Optical Biopsy of Sinonasal Tumors using Confocal Laser Endomicroscopy: A Clinical and Deep Learning-based Assessment and Visualization

• Applicants: Professor Dr.-Ing. Marc Aubreville; Privatdozent Dr. Miguel Goncalves
• Subject Area: Medical Physics, Biomedical Technology ; Otolaryngology, Phoniatrics and Audiology
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 545049923

Confocal laser endomicroscopy (CLE) is a promising imaging technology that provides a magnified, cellular-level view of superficial epithelium. It utilizes a small laser scanning probe applied to the area of interest, reaching a magnification of up to 1000 times in-vivo. The main motivation for non-invasive in-vivo diagnostics is the more precise definition and control of resection margins and diagnosis in functional critical areas. Thus, CLE could offer the possibility to keep tumor resection volume as low as possible and to realize a sufficient evaluation of the entire resection margin at the same time. Although our research group has achieved promising results regarding the use of CLE in the aerodigestive tract (mainly: the oral cavity, pharyngeal and laryngeal region), a systematic analysis of confocal laser endomicroscopy in the sinonasal tract has not yet been performed by us or other research groups. The sinonasal tract is anatomically challenging, since the internal carotid artery, optic nerve, orbit and skull base are in close vicinity. Over-sampling of resection margins by invasive biopsies can result in catastrophic complications and under-sampling in recurrence. Navigation-guided surgery using high resolution computed tomography (photon-counting CT, PCCT) fused with MRI enables an exact placement of the CLE probe on the interest area. The matching between CLE, endoscopic finding, imaging-based examination (MRI/CT) and histological is therefore possible with high precision and validity. With this project we intend to create during the 36 months funding period a database of CLE image data for the sinonasal tract which is suitable for the training of physicians as well as machine learning algorithms. This database will include inflammatory and tumoral (benign, such as inverted papillomas, as well as malign) mucosal lesions and physiological mucosa. Next, we intend to define, validate and improve clinical evaluation parameters and scoring systems for the assessment of CLE video sequences of the sinonasal tract based on our acquired data. Based on the expert-curated database, automatic detection and classification of the different affections of the sinonasal tract based on CLE images up to a clinically applicable level of accuracy and robustness are to be developed. Finally, through the possibility of intraoperative navigation we intend to generate a prototype of virtual CLE-enhanced endoscopy, i.e. a 3D visualization of anatomically mapped automatic tumor analysis. In sinus surgery, as the frontier between mucosa/brain and other sensible structures such as the orbital content, CLE is conceptually an ideal technique to limit over- and undersampling and could thus enable precision oncologic surgery and complete removal of the tumor, which is the most important survival factor for these patients.

• DFG Programme: Research Grants
• Major Instrumentation: Konfokales Laserendomikroskop
• Instrumentation Group: 5090 Spezialmikroskope