Endoscopic detection of cancer with lensless radioluminescence imaging and machine vision

Abstract

Complete removal of residual tumor tissue during surgical resection improves patient outcomes. However, it is often difficult for surgeons to delineate the tumor beyond its visible boundary. This has led to the development of intraoperative detectors that can image radiotracers accumulated within tumors, thus facilitating the removal of residual tumor tissue during surgical procedures. We introduce a beta imaging system that converts the beta radiation from the radiotracer into photons close to the decay origin through a CdWO4 scintillator and does not use any optical elements. The signal is relayed onto an EMCCD chip through a wound imaging fiber. The sensitivity of the device allows imaging of activity down to 100 nCi and the system has a resolution of at least 500 μm with a field of view of 4.80 × 6.51 mm. Advances in handheld beta cameras have focused on hardware improvements, but we apply machine vision to the recorded images to extract more information. We automatically classify sample regions in human renal cancer tissue ex-vivo into tumor or benign tissue based on image features. Machine vision boosts the ability of our system to distinguish tumor from healthy tissue by a factor of 9 ± 3 and can be applied to other beta imaging probes.

[18F]fluoroethyltyrosine-induced Cerenkov Luminescence Improves Image-Guided Surgical Resection of Glioma

“…These problems can be addressed for glioma resection through several strategies including the use of neuroendoscopy and probes for detecting β particles, which would allow real-time “lights on” surgical imaging, and by the use of higher energy positron-emitting radionuclides such as 11 C-methionine. This shows similar glioma targeting via LAT1 as FET, but produces over a five times greater photon yield than 18 F for CLI 21 , 51 - 54 . FET-induced CLI would also complement emerging interventional PET imaging, allowing intraoperative CLI of residual tumor in the surgical field and tomographic imaging of residual tumor in the whole brain, respectively 55 .…”

A Deep Learning Model for Striae Identification in End Images of Float Glass

“…However, a small loss in some data does not necessarily mean the loss in all data is small. As the result, the SGD algorithm sometimes cannot converge to the global optimal solution [29]. Considering the memory limit of computer and experimental results, the authors decided to contain 32 images in each batch of training samples, aiming to clarify the direction of gradient descent, control training oscillations and reduce the number of iterations.…”

Real-time cancer detection with an integrated lensless fluorescence contact imager

“…Goggle-based approaches [30,31] are easier to use, but suffer from low sensitivity due to the distance from the tumor bed and cannot image the complex tumor cavity sidewalls. A variety of handheld devices for residual cancer detection have been designed that use optical coherence tomography (OCT) [32], spectroscopy [33][34][35], birefringence [36], magnetic tracers [37], radio tracers [38], or fluorescence [39]. These are either too large to fit inside the resection cavity itself [32] or only image single-points [33][34][35][36][37].…”

“…These are either too large to fit inside the resection cavity itself [32] or only image single-points [33][34][35][36][37]. Techniques using fiber optic bundles [38,39] suffer from a fundamental trade-off of flexibility versus imaging area, prohibiting rapid imaging of the entire tumor bed surface in a small, complex-shaped, resection cavity.…”