Can you estimate body composition in dogs from photographs?

BMC Veterinary Research

Table 3 Results of single-observer validation studies

Method		vBCS_measured	vBCS_subjective	vBCS_adjusted
Versus BF% (R_s)		0.51	0.75	0.65
Versus BF% (R_s)		P < 0.001	P < 0.001	P < 0.001
Versus BCS _actual (Kappa)		0.51	0.63	0.63
Versus BCS _actual (Kappa)		P < 0.001	P < 0.001	P < 0.001
Overall scoring accuracy	Correct	83/125 (66 %)	35/51 (71 %)	90/125 (72 %)
	Under	19/125 (15 %)	13/51 (25 %)	16/125 (13 %)
	Over	23/125 (18 %)	2/51 (4 %)	19/125 (15 %)
Scoring accuracy (Ideal weight)	Correct	16/24 (67 %)	13/14 (93 %)	19/24 (79 %)
	Under	n/a	n/a	n/a
	Over	8/24 (33 %)	1/14 (7 %)	5/24 (21 %)
Scoring accuracy (Overweight)	Correct	19/37 (51 %)	10/20 (50 %)	19/37 (51 %)
	Under	3/37 (8 %)	9/20 (45 %)	4/37 (11 %)
	Over	15/37 (41 %)	1/20 (5 %)	14/37 (38 %)
Scoring accuracy (Obese)	Correct	48/64 (75 %)	13/17 (76 %)	52/64 (81 %)
	Under	16/64 (25 %)	4/17 (24 %)	12/64 (19 %)
	Over	n/a	n/a	n/a

The columns represent data for the three methods for visual body condition scoring (BCS) using photographs. vBCS_measured: BCS based upon abdominal width to thoracic width ratio measured from a dorsal photograph; vBCS_subjective: BCS semi-quantitatively assessed from a non-standardised photograph using visual descriptors; vBCS_adjusted: a refinement of vBCS_measured, whereby the A:T ratio was first used to estimate BS, but the observer could then modify after examining standardised dorsal and lateral photographs and applying visual BCS descriptors. Rows represent results of performance of each method determined using different parameters. BF% (R_S): correlation between vBCS method and body fat percentage using Spearman’s rank correlation; BCS_actual Kappa: agreement between vBCS method and the actual BCS (9-integer unit system [5] determined by a single observer); Correct BCS assigned: proportion and percentage of dogs correctly scored using each vBCS method and the actual BCS

ISSN: 1746-6148