br on a micrometer scale Vernier caliper An
on a micrometer scale (Vernier caliper). An image of the tumor section was acquired and stored in JPEG format. The camera lens was posi-tioned vertical to the tumor section during image acquisition (Fig. 1b). Then, the surgical specimens were preserved in formalin solution. Pa-thological sections and hematoxylin and eosin (H-E) staining were performed on the sections from which the maximum DOI was mea-sured. The obtained sections were scanned digitally with a scanner (NanoZoomer 2.0-RS, Hamamatsu Photonics K.K., Japan). The max-imum infiltration depth was measured in the software (Fig. 1c). For data comparability, the depth of tumor infiltration for these Ac-DEVD-CHO measure-ments was the vertical distance between the simulated normal mucosal junction and the deepest point of tumor infiltration. For exogenous tumors, the part above the mucosal surface was neglected, and for ul-cerative tumors, the invaginated part was added. The depth of in-filtration was measured by an experienced clinician (with over 5 years of experience) under the guidance of a more senior physician. If a disagreement occurred between the two physicians, the final decision was made after consultation with a radiologist.
Tumor images were intercepted and preserved in JPEG format from preoperative MRI, intraoperative tumor section images, and post-operative pathological section images. The above images contained a line indicating the specific depth to facilitate restoration of the true size of the image. Mimics 15.0 (Materialise Company, Belgium) was used to open the JPEG file and set the threshold, generate a mask, edit the mask, and calculate and generate entities in the Standard Template Library (STL) format. The generated STL file was imported into Geomagic Control 16.0 (Geomagic, America). The pathological data were used as the reference template for measurement, while the MRI data were used as the test object. The best fit and artificial registration were used to fit the pathological data and MRI data in the same co-ordinates for two-dimensional comparison. The registration criterion was the best fit for mucosal tumor morphology. After analysis, a report file was generated to record the average diﬀerence between the two-dimensional tumor margins (Fig. 2a–k). All the data were obtained with informed consent and permission from the Ethics Committee of Beijing Stomatological Hospital, Capital Medical University.
All patients were followed up to record local control and regional distant metastasis. The patients were followed up once every 2 months
Fig. 1. Measurement of invasion depth on MRI (a), intraoperatively (b) and pathologically (c). Measurement of the DOI based on the adjacent normal mucosal junction to the deepest infiltration point.
Fig. 2. Two-dimensional analysis of tumor margins. a. Preoperative intraoral image of squamous cell carcinoma of the right lingual margin. b. Intraoperative tumor section image. c. Range of the tumor on magnetic resonance imaging based on the T2 sequence. d. Postoperative pathological scan. e–g. Separation and extraction of tumor images from intraoperative, MRI and pathological images. h–i: The Best Fit and artificial registration were used to fit the MRI data and intraoperative data; the pathological data and MRI data are phosphate group in the same coordinates for two-dimensional comparison. j–k. Average diﬀerence of TD-DMI and TD-DMP.
during the first year after the operation and every 3 months during the second and third years. The examination included MRI/CT, chest radiography, abdominal B ultrasound, and PET-CT when necessary.
The relationships among the DOI measurements by MRI and by analysis of the intraoperative and pathological specimens were esti-mated using the Bland-Altman plot, t-tests were used to analyze the agreement between the three values, and multiple linear regression was used to analyze the associated factors. The cutoﬀ values of the DOI for predicting nodal status, overall survival (OS), and disease-specific sur-vival (DSS) were determined using a receiver operating characteristic (ROC) curve analysis. Kaplan-Meier plots were constructed to present cumulative survival outcomes. Analyses of the relationships between clinicopathologic parameters, DOI, disease-free survival (DFS) and DSS were performed using Cox proportional hazard models. The hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated. All statistical analyses were performed using IBM SPSS software version 25.0 (IBM Corp., Armonk, NY). A two-sided p value < 0.05 was con-sidered statistically significant.