Learned end-to-end high-resolution lensless fiber imaging toward intraoperative real-time cancer diagnosis

Abstract

Endomicroscopy is indispensable for minimally invasive diagnostics in clinical practice. For optical keyhole monitoring of surgical interventions, high-resolution fiber endoscopic imaging is considered to be very promising, especially in combination with label-free imaging techniques to realize in vivo diagnosis. However, the inherent honeycomb-artifacts of coherent fiber bundles (CFB) reduce the resolution and limit the clinical applications. We propose an end-to-end lensless fiber imaging scheme toward intraoperative real-time cancer diagnosis. The framework includes resolution enhancement and classification networks that use single-shot fiber bundle images to provide both high-resolution images and tumor diagnosis result. The well-trained resolution enhancement network not only recovers high-resolution features beyond the physical limitations of CFB, but also helps improving tumor recognition rate. Especially for glioblastoma, the resolution enhancement network helps increasing the classification accuracy from 90.8% to 95.6%. The novel technique can enable histological real-time imaging through lensless fiber endoscopy and is promising for rapid and minimal-invasive intraoperative diagnosis in clinics.