In-vivo longitudinal imaging of microvascular changes in irradiated oral mucosa of radiotherapy cancer patients using optical coherence tomography

Mucositis is the limiting toxicity of radio(chemo)therapy of head and neck cancer. Diagnostics, prophylaxis and correction of this condition demand new accurate and objective approaches. Here we report on an in vivo longitudinal monitoring of the oral mucosa dynamics in 25 patients during the course of radiotherapy of oropharyngeal and nasopharyngeal cancer using multifunctional optical coherence tomography (OCT). A spectral domain OCT system with a specially-designed oral imaging probe was used. Microvasculature visualization was based on temporal speckle variations of the full complex signal evaluated by high-pass filtering of 3D data along the slow scan axis. Angiographic image quantification demonstrated an increase of the vascular density and total length of capillary-like-vessels before visual signs or clinical symptoms of mucositis occur. Especially significant microvascular changes compared to their initial levels occurred when grade two and three mucositis developed. Further, microvascular reaction was seen to be dose-level dependent. OCT monitoring in radiotherapy offers a non-invasive, convenient, label-free quantifiable structural and functional volumetric imaging method suitable for longitudinal human patient studies, furnishing fundamental radiobiological insights and potentially providing useful feedback data to enable adaptive radiotherapy (ART).

The painful inflammation and possible ulceration of the mucous membranes lining the digestive tract known as mucositis is most common limiting toxicity of radio(chemo)therapy of head and neck cancer 1,2 . Its pathogenesis is complex and is based on the interaction of various cell and tissue factors, including both the effects of the oral cavity microflora and the development of vascular reactions [3][4][5] . Activation of the coagulation system is also known to play an important role in the development of acute radiation reactions 6 . The direct effect of the exposure of mast and endothelial cells to ionizing radiation is the generation of thrombin, and the release of histamine and prostaglandins I2 and Е2, leading to vasodilation and to increased vascular permeability resulting from the adhesion of neutrophils to the endothelial surface in the first several hours following irradiation [7][8][9][10] .
All modern systems of grading the severity of radiation mucositis are based on patient symptoms and complaints, and visual changes in the mucosae identified during examination of the oral cavity; these cannot provide objective and quantifiable information 11 . Experimental studies using repeated biopsies and histological investigations for evaluating reactions of the microstructure of the mucosa have been reported 12 , but these are rare and have little appeal for clinical practice adoption. Clearly, a noninvasive in-vivo method for detecting subclinical and quantifiable changes of the irradiated mucosa that are impossible to register during a standard visual examination is highly desirable. It would enable currently-impossible clinical tasks, including (1) detecting mucosal reactions before the onset of their clinical manifestations, thus offering the possibility of early intervention in suitable patients, (2) properly evaluating the effectiveness of different preventive measures and treatments of the mucositis, (3) monitoring of the radiobiological response of the different mucosal components such as small versus large vessels RT effects, their temporal dynamics relative to mucositis manifestation and relative to structural tissue changes (e.g., epithelium nature and thickness), and (4) possible alterations in the radiation treatment delivery in the context of ART (adaptive radiotherapy 13 ). Such a method should be ideally noninvasive, contrast-agent-free, suitable for repeated in-vivo investigations, well tolerated by patients, and provide useful information in real-time; the ability to detect tissue functionality including mucosal microcirculation during the course and after irradiation would also be desirable 14,15 .
Optical coherence tomography (OCT) is a medical noninvasive imaging modality that seems suitable for the task mentioned above at hand, with some pre-clinical and clinical experience in the diagnosis of malignant tumors and detailed in-vivo tissue assessment for over 20 years 16,17 . In recent years, works have been published on OCT application in evaluating the consequences of radiation/chemoradiation therapy of oral/pharyngeal tumors 18,19 . It has been previously shown that normal oral mucosa is easily accessible for OCT imaging with a suitably-designed probe, and exhibits a well-delineated high-contrast stratified structure on OCT 19 .
In addition to high-resolution microstructural tissue imaging, alternate OCT contrast mechanisms have enabled sensitive imaging of the microvasculature based on Doppler-and speckle-based OCT 20,21 .
Recent microvascular OCT studies in the oral mucosa of healthy human volunteers have also been performed 22 . Experimental preclinical 23 and preliminary clinical studies 18 have examined the possibility of OCT monitoring of the onset and progression of radiation mucositis (but without microvascular assessment). Wilder-Smith et al. 24 offered a semi-quantitative method based on the combined evaluation of structural OCT changes and damage to the microcirculation in chemotherapy-induced oral mucositis on hamsters as determined by Doppler studies 24 (although unable to detect/assess the fine changes in capillary-like vessels studied here). OCT studies for monitoring late oral radiation toxicity based on microstructural and limited microvascular data (including approximate categorizing of vessels with diameter >50 μm) have also been reported 25,26 . Given this promising research activity, the potential importance of blood microcirculation in mucositis pathogenesis 27 , and recent technological advances in multifunctional OCT (including microcirculation imaging 28 , with its much improved microvascular visualization and quantification capabilities), a pilot study of microstructural and microvascular OCT in-vivo imaging in head and neck radiotherapy patients is reported. Our OCT study quantifies RT-induced changes in the microvasculature, including in smallest capillary-like (<15 μm in diameter) vessels in the human oral mucosa of radiotherapy patients. Fine blood microcirculation details can potentially be detected by other techniques such as direct oral microscopy 29 , conventional capillary microscopy 30,31 , orthogonal polarization spectroscopy (OPS) 32 , side stream dark field (SDF) imaging [31][32][33][34] , and narrow band imaging (NBI) [35][36][37][38] . In comparison to these, OCT looks particularly promising due to its signal information content, imaging speed, resolution, penetration depth, access to different clinical sites via flexible fiber optic probes, and relative robustness in realistic clinical imaging conditions (e.g., tissue/patient motion artefacts).
To the best of our knowledge, previous works do not closely overlap with the current study and its intent to (1) show the technical feasibility of OCT monitoring in radiotherapy patients (with focus on quantifying changes in capillary-like vasculature), (2) evaluate its compatibility with clinical radiotherapy patient workflow, and (3) elucidate the initial radiobiological trends of mucosal tissue response (including mucositis development).

Results
The OCT angiographic images of the normal mucosa show dense, volumetrically uniform microvascular networks, mostly consisting of relatively large vessels (Fig. 1a). At doses of 4-8 Gy, all patients exhibited an increased vessels density (Fig. 1b). Continued dose accumulation caused an increase in the clinical manifestations of radiation reaction, as also seen by the 2D OCT angiographic maps (Fig. 1c,e,g). After start of anti-mucositis therapy (typically chamomile and antiseptic washes in case of grade 1 (Fig. 1d); washes and analgesics in case of grade 2 (Fig. 1f); analgesics, antibacterial and antifungal therapy in case of grade 3 (Fig. 1h)), only in the case of grade 3 mucositis, the anti-mucositis therapy tended to slow down the microvascular increase (Fig. 1h). These tentative visual conclusions were confirmed by quantitative processing of the OCT images; we chose two quantification metrics to analyze these 2D OCT angiographs, as described in Materials and Methods section (Fig. 1i,j). Especially significant changes of microvascular parameters (compared to their pre-RT values) were detected when grade two and three mucositis developed (Fig. 1i,j).
In addition to these general trends based on the analysis of all examined patients, the real potential of OCT microvascular monitoring was evident in patients who received Intensity Modulated Radiation Therapy (IMRT). The first clinical symptoms of mucositis usually occurred after accumulation of 10-15 Gy in the buccal tissue. This may correspond to different dose accumulated in the Planning Target Volume (PTV), dependent on PTV location and treatment plan. Two case reports are presented below (Figs. 2 and 3).
Case report 1: male, age 55. Diagnosis: nasopharyngeal cancer, T2N2M0, a low-differentiated squamous cell cancer. Before radiation therapy, four cycles of inductive chemotherapy were carried out. Patient was treated by IMRT and, accordingly, the treatment plan involved a conformal dose distribution with high dose gradients near the normal tissue borders. The Dose-Volume Histogram (DVH) for right and left cheek (monitoring sites) were similar (Fig. 2b) and corresponded to 40-45% of tumor (PTV) dose levels (Fig. 2a). The first clinical symptoms of mucosal reaction (hyperemia) occurred after a total PTV dose of 22 Gy, corresponding to ~10 Gy on right and left buccal mucosa.
Mucositis reached no more than first degree for 10 days. After beginning of radiation reaction (occurring at the total PTV dose of ~22 Gy), the patient was treated by chamomile and antiseptic washes. After that, visual symptoms of mucositis began decreasing.
OCT monitoring of the buccal mucosa showed similar microvascular changes in the right and left cheeks (Fig. 2c,d), and the appearance of a vascular reaction prior to clinical manifestations and symptoms of mucositis. At their maxima at doses of ~30 Gy, the total length of <15-μm-diameter vessels increased by 40% and 35% on the right and left cheeks, respectively. The similarity in the values the microvascular metrics and their time-course changes for the two cheeks is indicative of their similar doses; importantly, these metrics show significant changes well before clinical manifestations of mucositis.
Case report 2: female, age 34. Diagnosis: cancer of the minor salivary gland of the right cheek T1N0M0, mucoepidermoid carcinoma after a non-radical resection of the tumor. Postoperative IMRT (Fig. 3a) on the tumor bed was performed with total dose = 44 Gy, dose/fraction = 2 Gy (22 fractions). In this patient, the PTV and right cheek's tissue including mucosa coincided. The first clinical symptoms of oral mucosa reaction (h) -after initiation of anti-mucositis therapy. Corresponding to these 8 representative svOCT image panels, summary statistics from the entire patient cohort are summarized in (i) -average vascular density; and (j)total length of <15-μm-diameter vessels. In (i,j), data shown are mean ± SD; number of analyzed 3D image data sets and number of patients = 61 and 25 (before RT), 57 and 25 (before visual signs), 19 and 5 (grade 1 mucositis), 21 and 5 (after initiation of anti-mucositis therapy), 20 and 12 (grade 2 mucositis), 24 and 12 (after initiation of anti-mucositis therapy), 21 and 5 (grade 3 mucositis), 22 and 5 (after initiation of anti-mucositis therapy). Blue bar is a peak of symptoms of mucositis; red bar is after initiation of anti-mucositis therapy (see text for details). *Statistically significant difference compared to pre-RT levels (one-tailed t-test, p ≤ 0.05).
(hyperemia) occurred after a total PTV dose of ~14 Gy. Since the left cheek was more distant from PTV region, its dose was smaller, which caused asymmetry in the right/left cheek responses. The appearance of second grade mucositis (patches and erosions) occurred after a dose of 20 Gy on the right cheek and persisted for 6 days. On the left cheek mucosa, the severity of mucositis was less significant and was limited to hyperemia and single fibrous patch.
Analysis of OCT images of the right cheek indicated the increase of the total length of <15-μm-diameter vessels and the vascular density immediately after beginning of radiation therapy (Fig. 3c,d). When the total dose was 8 Gy, the number of small vessels on the right buccal mucosa increased by 17.5%, and at the peak of the reaction (dose ~20-22 Gy) by more than 56%. After beginning of antibacterial (anti-mucositis) therapy, these indicators decreased and reached an initial level at the dose of 30 Gy. We also detected and quantified the difference in the microvascular changes on the right and left cheeks, as expected from their different dosimetry (Fig. 3b). The number of <15-μm-diameter vessels on the left buccal mucosa showed a much smaller increase, reaching its maximum of 21% at a PTV dose of 36 Gy that corresponded to a dose of approximately 10 Gy to left cheek mucosal tissue.

Discussion
The development and clinical adoption of conformal irradiation including IMRT improved dose delivery to the tumor and decreased dose received by organs at risk. However, the significant problem of oral cavity mucositis has not disappeared. In fact, it seems as if IMRT's highly conformal dose distributions and high dose gradients in both pathologic and normal tissues have made its appearance and severity more unpredictable and plan/patient specific 39 . Hence, diagnosis, prophylaxis and corrective treatments of radiation induced mucosal toxicity demand correspondingly more accurate and objective approaches. OCT is a non-invasive, label-and contact-free, convenient, high-resolution 3D structural and functional subsurface imaging modality that may meet some of these demands, including early detection prior to clinical manifestation of toxicities. While other technologies may potentially be adapted to monitor microvascular changes in RT patients (e.g., direct oral microscopy 29 , conventional capillary microscopy 30,31 , orthogonal polarization spectroscopy 32 , side stream dark field imaging 31-34 , narrow band imaging [35][36][37][38] , laser Doppler flowmetry, high-frequency Doppler ultrasound, and confocal or intravital endoscopy 40 ), OCT's unique combination of capabilities and strengths bodes well for its increased use in many potential scenarios in radiotherapy, in particular those with microvascular involvement. This study reports the first use of OCT for in-vivo longitudinal monitoring of microvascular monitoring of oral mucosa in the course of radio(chemo)therapy in patients with oral, oropharyngeal and nasopharyngeal cancer. Microvascular reactions as imaged and quantified by OCT were dose-level dependent and were detected before clinical mucositis was observed. These findings may have relevance in the context of early intervention in patients who, on the basis of early-stage OCT microvascular data, are likely to develop severe oral complications; as such, this may enable data-informed adoption of ART scenarios. In the cases of IMRT patients, similarity or dis-similarity of microvascular metrics in the different mucosal sites tracked with corresponding doses/DVHs (Figs. 2 and 3). The total dose to the mucosa of each cheek depended on the treatment plan and in some cases (e.g., IMRT) could significantly differ from the right and left sides. In this and similar situations, OCT proved to be a very useful tool to identify subtle dose-level-dependent changes in the microcirculation, which are not determined by visual inspection alone.
Further, the study demonstrated the technical feasibility and clinical use of a prototype OCT system suitable for in-vivo, longitudinal, objective and quantifiable high-resolution imaging of the subsurface microstructure and microvasculature of oral mucosa in RT patients. Importantly, the OCT examination procedure was well tolerated by most consenting patients.
The presented OCT vascular imaging and quantification method has some methodological limitations. Indeed, mucositis is a complex pathological process that affects not only tissue microvasculature but its microstructure/morphology as well (e.g., thickness of mucosal layers, scattering properties). As these in turn may (slightly) affect OCT vessel visualization, the derived microvascular metrics may also reflect slight contributions from tissue morphology alterations.
Further, motion artefacts are typically present in OCT angiographic images and reduce their quality. Attempts to eliminate/compensate for the artefacts may lead to loss of the useful signal as well (like in other in vivo imaging techniques). Further development of robust algorithms to avoid these losses are needed.
Overall, this study represents early stage phase I research with a limited patient cohort, and more work is needed to better assess technological feasibility and potential clinical utility of OCT microangiography in radiotherapy. In particular, more extensive exploration of correlation between the early OCT-detected mucosal microvascular changes in the course of RT and the early and late RT toxicities is planned. Further research will also encompass a more thorough mucosal assessment by multifunctional OCT, including quantitative (d) -total length of < 15-μm-diameter vessels. The patient was treated by chamomile and antiseptic washes to combat RT toxicity (after ~16 Gy on the right cheek; left cheek mucositis manifestations were very limited; for details, see text). Note that both microvascular metrics exhibit significant changes much earlier than the clinical manifestation of mucositis on the right cheek. evaluation of the microstructure of the mucous membrane layers and the state of its connective tissue matrix (via polarization-sensitive OCT). This will provide more accurate information about subtle changes in the various mucosal compartments (epithelium, connective tissue, microcirculatory bed) during irradiation. Objective testing of the efficacy of various agents for the prevention and treatment of mucositis via multifunctional OCT is also envisioned.

Conclusion
The study showed that longitudinal OCT angiographic monitoring can be used for objective evaluation of radiation induced microvascular volumetric changes in the oral mucosa, thus 'shedding light' on the temporal sequence of early functional and structural radiation toxicities. This may potentially play a role in the design and effectiveness evaluation of anti-mucositis treatment and prophylaxis modalities, and in the implementation of adaptive radiotherapy protocols.

Materials and Methods
Patients' characteristics. Longitudinal imaging results from twenty-five (25) patients with stage I-IV of oral, oropharyngeal and nasopharyngeal squamous cell carcinoma are reported. Patient study was performed in the department of radiation oncology of Nizhny Novgorod Regional Oncology hospital and approved by the Research Ethics Board of the Nizhny Novgorod State Medical Academy. Informed consent was obtained from all participants and/or their legal guardians enrolled in the study. All methods were performed in accordance with the relevant guidelines and regulations. The patient characteristics and treatment methods are summarized in Table 1. All target volumes were irradiated to a total dose of 46-70 Gy as deemed clinically useful for tumor cure, but which can also cause radiation side-effects in the mucosa (oral cavity radio-toxicity such as mucositis). Irradiation was performed using a linear accelerator (Varian Clinac 600) or Cobalt-60 system (Terabalt). Mucositis degree was scored by Radiation Therapy Oncology Group and European Organization for Research and Treatment of Cancer (RTOG/EORTC) scale. All 25 patients received prophylaxis and treatment of mucositis according to our hospital standards (diet and oral care). From the first day of RT, patients were advised to rinse oral cavity and pharynx by chamomile as often as possible. Patients with grade 2 mucositis (single erosions and plaques) were assigned antiseptic washes and analgesics. For grade 3 mucositis, analgesics, antibacterial and antifungal therapy were assigned.
Multifunctional OCT imaging. OCT imaging was performed twice a week, starting from the first day of irradiation or chemotherapy on two standard, symmetric sites inside both cheeks (Fig. 4). In all cases, the minimal distance from visual tumor (PTV) border to the cheek monitoring sites was >3 cm, which allowed us to minimize the influence of tumor angiogenesis effects. A modified ophthalmic head restrainer was used, with chin rest and forehead strap ensuring adequate immobilization for the OCT imaging session. After initial exploration and mapping of the oral cavity, sites along the centre line connecting the secretory duct of the salivary gland and the angle of the mouth were chosen. Such anatomical referencing allows for reproducible OCT angiographic images (Fig. 4).
Buccal mucosa (cheek tissue) was selected for OCT monitoring. It is an accessible, convenient and clinically important site for the introduction and reliable positioning of the OCT probe, which is important for image quality; this location also minimizes patient discomfort during the imaging study. It was verified that the OCT images of the buccal mucosa were visually and qualitatively similar with the probe re-positioning of ±several millimeters; therefore, probe positioning based on visual landmarks was sufficient to provide consistent results. Typically, six 3D data sets from two sites in each patient (with three repeats) were collected during each imaging session. In the pre-RT examinations, no statistically significant differences were found between left and right cheek within three repeats as shown in Fig. 4(g,h) (although the tumor was asymmetrically located at the left side of the tongue in this case). This demonstrates our ability for robust probe repositioning in the context of a longitudianl imaging study. Overall ~70% of OCT angiographic images were deemed visually appropriate, but for accurate quantification of microvascular parameters only ~40% were used, since some of these actually suffered motion artefacts manifest as bright stripes on OCT angiographic images. The OCT imaging procedures were well tolerated by consenting patients. No statistically significant differences were found between separate measurements near same anatomical reference points.  [41][42][43] . Infrared laser power incident on tissue was low (~2 mW), and the imaging sessions were relatively fast (26 seconds beam-on scanning time, ~3 minutes total for two oral mucosa locations including set-up and alignment). The oral OCT probe (7 cm length × 1.1 cm diameter) was based on common-path interferometry scheme to provide convenient operation during the clinical study. The electromechanically actuated cantilever with optical fiber tip and focusing micro-lens was placed in the stainless steel tube with fused silica output window, to provide biocompatibility and ease of sterilization between imaging sessions. Patient's head was immobilized in a head support frame to minimize motion artefacts. The optical probe was positioned at the oral mucosa site with gentle contact using an articulated arm apparatus (Fig. 5). The lens system formed a 15 µm dimeter focal spot 500 µm distal to the end face of the probe's output window.
Based on temporal speckle variations as the source of angiographic image contrast, 3D OCT angiographic images (3 mm × 3 mm laterally, ~1.5 mm depth) were obtained. The angiographic calculation algorithm was based on image phase alignment to reduce motion artefacts and high-pass filtering along slow scanning axis 28,44 . Optically linearized spectrometer based on our previous development 40 reduced signal computational complexity and allowed real time visualization of angiographic images. Between patients, the probe underwent cold chemical sterilization in a polyhexamethylene biguanide hydrochloride solution.
Processing of OCT images. For OCT angiographic analysis and quantification, each obtained 3D microvascular network was converted to 2D images with Maximum Intensity Projection (MIP) by projecting the whole imaged depth (~1 mm) onto the plane. The choice of 2D analysis helps to minimize the influence of shadow artefact cast by flowing blood below the true vessels 45 . Resulting 2D images were binarized and skeletonized. Vessel thickness was calculated as 2x the distance between the vessel's binary image border and its skeleton; vessels with overlapped binary borders and skeletons were assigned a thickness of 1-pixel. For the lateral resolution of the current OCT imager, this corresponds to vessel diameters less than ~15 μm. The total length of such thinnest vessels served as one quantified microvascular metric. The second derived metric was vessel density, which was calculated as the number of pixels of all vessel skeletons in the analyzed image area divided by the total number of pixels in this area.
Other more advanced microvascular quantification biomarkers are currently being explored (fractal dimension, Euclidian segment lengths, etc -see for example 46 ), but these two selected ones sufficed for the current feasibility study.

Statistical analysis.
Statistical analysis used the one-tailed Student's t-test (Statistica 6.0), with p ≤ 0.05 being considered statistically significant. The values of the two reported microvascular metrics throughout the course of therapy were compared to pre-RT levels.
Data availability statement. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.