Angiogenesis in solid tumors gives rise to a relatively high fraction of immature blood vessels, which represents an attractive therapeutic target. Non-invasive methods that probe vascular function would have potential in characterizing individual tumors, predicting treatment susceptibility, and monitoring response. Blood oxygenation level-dependent (BOLD) MRI in conjunction with hyperoxic/hypercapnic gas stimuli has been proposed as a functional biomarker of vessel maturity^1–3. Immature vessels lacking appropriate smooth muscle vasculature should be unable to dilate or constrict in response to vasoactive challenges, whereas surrounding healthy mature vessels should respond³. However, previous optical imaging studies have suggested that these measurements may be confounded by the body’s low frequency hemodynamic fluctuations⁴. In this study we evaluated BOLD response to carbogen interleaved with air and oxygen, relative to an air-only control experiment in healthy breast parenchyma. We also tested the relative efficacy of carbogen (5% CO₂, 95% O₂) and ‘carbogen-light’ (2% CO₂, 98% O₂) in inducing a BOLD vasomotor response. The optimal stimulus design was evaluated in a small patient cohort.

Data Acquisition: MR imaging was performed at 3T (MR750, GE Healthcare, Waukesha, WI) in 19 healthy premenopausal volunteers (aged 22–39 years) using a multi-phase single-shot fast spin echo sequence to acquire sequential T₂-weighted sagittal breast images at a single slice location. Imaging parameters were as follows: TR 4s, TE 58ms, bandwidth ±83kHz, matrix size 128x128, FOV 20cm, slice thickness 5mm. The modulated gas stimulus design consisted of 2% (n=9) and 5% (n=10) carbogen gas mixtures interleaved with either medical air or oxygen in two-minute blocks, for four cycles (Figure 1). A twelve-minute medical air breathing period was used to determine background physiological noise. Gases were administered to the subject via an OxyMask™ (Southmedic Inc., Barrie, ON) at a flow rate of 14L/min, with switching controlled by an in-house gas delivery system. The number of days from participants’ last menstrual period at the time of the MR examination was noted. Five patients (aged 29–59 years) with histologically proven breast cancer were scanned using the same BOLD MRI protocol, but only the optimal stimulus design was delivered. Tumor ROIs were delineated by a radiologist on slice-matched post-contrast images. Data Analysis: A non-rigid registration algorithm⁵ was employed to mitigate respiratory motion artifacts. The first cycle of data was discarded to allow equilibration of the gas inhalation regime. Baseline subtraction of the line of best fit through the data was performed to eliminate linear drift. Signal intensity response for each pixel meeting a certain signal-to-noise threshold was then cross-correlated with a sine and cosine function at the stimulus frequency (0.0042 Hz), yielding pixel-wise measures of the correlation and phase lag between the delivered stimuli and measured BOLD contrast. The same frequency was used for the baseline air-only data, even though there was no imposed periodicity in the stimulus. Paired (across subjects) t-tests were performed to compare response to the alternating stimuli and the air-only state. The ratio of median correlation coefficients was calculated for the oxygen/carbogen stimulus relative to each subject’s all-air control. This metric was compared between two groups of volunteers scanned in different phases of the menstrual cycle to explore the effect of hormonal changes on BOLD response.

Activation maps and histograms of pixel-wise correlation coefficients and phase lags for the air/carbogen and oxygen/carbogen stimuli, compared to air-only breathing are shown in Figure 2. Of the four stimulus combinations tested, oxygen/carbogen produced a response that was significantly stronger (p<0.05) than the air-only control in volunteers (Table 1). Subjects imaged during the follicular phase of their cycle (~days 10-20), when estrogen levels typically peak⁶, exhibited a significantly smaller BOLD response (p=0.011), compared to those in the menstrual or luteal phases of their cycle (Figure 3). Response to the oxygen/carbogen stimulus in malignant tissue was variable, with two lesions exhibiting a significant response to the gas stimulus, whilst three showed a diminished response relative to surrounding parenchyma (Figure 4).A measurable BOLD effect was observed in healthy breast parenchyma in response to oxygen vs. 5% carbogen, above the demonstrated level of normal physiological fluctuations seen with air. This is consistent with the opposing vasomotor effects of these two gases. Our results suggest changes should be characterized with respect to normal air variations to help account for inter-subject variations in background physiological noise. Additionally, higher estrogen levels appear to suppress vasomotor response. The diversity of response seen in tumors suggests this technique may be sensitive to the underlying vasculature, which warrants further investigation to correlate BOLD contrast changes with histological markers of vessel maturity and with therapy response.