16th DDc Lunch Retreat: Exploiting Deep Learning and Lensless Optics for Paradigm Shifts in Biomedicine

Light has the potential to recognize the origins of diseases, to prevent them, or to cure them early and gently. The early diagnosis is the key to improve the survival rate and cure rate of patients. Endoscopy plays an important role in the early stages of diagnosis by guiding biopsy extraction. Conventionally, it takes several hours to a few days for the surgeon to know the results of the diagnosis. Optical biopsy offers real-time intraoperative diagnosis and reduces the risk of surgical resection.

The state of the art are fiber endoscopes, but the lens optics limits the size. Lensless fiber endoscopy offers both small diameters of a few 100 microns and the suitability as single-use probes, which is beneficial in sterilization. We demonstrate an end-to-end lensless fiber imaging using deep neural networks. The well-trained resolution enhancement network not only recovers high-resolution features beyond the physical limitations, but also helps improving tumor recognition rate. This paradigm shift enables imaging with lensless fiber endoscopy and is promising especially for minimally invasive intraoperative treatment and cancer diagnosis in neurosurgery.