Breast Diffusion Weighted Imaging

Introduction

This lecture explains the basics of diffusion-weighted imaging (DWI) of the breast and describes its role in breast cancer screening, staging, and treatment monitoring. Dynamic contrast-enhanced (DCE) MR is the most sensitive imaging modality for breast cancer, but specificity remains a problem, and patients who undergo annual screening are not fond of an annual dose of contrast. Diffusion may help improve specificity and, in the future, may supplement or partly replace contrast-enhanced MR with a non-contrast technique. Diffusion is not yet sensitive enough to do that on its own, but it is already useful enough to be incorporated into the clinical standard of care. The lecture covers the physics behind diffusion, the apparent diffusion coefficient (ADC), a working 3T protocol, clinical examples, the main challenges and their fixes, and future developments.

Why Perform Breast DWI

Diffusion imaging reflects tumor cellularity and is quantifiable, most commonly through the ADC. It is under assessment in the ACRIN 6698 trial.

Key Points

• Early papers showed DWI can demonstrate response to treatment earlier than contrast-enhanced MR, sometimes within a few days of neoadjuvant chemotherapy, whereas changes in enhancement and tumor size on contrast MR are delayed.
• Diffusion may help characterize breast lesions and improve specificity, which could eliminate some unnecessary biopsies.
• Lymph nodes, normal and malignant, appear like light bulbs on standard DWI, which makes them easy to find and can improve the search pattern for nodes with abnormal morphology.
• Future goals: a non-contrast breast evaluation, an "imaging biopsy" that predicts tumor pathology, and shorter scan time by eliminating contrast and dynamic phases.

MR Basics in Breast Imaging

The patient lies prone in a dedicated breast coil, which matters for diffusion because it gives the best possible signal-to-noise ratio (SNR).

Key Points

• The body is about 70% water; standard MRI detects hydrogen protons from that water, aligned in the main magnetic field.
• Additional gradient fields excite the tissue of interest; protons flipped perpendicular to the main field are detected by the receiving coil, data are collected in k-space, and the inverse Fourier transform yields the image.
• DCE MR detects contrast uptake and washout. Tumors have increased angiogenesis and immature, permeable vessels, so they take up contrast quickly and wash it out.
• DCE MR sees quite small tumors and has been reliable for nearly 20 years, especially for women at high risk for breast cancer.

Basics of Diffusion

Rather than macroscopic permeability, diffusion looks at the microscopic motion of water molecules, including the intracellular component. The average distance traveled by molecules determines the diffusion coefficient, which varies by tissue type.

Key Points

• Cancer generally has higher cellularity, so diffusion is hindered between cells; restricted diffusion is the intracellular component and reflects a high nuclear-to-cytoplasm ratio.
• Hindered and restricted diffusion together give breast cancer cells higher signal than healthy cells.
• Hindered water moves with a Gaussian (root-mean-square) distribution; restricted water, stuck inside a cell, does not follow a Gaussian distribution.
• Three compartments: free water, which disperses wherever it wants; hindered water, stuck between tightly packed cells (seen in cancer, edema, and infection); and restricted water, stuck within cells.
• At higher b-values overall SNR drops, but contrast-to-noise improves because cancer cells remain bright while extracellular water signal falls away.
• The spleen, like lymph nodes, restricts at baseline and remains bright as b-value increases while surrounding tissue loses signal; diffusion is therefore not valuable for separating benign from malignant lymph nodes or spleen masses.

b-Values and the Pulse Sequence

The b-value is a reflection of pathophysiology, not an arbitrary choice: it determines what feature is being examined.

Key Points

• Higher b-values allow water to move for a longer period; snapshots at different diffusion times show where water moved and where it did not.
• Some vendors offer synthetic b-values that increase conspicuity, but they add no information about pathophysiology and do not help differentiate restricted from hindered diffusion.
• Diffusion imaging is usually a spin-echo sequence with bipolar gradients that phase and dephase moving spins, followed by an echo-planar readout.
• The strength of the diffusion weighting (the b-value) depends on those bipolar gradients; signal intensity changes with the b-value chosen, which also affects the measured ADC.
• The echo-planar readout is the source of several artifacts particular to breast diffusion.

The Apparent Diffusion Coefficient (ADC)

ADC quantifies tissue cellularity with imaging, before a needle goes into the patient. Automated systems produce ADC maps once images at different b-values are collected; the reader draws a region of interest and the ADC is calculated.

Key Points

• ADC is expressed in millimeters squared per second; the b-value is expressed in the opposite units, seconds per millimeter squared.
• A low ADC reflects restricted diffusion, so dark areas on the ADC map are more suspicious for cancer. Cancer tends to have lower ADC values than benign lesions.
• ADC values of benign and malignant lesions overlap, in the breast and in every other part of the body, which limits DWI.
• Change in ADC relative to a baseline study in the same patient is valuable for evaluating response to treatment despite that overlap.
• A common higher b-value is b800, paired with a b0 or b50; at least two b-values are needed to fit the curve and quantify ADC.
• Other dark areas appear across the ADC map, so the ROI must be placed using the other sequences, not drawn at random.
• An increased ADC after treatment demonstrates decreased tumor burden; a decreased ADC demonstrates increased tumor burden. Tumor ADC should theoretically be an improved marker relative to tumor size measured by DCE.

Breast Diffusion Protocol (UC San Diego)

All breast imaging at the institution is done at 3T, not by requirement but because the breast clinic building has only 3T scanners, and MR imaging should be done on the same scanner used for MR-guided biopsy. 3T does exacerbate some artifacts, and there is discussion of the value of lower field strengths, but 3T is a perfectly good field strength for breast diffusion.

Key Points

• Spin-echo sequence with echo-planar readout; TR 4.5; flip angle 90; slice thickness 5 mm.
• b-values collected include 600 and 800, in line with the ACRIN 6698 trial.
• The gap between 5 mm slices can cause an entire small lesion to be missed, which is a problem for screening use and an area for future development.
• Field of view matches the DCE acquisition; matrix 128 by 128 to maintain SNR and keep acquisition time down; higher matrix sizes may become feasible as SNR and speed improve.
• Fat saturation must be on. Diffusion without fat saturation is useless.
• Acquire in the same plane as the dynamic acquisition so findings can be correlated.

Clinical Examples

Key Points

• Cysts: the b0 image is essentially a T2 fat-sat sequence, so cysts are bright and cause T2 shine-through, especially at lower b-values; the signal drops at higher b-values.
• Post-biopsy hematoma adjacent to a cancer: standard DWI reflects any high-cellularity process (edema, infection, hematoma), so the T1 non-fat-sat and T2 sequences are used to separate the cancer from the hematoma, which shows a hemosiderin rim and is bright on T1 and T2 without contrast.
• Large breast cancer measuring 5.3 cm on DCE: a large ROI over the lesion gave an ADC of 0.8. As a rule of thumb, an ADC below 1 raises suspicion for cancer, with the caveat of benign-malignant overlap. A commercially available software package was used.
• Response to chemotherapy, example 1: a 12.5 cm right mammary carcinoma with chondroid metaplasia (with incidental diffuse left DCIS) had a pre-treatment ADC of about 1.06; after one course of chemotherapy the mass measured 8.5 cm and the ADC rose to 1.4, demonstrating response both by size and by ADC. ADC values are always times 10 to the minus 3, so check the units.
• Response to chemotherapy, example 2: a 3.5 cm left breast mass with ADC 1.4 shrank to 2.1 cm after treatment, but the ADC fell to 1.3. The size decrease suggests response, but the ADC decrease is concerning for progression: chemotherapy can kill vessels so vascularity is no longer detected, while diffusion reflects the cellularity that remains.
• Response to chemotherapy, example 3: ADC increased from 1 to 1.95, concordant with the kinetic and morphologic response on DCE, demonstrating a significant favorable response. The initial cancer ADC was 1, again overlapping with benign lesions.
• Post-surgical fat necrosis after implant removal: rim enhancement and diffusion signal in the lower left breast; the T1 non-fat-sat sequence showed florid fat in the area, making the diagnosis of benign fat necrosis. Use every sequence acquired.

Challenges of Breast Diffusion

Key Points

• Distortion from B0 inhomogeneity is the main challenge. Unlike the brain, which sits round in the center of a homogeneous field, the breasts are oblong, off-center, bordered by a large air-tissue interface, with the beating heart behind them. In most current standard-of-care acquisitions this is not corrected.
• Edema and hematoma can look bright on diffusion. Very high b-values (3,000 to 4,000) are less susceptible to the edema contribution.
• The ideal b-value is still open; multiple b-values are being evaluated and the cost-benefit of higher b-values versus acquisition time will be debated for years.
• There is no standard way to normalize ADC across scanners, patients, field strengths, and vendors; this needs development and vendor cooperation.
• Failed fat saturation, as in a patient with excess axillary fat and an unusual body habitus, leaves fat signal at both b-values and makes tumor detection and the ADC calculation unreliable.
• Sharp fat-glandular interfaces in the breast produce chemical shift artifacts, and 3T exacerbates them.
• Early issues encountered when first using diffusion: low SNR at different b-values, inhomogeneous fat saturation (making shimming extremely important), and aliasing.

Correcting and Minimizing Distortion

Using a breast phantom from the National Institute of Standards and Technology across multiple scanners of the same vendor and at different field strengths, the speaker's group showed uncorrected B0 distortion leaves the breasts angled and distorted ("flowing in the wind"), which degrades precision in identifying a tumor.

Key Points

• Reverse polarity gradient (RPG) technique: with single-shot echo-planar imaging, acquiring the same k-space data in the opposite phase-encode direction allows the distortions to be calculated and corrected. Some vendors now carry this.
• Parallel imaging by itself minimizes distortion.
• Reduced field of view improves distortion. It cuts out part of the axilla, but as a complementary sequence this is reasonable: diffusion shows lymph nodes but cannot separate benign from malignant nodes, because all nodes are bright at high b-value.
• Across four scanners of the same vendor, diffusion signal varied markedly; normalizing with a small tube of water in the field of view reduced that variability.

Future Developments

Key Points

• Standardize diffusion across magnet manufacturers and within the same manufacturer.
• Identify the "ideal" b-value, a moving target that depends on what is being sought and how much time is available.
• Correct for artifacts, including distortion and fat saturation: use parallel imaging, consider reduced field of view, and use RPG or similar if available.
• Very high b-values (up to 4,000) with distortion correction are under investigation to differentiate intracellular restricted diffusion from hindered diffusion. In a lactating patient whose cancer was hard to see against background enhancement, b4,000 improved conspicuity of the lesion.
• Separating signal into free, hindered, and restricted components may classify tissue and move toward the imaging biopsy.
• Every part of the body is different; meaningful results require accounting for the tissue type being evaluated. Use the basic technique now, and watch for the up-and-coming technology.

Knowledge Check

Key Points

• An increase in ADC in breast cancer represents a quantitative measure reflecting decreased tumor burden.
• ADC units: millimeters squared per second. b-value units: seconds per millimeter squared.
• A higher b-value reduces the contribution of perfusion effects on the ADC; perfusion contributes most at b200 and below. Some techniques use low b-values to estimate perfusion without contrast, but SNR is poor and contrast is a direct measure of perfusion.
• In women with normal menstrual cycles, the best time for contrast-enhanced imaging is week two, although breast MR is now performed whenever needed because the cancer can still be seen and care should not be delayed; knowing the last menstrual period helps interpret background enhancement.
• ADC in normal breast tissue decreases during week two, related to hindered water in the breast.

Summary

Key Points

• Benign lesions: intermediate to low diffusion signal, increased ADC, reflecting lower cellularity.
• Malignant lesions: higher signal on raw diffusion images, decreased ADC, reflecting high cellularity.
• The technique is not helpful for DCIS because of its low cellularity.
• A favorable response to chemotherapy is shown by an increased ADC across two time points.
• Lesion cellularity affects T2 signal as well as DWI and ADC; ADC is inversely correlated with cellularity; benign and malignant values overlap, which limits use as a screening technique for now.
• ADC values are highly dependent on the magnet manufacturer and the b-values used.

Questions and Answers

Key Points

• Cost and framework for adopting breast DWI now: the main cost is about five minutes of scan time. Any scanner that does brain diffusion already has the sequence; most sites, including outside institutions, are starting to include diffusion in their acquisition.
• Colorization in the images: standard ADC maps are grayscale. Red, blue, and yellow are the DCE conventions for type one, two, and three curves; in the lecture the color simply delineated the ROI drawn from the ADC map and DCE together.
• Delta to the one-half power represents the square root: diffusion time is inversely related to the root mean square of the distance.
• What "b" stands for in b-value: the speaker does not know; every source simply says b-value.