Abstract

HyperSpectral (HS) images have been successfully used for brain tumor boundary detection during resection operations. Nowadays, these classification maps coexist with other technologies such as MRI or IOUS that improve a neurosurgeon’s action, with their incorporation being a neurosurgeon’s task. The project in which this work is framed generates an unified and more accurate 3D immersive model using HS, MRI, and IOUS information. To do so, the HS images need to include 3D information and it needs to be generated in real-time operating room conditions, around a few seconds. This work presents Graph cuts Reference depth estimation in GPU (GoRG), a GPU-accelerated multiview depth estimation tool for HS images also able to process YUV images in less than 5.5 s on average. Compared to a high-quality SoA algorithm, MPEG DERS, GoRG YUV obtain quality losses of −0.93 dB, −0.6 dB, and −1.96% for WS-PSNR, IV-PSNR, and VMAF, respectively, using a video synthesis processing chain. For HS test images, GoRG obtains an average RMSE of 7.5 cm, with most of its errors in the background, needing around 850 ms to process one frame and view. These results demonstrate the feasibility of using GoRG during a tumor resection operation.

Highlights

  • The motion flag is a boolean image that represents the differences between the current frame and the previous one. It is used both in the homography and graph cuts stages to ensure temporal consistency and to speed-up the process, as the parts that did not change are not updated. It is calculated with an average difference user-parametrizable window that compares, using the L1 norm, the current frame with the previous one; if the value obtained is greater than a threshold, which is another parameter, the pixel is considered changed

  • They are compared to Depth Estimation Reference Software (DERS) in the same conditions to analyze what is the quality-time trade-off achieved by GoRG, considering DERS as the quality reference

  • These results show how GoRG depth maps are well estimated for the objects in the scene, preserving borders

Read more

Summary

Introduction

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. In the particular case of brain cancer detection, addressed in this paper, several works benefited from using HS images in the operation room [13,14,15] to precisely detect the borders of a tumor and to help neurosurgeons during the resection process All of these works aimed to improve both (i) the quality of the classification maps and the results of HS images, and (ii) the acceleration of the algorithms as much as possible, so the process can be executed in real time. YUV images at the cost of a huge processing time This tool was GPU accelerated to prove its viability in producing high-quality depth maps in an operating room at real time.

Related Works
Background
Read Cameras
Homography Cost
Graph Cuts
Interpolation
Reliability Map
Smoothing Map
Motion Flag
Proposal
Reliability and Smoothing Map
YUV Material
HS Material
Results
Kernel Performance and Optimization
Graph Cuts Filtering
Interpolation Analysis
GoRG Evaluation Results with YUV Images
GoRG Evaluation Results with HS Images
Conclusions
Future Lines
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call