New opportunities in image- guided surgical oncology

Complete surgical resection of neoplastic tissue is critical to improving cancer outcomes. Although surgical oncologists attempt to remove as much of the tumor as is safely possible, remaining tumor cells can lead to tumor recurrence; conversely, overly aggressive removal of healthy tissue may result in poor cosmetic outcome and/or excess morbidity in patients with breast, brain, skin, and head-and-neck cancers. Thus, there is a critical need to distinguish tumor from normal tissue.

Although fluorescence image guidance has recently shown promise in identifying microscopic disease during surgery,^1–3 few optical dyes are available for clinical use and no target-specific agents have been approved by the US Food and Drug Administration. Alternatively, a wide range of disease-specific positron emission tomography (PET) radiotracers are employed routinely in clinical diagnostic imaging,⁴ and development of new PET probes is incentivized by immediate commercial application. A new method to image PET radiotracers, termed Cerenkov luminescence imaging (CLI), has emerged since the discovery that modern optical imaging research systems can detect the Cerenkov photons emitted by highly-energetic charged-particle (β⁺ or β⁻) emissions.^5–9 The underlying physics is shown in Figure 1. The ability to perform CLI with conventional optics provides an opportunity to use Cerenkov light to guide intervention and improve surgical resection margins.

Figure 1. Schematic of the physics of Cerenkov luminescence imaging (CLI). Optical photons are emitted when charged particles emitted from a radiotracer exceed the relativistic speed of light in the medium (in this case, tissue). An optical probe, such as a flexible fiber bundle, would be a practical clinical tool to enable CLI-guided surgical resection.

Our lab has recently proposed a new oncological surgery paradigm using intraoperative CLI.¹⁰ Patients are injected with a radiotracer prior to surgery, and a PET image is used to identify tumor location for surgical planning and to determine whether the selected PET tracer is appropriate for CLI. Surgical resection follows as in standard practice. In regions where the surgeon believes the tissue is free from tumor, the CLI fiberscope is introduced to locate regions of cancerous tissue. After this tissue is removed, the Cerenkov fiberscope is used ex vivo to verify that the appropriate tissue was excised. We have demonstrated this new surgical device for two clinical uses: bulk tumor resection, demonstrated in a subcutaneous xenograft mouse model, and resection of tumor invasion, performed in a transgenic preclinical murine model.

To demonstrate the performance of CLI for bulk tumor resection, five mice had subcutaneous tumors grown from a C6 glioma cell injection. After intravenous injection of fluorodeoxyglucose (¹⁸F-FDG) via the tail vein, we performed CLI imaging in a dark chamber through a coherent optical fiber imaging bundle coupled to a CCD. The distal end of the optical fiber was coupled to an 8mm F/1.4 lens with a 1cm working distance (Schneider Cinegon). We surgically exposed the tumor tissues and CLI was performed on each mouse before and after surgical removal of the tumor using the fiber-based imaging system (see Figure 2). Tumor tissues showed significant preferential uptake of ¹⁸F-FDG compared with normal tissues, with nearly a 20% increase. After tumor removal, the Cerenkov signal from the surgical cavity dropped to the noise floor.

Figure 2. White light and CLI overlay images acquired from a commercial imaging system (top row) and a fiberscopic system (bottom row), showing the (a)–(d) exposed tumor and (e)–(h) excised tumor.

The ability to identify the extent of tumor invasion is also an unmet clinical need, especially in prostate cancer. The feasibility of identifying prostate cancer invasion beyond the capsule in an mouse model of an invasive tumor was studied with PET/computed tomography (CT) and CLI using a transgenic adenocarcinoma mouse model of the prostate (TRAMP). We performed PET imaging with a Siemens Inveon scanner, and all Cerenkov images were collected with a commercial optical imaging system (Perkin-Elmer). We injected six mice (four TRAMP animals and two healthy C57BL6 mice) via the tail vein with ∼1mCi of ¹⁸F-FDG. Cerenkov-guided surgery commenced immediately, with Cerenkov images collected to delineate tumor-containing tissues at the surgical margin. A progression of subsequent tumor excisions is shown in Figure 3. We performed image analysis to determine the signal-to-background ratio of the Cerenkov signal of tissues that contained tumors and those that did not (as verified by hematoxylin and eosin histological staining). All four TRAMP animals exhibited extensive tumor invasion in the prostate with subsequent invasion into the seminal vesicles. No tumor tissue was identified in the other tissues, as confirmed by pathology analysis. CLI was able to significantly differentiate between tissues involved with tumor and those with no tumor (P<0.01).

Figure 3. Overlays of CLI images onto ambient light images of a transgenic adenocarcinoma mouse model of the prostate (TRAMP mouse): (a) pre-prostatectomy, (b) post-prostatectomy, and (c) after removal of invasive tumor tissue.

These studies explored the feasibility of using CLI for the detection of tumor tissue, and determined that Cerenkov images from ¹⁸F-FDG were able to significantly differentiate between tumor and healthy tissue. We next plan to use PET/CT to help quantitatively guide CLI tumor resection in mice. We expect that the further synergy of pre-operative PET/CT and CLI will provide improved tumor-to-normal contrast and enable healthy/tumor tissue signal thresholding. We envisage that this fusion strategy of pre- and intraoperative imaging might have significant value in the clinic, and higher specificity than CLI alone.

This research was funded by the National Institutes of Health (grants NCI R01 CA128908 and NIH ICMIC P50CA114747), the Department of Defense Breast Cancer Postdoctoral Fellowship BC097779, and the Center for Biomedical Imaging at Stanford (CBIS). The author acknowledges the critical input and efforts of Hongguang Simon Liu, Guillem Pratx, Conroy Sun, Zhen Cheng, Sam S. Gambhir, and Prof. Lei Xing.