A comparison of four different imaging modalities - Conventional, cross polarized, infra-red and ultra-violet in the assessment of childhood bruising

Background

It is standard practice to image concerning bruises in children. We aim to compare the clarity and measurements of bruises using cross polarized, infra-red (IR) and ultra-violet (UV) images to conventional images.

Methods

Children aged <11 years with incidental bruising were recruited. Demographics, skin and bruise details were recorded. Bruises were imaged by standard protocols in conventional, cross-polarized, IR and UV lights. Bruises were assessed in vivo for contrast, uniformity and diffuseness, and these characteristics were then compared across image modalities. Color images (conventional, cross polarized) were segmented and measured by ImageJ. Bruises of grey scale images (IR, UV) were measured by a ‘plug in’ of ImageJ. The maximum and minimum Feret's diameter, area and aspect ratio, were determined. Comparison of measurements across imaging modalities was conducted using Wilcoxon rank sum tests and modified Bland-Altman graphs. Significance was set at p < 0.05.

Results

Twenty five children had 39 bruises. Bruises that were of low contrast, i.e. difficult to distinguish from surrounding skin, were also more diffuse, and less uniformity in vivo. Low contrast bruises were best seen on conventional and cross-polarized images and less distinctive on IR and UV images. Of the 19 bruises visible in all modalities, the only significant difference was maximum and minimum Feret's diameters and area were smaller on IR compared to conventional images. Aspect ratios were not affected by the modality.

Conclusions

Conventional and cross-polarized imaging provides the most consistent bruise measurement, particularly in bruises that are not easily distinguished from surrounding skin visually.

•

Diffuse bruises may be measured on conventional & cross polarized imaging.

•

Infrared or ultraviolet imaging may not show bruises which are difficult to see in vivo.

•

Measurements of bruises using infrared imaging may be smaller than other modalities.

Optimising the measurement of bruises in children across conventional and cross polarized images using segmentation analysis techniques in Image J, Photoshop and circle diameter measurements

BACKGROUND

Bruising is a common abusive injury in children, and it is standard practice to image and measure them, yet there is no current standard for measuring bruise size consistently. We aim to identify the optimal method of measuring photographic images of bruises, including computerised measurement techniques.

METHODS

24 children aged <11 years (mean age of 6.9, range 2.5-10 years) with a bruise were recruited from the community. Demographics and bruise details were recorded. Each bruise was measured in vivo using a paper measuring tape. Standardised conventional and cross polarized digital images were obtained. The diameter of bruise images were measured by three computer aided measurement techniques: Image J (segmentation with Simple Interactive Object Extraction (maximum Feret diameter), 'Circular Selection Tool' (Circle diameter), & the Photoshop 'ruler' software (Photoshop diameter)). Inter and intra-observer effects were determined by two individuals repeating 11 electronic measurements, and relevant Intraclass Correlation Coefficient's (ICC's) were used to establish reliability. Spearman's rank correlation was used to compare in vivo with computerised measurements; a comparison of measurement techniques across imaging modalities was conducted using Kolmogorov-Smirnov tests. Significance was set at p < 0.05 for all tests.

RESULTS

Images were available for 38 bruises in vivo, with 48 bruises visible on cross polarized imaging and 46 on conventional imaging (some bruises interpreted as being single in vivo appeared to be multiple in digital images). Correlation coefficients were >0.5 for all techniques, with maximum Feret diameter and maximum Photoshop diameter on conventional images having the strongest correlation with in vivo measurements. There were significant differences between in vivo and computer-aided measurements, but none between different computer-aided measurement techniques. Overall, computer aided measurements appeared larger than in vivo. Inter- and intra-observer agreement was high for all maximum diameter measurements (ICC's > 0.7).

CONCLUSIONS

Whilst there are minimal differences between measurements of images obtained, the most consistent results were obtained when conventional images, segmented by Image J Software, were measured with a Feret diameter. This is therefore proposed as a standard for future research, and forensic practice, with the proviso that all computer aided measurements appear larger than in vivo.