This retrospective study occurred between 2018-2020 in a single center and received Institutional Review Board approval. Inclusion criteria was determined as patients 18 years or older presenting with either a progressively growing tumor, demonstrated by two sequential MRI scans taken 3 months apart, or with a highly symptomatic tumor that was unresponsive to conventional analgesic treatment. All the patients in this study were reviewed by the institutional multi-disciplinary tumor board, comprised of an oncologist, radiologist, pathologist, and orthopedic oncologist. All treatment options were considered prior to transitioning from a hands-off, wait-and-see policy to cryosurgery.
Preoperative planning protocol
3D modules reconstruction
An MRI scan was used to generate a computerized three-dimensional module. The module was generated using Mimics® software (v23, Materialise, N.V. Leuven, Belgium). A customized preoperative plan was constructed for each individual patient (Figure 1). The module also considered potential at-risk structures such as major vessels, nerves, and vital organs, to ensure that any potential plan would not jeopardize them. To achieve optimal tumor coverage, the number of needles, trajectory, and placement were planned based on the generated modules. An ice ball with a radius of 4 cm was generated at the end of each cryo-needle to use a reference for the ablation zone of each needle. The total ablation zone was determined to equal the sum of all the individual ice balls.
Pre-operative MRI and fiducials markers placement
One day prior to ablation, skin fiducial markers (pinpoint, beekly medical LTD) were scattered in a random pattern around the tumor and a preoperative MRI scan was performed. The markers were kept in place for surgery the following day to ensure that the intraoperative registration included the precise location of the markers displayed in the navigation system. In addition, the MRI scans were used to re-evaluate the tumor size, location, and composition in order to assess for any deviations from the preoperative plan (Figure 3).
Preparation and image registration
All the interventions were performed by a fellowship-trained orthopedic surgeon specializing in orthopedic oncology. Surgeries were performed in an operating room and the patients were placed under general anesthesia. Stealth-station navigation system (StealthStation®, Medtronic Sofamor, Danek) was set to cranial model in order to allow optimal soft tissue demonstration. Then, the MRI scans taken the previous day were uploaded to the stealth station as DICOM files. Registration was performed for each individual skin fiducial marker using pointer (medtronics). At the end of the registration and calibration process, the skin fiducial markers were removed. Each individual cryo-needle was calibrated to the generated image using the sure-tract system (Danek and Sure-Track system by Medtronic). Using MRI based navigation, the cryo needles (Galil medical-ice rod) were placed inside the tumor mass according to the preoperative plan. The location and position of each needle were confirmed using an ultrasound (US) performed by I.D, a fellowship trained radiologist. After placing all the cryo needles, a Cone beam CT scan was performed as a cautionary measure to ensure that the chosen needle location was identical to the preoperative model (O‐arm scanner Medtronic Sofamor; Dannek). Subsequently, the ablation protocol began.
The treatment protocol included ten minutes of freezing followed by five minutes of thawing and another ten minutes of freezing. Neuromonitoring was used as a precautionary in procedures that were performed close to major nerves. Skin temperature was closely monitored to avoid any thermal damage. To protect the underlying skin, a barrier was created by injecting continuous warm normal saline into the subcutaneous layer to hydrodissect the skin from underlying tissue. Patients were permitted to return home one day after the procedure
Postsurgical surgical image analysis
Patients underwent MRI scanning at 3, 6 and 12 months post ablation. Following surgery, the IntelliSpace Discovery software platform (Philips Healthcare) employed a novel Gaussian mixture model (GMM) algorithm developed specifically for post-surgical imaging analysis This method employed a semi‐automatic 3D segmentation tool to successfully identify variations in complex tissues at the macroscopic volumetric segmentation of the desmoid tumor tissue of each patient (Figure 3). Images were evaluated before and after ablation. Total Tumor Volume (TTV) was evaluated and compared to the preoperative protocol. Tumors were then further categorized into viable (high intensity/ bright) and non-viable tumors (low intensity/ dark) and were evaluated semi-automatically using Phillips software according to their density. Tumor dimension and composition were assessed using a series of consecutive postoperative MRI scans (3,6, and 12 months). Minimum follow-up time was 12 months.
Data collection and follow up
Medical records were reviewed for demographic and clinical data, tumor volume, tumor location, and symptoms prior to the intervention. Additionally, prior treatments were reviewed and documented. Intra-operative CT scans were used to compare each element of the procedure with the pre-operative 3D model, including the number of needles used and potential involvement of untargeted structures. Any discrepancy between the pre-operative plan and the execution was documented. Patient follow-up was conducted in the outpatient clinic at three, six, and twelve months. Pre-operative and post-operative symptoms were compared. An SF-36 was administered pre- and 12 months post-operation to evaluate individual patient health status and compare disease burden.
All continuous data was presented as mean ± SD and compared between pre- and post-operation using a paired t-test. Significance threshold was set at p-value<0.05. Comparisons were conducted using IBM SPSS software (V25).