Since we acquired more than 14 images from more than 3 patients on a daily average, we had to adapt the workflow for the associated high-throughput demands. The concept of the SRH system suggests the use directly in the OR. In our experience, tissue acquisition occurred rather simultaneously in different concomitant operations. Therefore, it would be bothersome to transfer the imaging system fast enough. To optimize the workflow for our needs and adapt it to our OR infrastructure, we have placed the device at a fixed location in a research laboratory co-located at the OR floor and the samples were transferred directly after acquisition into this laboratory for further processing. The SRH images were initially presented to the surgeon via a tablet with Wi-Fi connection, which was recently complemented by connection of the device to our clinic’s imaging system IMPAX VIII (Agfa, Mortsel, Belgium) (Fig.
3).
Due to the high sample volume, meticulous and systematic documentation on paper and digitally was necessary in order to prevent any loss of information for a specific sample. The documentation of the distinct extraction point of the tissue in the neuronavigation software proved time consuming in the beginning.
Discussion
Standard-of-care intraoperative histopathology is the gold standard and helpful for operative decision-making in neurosurgical oncology. Nevertheless, it is time consuming and laborious. SRH is a new and innovative technique for near real-time intraoperative tissue imaging. It is easy to implement into the neurosurgical workflow and allows to rapidly obtain valuable and additional information relevant for the neurosurgeon. Therefore, SRH may indeed become a game changer for intraoperative tissue diagnostics during neurosurgical tumor resection, even if several technical alternatives for SRH in the field of neurosurgery already exist.
Next to sampling with subsequent analysis, there is the possibility of using a handheld probe for intraoperative tissue analysis on the basis of Raman spectroscopy [
4,
21]. However, this technique has several disadvantages, as the interpretation of the Raman spectra is solely possible via computational algorithms. A near real-time user-based analysis, which is one of the advantages of SRH, is not possible for the handheld probe and also not for table-top Raman spectroscopy [
15]. Moreover, white-light microscopy has to be paused while using a handheld Raman spectroscopy probe, which leads to a disruption of the neurosurgical workflow. Last but not least, the technique is easily disturbed by small movements.
Another alternative for intraoperative digital histopathology is fluorescein-assisted confocal laser endomicroscopy — the Convivo® System (Carl-Zeiss-Jena, Jena, Germany) [
8]. This system uses a handheld confocal laser endomicroscopy probe, which allows image acquisition in the brain without resection of the tissue. Due to the nature of confocal microscopy, it allows to capture images within a z-stack of 30-µm depth. However, the intravenous application of fluorescein as a dye is required to this date. The use of the Convivo®-device ex vivo was shown to be non-inferior to the gold standard [
1] and similar results may be expected also in vivo.
A major application of SRH is obtaining diagnostic information about the entity of resected putative pathological tissue. As described above, SRH images have now been integrated as DICOM files in our clinic’s imaging system IMPAX VIII immediately after imaging. Therefore, neuropathological assessment is possible at yet unprecedented speed. Previous studies [
5] and very recent studies from our group (cf. Straehle et al., [
18]) showed a non-inferiority of the assessment of SRH images by a board-certified neuropathologist compared to the assessment of H&E-stained frozen sections.
Another potential use case of Raman-based methodologies is the detection of metabolic changes in tumor tissue, which could be a surrogate parameter for specific mutations. The relevance of molecular diagnostics in brain tumors is high, as the new 2021 WHO classification of tumors of the central nervous system places a higher priority on molecular changes than ever before [
12]. Raman spectroscopy in general has shown to be a precise and reliable tool to determine changes at a molecular level. For example, the non-invasive Raman-based measurements of blood glucose levels are now established [
11]. IDH mutations are also causing metabolic changes, which are detectable by Raman spectroscopy [
20]. In future, research on modifications of hardware and software could also facilitate the intraoperative direct prediction of crucial molecular changes in the tumor.
SRH imaging may also be useful as an additional tool to achieve a greater extent of resection in diffusely infiltrating brain tumors. However, one should consider that there are differences between a real H&E image and the virtual H&E-like SRH image. For example, protein-rich extracellular fibers are usually red in the conventional H&E imaging, but may appear eosinophil in the virtual H&E-like SRH image. In order to use SRH as a robust modality to determine tumor histology or tissue diagnosis at the suspected tumor border, it is inevitable to fully understand the limitations and alternative illustration of SRH in comparison to classical staining in health and disease. These are also absolute prerequisites in order to use the SRH technology as an intraoperative tool for improvement of the extent of resection. One problem is the fact that the information provided by SRH imaging of white matter is very sparse (cf. Straehle et al., [
18]). In turn, the information-richness of SRH goes far beyond conventional staining. Given the molecular complexity of tissues, additional molecular markers may be identified and allocated via their characteristic vibrational signatures (i.e., at different Raman shifts). Thus, additional false-color maps illustrating particular molecular distributions or the location of specific molecular changes at the same sample may yield multiple yet inherently correlated images each highlighting a particular molecular feature still label-free.
The near real-time character of tissue analysis predestines SRH imaging for quality control of intraoperative biopsies. Still, if small lesions are resected or are subject to stereotactic biopsy, it is of utmost importance to increase the diagnostic yield of the specimen. Using SRH, an additional quality control step is implemented in addition to the standard of care intraoperative neuropathological assessment by H&E staining of frozen sections without disturbance of the neurosurgical workflow. One technical limitation of SRH is that the examined tissue has to be readily and uniformly compressible. Therefore, at the moment, it is not possible to image stiff specimen, which reduces the applicability of this method. On the other hand, SRH is not only limited to the field of neurosurgery. Other applications for example in Head and Neck surgery were also reported [
7]. In general, every tissue type that is compressible is suitable for analysis via this technique.
Our study shows that use of SRH imaging in a major European neurooncological center is feasible although the measures taken for this cause are not without drawbacks. Using the imaging system for multiple ORs can lead to delays in tissue processing, if the system is already imaging. This has never exceeded more than a few minutes and did not lead to noticeable lower quality of SRH imaging, but it may lead to suboptimal tissue handling in these exceptional cases.
Adaptations to the high throughput application of SRH imaging proved themselves effective. Nevertheless, some difficulties regarding the exact registration and documentation of the site of tissue acquisition remain. This issue refers to the tissue acquisition, for example, in a brain tumor resection but not for stereotactic biopsy. Of course, annotation points were ideally inserted for every sample in the MRI 3D dataset of the neuronavigation system. Yet, system inherent factors like inaccuracies of the initial patient registration or the selection of the exact extraction localization with the pointer are only minor problems. Brain shift effects on the accuracy of the neuronavigation system prevent a precise and valid documentation of the exact tissue extraction points.