On the feasibility of MRI‐guided navigation to demarcate breast cancer for breast‐conserving surgery

Abstract

The aim of this study was to investigate the feasibility of image-guided navigation approaches to demarcate breast cancer on the basis of preacquired magnetic resonance (MR) imaging in supine patient orientation. Strategies were examined to minimize the uncertainty in the instrument-tip position, based on the hypothesis that the release of instrument pressure returns the breast tissue to its predeformed state. For this purpose, four sources of uncertainty were taken into account: (1) U(ligaments): Uncertainty in the reproducibility of the internal mammary gland geometry during repeat patient setup in supine orientation; (2) U(r_breathing): Residual uncertainty in registration of the breast after compensation for breathing motion using an external marker; (3) U(reconstruction): Uncertainty in the reconstructed location of the tip of the needle using an optical image-navigation system (phantom experiments, n = 50); and (4) U(deformation): Uncertainty in displacement of breast tumors due to needle-induced tissue deformations (patients, n = 21). A Monte Carlo study was performed to establish the 95% confidence interval (CI) of the combined uncertainties. This region of uncertainty was subsequently visualized around the reconstructed needle tip as an additional navigational aid in the preacquired MR images. Validation of the system was performed in five healthy volunteers (localization of skin markers only) and in two patients. In the patients, the navigation system was used to monitor ultrasound-guided radioactive seed localization of breast cancer. Nearest distances between the needle tip and the tumor boundary in the ultrasound images were compared to those in the concurrently reconstructed MR images. Both U(reconstruction) and U(deformation) were normally distributed with 0.1 +/- 1.2 mm (mean +/- 1 SD) and 0.1 +/- 0.8 mm, respectively. Taking prior estimates for U(ligaments) (0.0 +/- 1.5 mm) and U(r_breathing) (-0.1 +/- 0.6 mm) into account, the combined impact resulted in 3.9 mm uncertainty in the position of the needle tip (95% CI) after release of pressure. The volunteer study showed a targeting accuracy comparable to that in the phantom experiments: 2.9 +/- 1.3 versus 2.7 +/- 1.1 mm, respectively. In the patient feasibility study, the deviations were within the 3.9 mm CI. Image-guided navigation to demarcate breast cancer on the basis of preacquired MR images in supine orientation appears feasible if patient breathing is tracked during the navigation procedure, positional uncertainty is visualized and pressure on the localization instrument is released prior to verification of its position.