Performance evaluation of 2D sensors for medical navigation camera systems

Abstract:

Abstract

Stryker’s FP6000 is the currently most accurate medical navigation camera on the market.

This high-performance system consists of three one-dimensional sensors. An evaluation

is presented, that analyzes, if it is possible to design a camera based on 2D sensors, that

satisfies the same accuracy as the FP6000 camera, but is designed of smaller dimensions,

reduced weight and lower cost. For the LED localization a glass-substrate with a certain

pattern is placed in front of the 2D sensor. Within the glass-material there are always

local inhomogeneities present, that influence the measurements. One hypothesis is, that

the influence of those inhomogeneities can be lowered by increasing the number of pattern

elements. In this way, the accuracy of the 2D sensors may be improved up to a certain level.

Furthermore, a medical navigation camera, that is designed with multiple 2D sensors,

represents an overdetermined system. It is analyzed, how strongly the RMS error for

a camera based on multiple 2D sensors can be decreased by exploiting the concept of

system over-determination. On a test bench the influence of different pattern elements

on the accuracy of various 2D sensors was evaluated. A Matlab code simulated the RMS

error of a camera system, that localizes LED positions with the tested 2D sensors. In this

implementation, overdetermined problems were solved through a combination of linear

triangulation and bundle adjustment. The accuracy tests confirmed, that the part of the

systematic error, that results from local inhomogeneities within the glass-material, could

be diminished with 30 holes inside the pattern. As a result of this, the accuracy of the 2D

sensors was improved. However, none of the tested sensors was accurate enough for the

design of a camera of small dimensions and low cost, that localizes LED positions with

the same accuracy than the FP6000.