Comparison of Processed and Unprocessed Digital Images for the Determination of Radiographic Bone Aluminum Equivalent (RBAE) Values in Equine Bone

Joe D. Pagan¹, Peter Kazakevicius², Alyx Swanhall¹, Emma Ford¹, and Abby Pritchard¹

¹Kentucky Equine Research, Versailles, Kentucky, USA; ² Ocala, Florida, USA

Radiographic photodensitometry to estimate bone density in horses was originally developed using conventional analog film radiography. Today, most researchers have switched to digital radiography (DR) to make these measurements. One important feature of DR is that the raw data is processed after acquisition. Processing algorithms are generally proprietary and specific to the DR vendor, but they encompass manipulations such as adjustment of contrast curves and application of nonlinear image filters to optimize image quality parameters such as contrast and noise. These processing steps have a profound effect on the final appearance of the radiograph, and they can also lead to artifacts unique to digital systems that may affect density measurements. This study measured radiographic bone aluminum equivalent (RBAE) values in the left third metacarpal (MCIII) of 16 Thoroughbred horses (3.2 ± 0.4 y) using raw (unprocessed) and processed images acquired using a portable X-ray unit and read by a DR system (Sound NEXT Equine DR). Radiographs of the dorsal-palmar and lateral-medial views of MCIII were taken. An aluminum step wedge placed in the same plane as the MCIII was exposed simultaneously to calculate RBAE values. Unprocessed and processed images were analyzed using ImageJ image-processing software. An optical density calibration curve was created for each image using the known thicknesses of the step wedge. This relationship was best described using a natural log equation in the unprocessed images and a 4th degree polynomial equation in the processed images. Image processing had a large, inconsistent effect on RBAE values. Peak densities were 32-68% higher in the processed images and many were above the density of the thickest step of the step wedge. In lower density areas, the processed images had lower densities than the unprocessed images, demonstrating the nonlinear nature of the processing algorithm. RBAE values obtained from unprocessed images were similar to values measured in previous KER studies using analog film radiology. Raw (unprocessed) images should be used with DR systems to determine RBAE in equine bone.

Radiographic bone aluminum equivalents (RBAE) (mm Al) from processed and unprocessed DR images.

	Dorsal-palmar view			Lateral-medial view
	peak medial	peak lateral	mean medullary	peak dorsal	peak palmar	mean medullary
unprocessed	27.6 ± 0.7a	25.7 ± 0.2a	24.0 ± 0.4a	23.7 ± 0.2a	21.2 ± 0.2a	18.3 ± 0.2a
processed	36.5 ± 0.8b	36.5 ± 0.2b	20.2 ± 0.7b	35.6 ± 0.1b	35.6 ± 0.3b	14.1 ± 0.2b
% difference	32%	42%	-16%	50%	68%	-23%
a,b values in the same column with different superscripts are different (p<.05)