Documentation of all gross specimen examinations includes the specimen’s dimensions, which is manually measured by the pathologist’s assistant or the pathologist, typically with a ruler or invariably by estimation. These dimensions are then typically dictated, via transcriptionist or automatic speech recognition, to be transcribed into the specimen gross description. Sources of error in these dimension values are present at the point of measurement, especially if the object lack definable landmarks or is an irregular shape, and in the transcription. Our team proposes a low-cost video image processing system to measure the three dimensional lengths of an object using three off-the-shelf high resolution webcams and custom software using open-source image processing software libraries. Real-time imaging process to determine an object’s shape and size is a mature technology but its use is mostly limited to manufacturing and parcel routing applications. Our proposed 3D object measurement system uses the same image processing software tools to reconstruct an objects three dimension lengths using images from three orthogonally oriented web cameras. These cameras are fixed and oriented towards the region of measurement, which is initially calibrated with a solid object of known dimension.

Hardware: three Logitech HD C310 webcams, desktop Windows computer
Software: Python for production product, prototyping in Matlab with Computer Vision Toolbox

In operation, the user places the specimen in the measurement field after an initial calibration. The three images are preprocessed (edge detection, low-band pass), and image data fused to estimate the objection dimensional lengths in the camera coordinate frame. The object is then rotated into the coordinate frame defined by its two longest orthogonal axes. The real-time display of this dimension estimate is updated at 5Hz. An Autohotkey shortcut triggered by a foot pedal or voice command (if using Dragon Dictaion or pastes the text of the current dimensions at the cursor location.

The advantages would be the system’s accuracy of measurement over manual methods and a modest improvement in workflow speed.

A future extension of this concept would bring real-time 3D object dimensions measurement to the consumer, using multiple bluetooth-networked mobile phone cameras. The cameras would be orientation-fixed prior to calibration with a standard known-size object.

ROBOTIC TECHNOLOGY & VIRTUAL ASSISTANTS

Author: Thomas Chong

Coauthor(s): None

Status: Project Concept

Funding Acknowledgment: Hardware and software are already available.