Methods: Retrospective analysis of data of seven patients with oral and maxillofacial tumors who underwent surgery between January 2019 and January 2021 using a combination of mixed reality and surgical navigation. Virtual surgical planning and navigation plan were based on preoperative CT datasets. Through IGT-Link port, mixed reality workstation was synchronized with surgical navigation, and surgical planning data were transferred to the mixed reality workstation. Osteotomy lines were marked with the aid of both surgical navigation and mixed reality images visualized through HoloLens. Frozen section examination was used to ensure negative surgical margins. Postoperative CT datasets were obtained 1 week after the surgery, and chromatographic analysis of virtual osteotomies and actual osteotomies was carried out. Patients received standard oncological postoperative follow-up.

Results: Of the seven patients, four had maxillary tumors and three had mandibular tumors. There were total of 13 osteotomy planes. Mean deviation between the planned osteotomy plane and the actual osteotomy plane was 1.68 ± 0.92 mm; the maximum deviation was 3.46 mm. Chromatographic analysis showed error of ≤3 mm for 80.16% of the points. Mean deviations of maxillary and mandibular osteotomy lines were approximate (1.60 ± 0.93 mm vs. 1.86 ± 0.93 mm). While five patients had benign tumors, two had malignant tumors. Mean deviations of osteotomy lines was comparable between patients with benign and malignant tumors (1.48 ± 0.74 mm vs. 2.18 ± 0.77 mm). Intraoperative frozen pathology confirmed negative resection margins in all cases. No tumor recurrence or complications occurred during mean follow-up of 15.7 months (range, 6-26 months).

CASE SUMMARY: We reported a case of sclerosing odontogenic carcinoma of the maxilla in a 62-year-old woman, who presented with an indolent right palatal swelling, which progressively increased in size over 7 years. Right subtotal maxillectomy with surgical margins of approximately 1.5 cm was performed. The patient remained disease free for 4 years following the ablation surgery. Diagnostic workups, treatment, and therapeutic outcomes were discussed.

APPROACH: In this paper, we propose a novel model called radiomics-reporting network (Radioport), which incorporates text attention. This model aims to improve the interpretability of deep learning radiomics in mammographic calcification diagnosis. Firstly, it employs convolutional neural networks (CNN) to extract visual features as radiomics for multi-category classification based on Breast Imaging Reporting and Data System (BI-RADS). Then, it builds a mapping between these visual features and textual features to generate diagnostic reports, incorporating an attention module for improved clarity.

MAIN RESULTS: To demonstrate the effectiveness of our proposed model, we conducted experiments on a breast calcification dataset comprising mammograms and diagnostic reports. The results demonstrate that our model can: (i) semantically enhance the interpretability of deep learning radiomics; and, (ii) improve the readability of generated medical reports.