis in stroke patients. A num­ber of researchers have found that the area of perfusion deficit seen in cerebral blood flow (CBF) and mean transit time (MTT) maps may extend beyond the area of hy­per­in­ten­si­ty seen in diffusion-weighted imaging and that the size of the infarct fi­nal­ly seen on delayed T2-weighted images matches the area of perfusion defect (Fi­gu­re 16-10) [⇒ Østergaard 1998, ⇒ Rosen 1989, 1991].

Figure 16-10:
Parametric images of the brain of a patient with recent stroke.
(a) Area under the curve image which is correlated to blood volume. The image was created by first con­vert­ing the time-intensity curves into time-concentration curves by using mathe­ma­ti­cal pro­ces­sing. Non-perfused areas have a flat curve, thus the areas are small and thus the non-perfused regions show as dark.
(b) A ROI has been drawn to indicate an ischemic region.
(c) Time-to-peak image of the same slice. The perfusion in the ischemic region is delayed; it shows up in light gray.

Tracer kinetics principles first employed in nuclear medicine can be applied to ge­ne­ra­te cerebral blood volume maps [⇒ Belliveau 1990, ⇒ Tofts 1999].

Ultimately, the goal of perfusion imaging remains the visualization of the pen­um­bra and thus the distinction between normal, salvageable, and irreversibly damaged tissue.

16-04-03 Heart Imaging

The main goal of myocardial perfusion imaging is the detection and delineation of hypo­per­fu­sion due to non-occlusive coronary artery stenosis. Screening for ischemic heart disease requires both high spatial and temporal resolution ima­ges with de­tec­tion and quantification of abnormal wall motion, evaluation of car­di­ac me­ta­bo­lism, and measurement of regional myocardial perfusion [⇒ Atkinson 1990, ⇒ Lombardi 1995; 2018].

In the heart, the much higher blood volume and the abundance of sus­cep­ti­bi­li­ty ar­ti­facts in plain imaging commend assessment of perfusion by T1-weighted dy­na­mic imaging during the first pass of an appropriately low dose of a pa­­ra­­mag­­ne­­tic con­trast agent.

Assessment of myocardial blood flow is difficult because a large fraction of ex­tra­cel­lu­lar contrast agents will extravasate into myocardial tissue during the first pass, making myocardial signal intensity dependent on both ex­tra­ction frac­tion and flow. However, some research groups have succeeded in obtaining ex­cel­lent de­li­ne­ation of hypoperfused areas using first-pass dynamic imaging with ECF-space contrast agents under stress, usually pharmacological stress [⇒ Higgins 1996]. Based upon de­fin­ed ROIs, parametric images of the heart can be ob­tain­ed by combining ana­to­mi­cal and functional information (Figure 16-11).

Figure 16-11:
In this case, the heart is semi-automatically divided into ROIs which follow the supply area of the coronary arteries. The pa­ra­met­ric image represents the cross-cor­re­la­tion coefficient (CCC) calculated one week after co­ro­na­ry in­farc­tion (dark area). See also Chapter 14.

16-04-04 Other Applications and Critical Remarks

Other Applications. There are numerous other applications of dynamic imag­ing, in­clud­ing imaging of the liver, the kidneys, muscles and joints, the urinary bladder, and the prostate.

Critical Remarks. Image processing is useless if applied ran­dom­ly with­out a well de­fin­ed aim. Many approaches to explain results of dynamic imaging and image pro­ces­sing are based on hypotheses which are still to be proved, and much research in this field is empirical and heuristic.