High-resolution ultrasound reflex transmission imaging and digital photography: potential tools for the quantitative assessment of pigmented lesions.

Background/aims: High-resolution ultrasound (HRU) is a relatively cheap imaging method that shows small quantitative differences between benign naevi and melanoma. Previous studies using B-mode display suggest that these arise from their differing attenuating properties. Attenuation characteristics, however, are better evaluated using reflex transmission imaging (RTI). White light clinical (WLC) photography is an even cheaper imaging method that is routinely used for monitoring but less frequently in everyday diagnosis. As features from each method may have an independent origin, two such modalities may be of greater diagnostic value than either method alone. However, although quantitative analysis of digital photographs is being developed to aid tumour diagnosis, in vivo RTI for the evaluation of pigmented skin lesions has not previously been described. This paper presents the feasibility of performing RTI in vivo and evaluates the reliability of the objective features used. The potential of the combination of quantitative RTI and white light (WL) digital photography data for the classification of pigmented lesions was assessed.

Methods: Randomly selected patients were recruited via a skin cancer screening clinic. RTI data were acquired from each index lesion with a 20 MHz single-element scanner. WL images were taken using a high-resolution (2.8 Mpixels) digital camera. Quantitative features calculated from both images were used to derive a discriminant rule. This equation was then applied to reclassify each case based on its quantitative criteria. The resultant classification was compared with histological diagnosis.

Results: Twenty-four lesions (10 melanoma and 14 naevi) were studied. On RTI, no subjective differences were observed between benign naevi and melanoma. Many lesions were either not visible on RTI or lacked clearly definable borders. Consequently, the WL photographs were used to draw lesion boundaries on RT images for feature calculation. Melanoma were less attenuating than naevi on RTI (P=0.026) and had greater red colour variegation on WL imaging (P=0.016). The combination of quantitative parameters (two from RTI and four from photographs) improved sensitivity for this sample without compromising the specificity of 100% compared with either modality alone. The procedure is highly reproducible (r=0.85 between two operators).

Conclusions: Pigmented skin lesions can be quantitatively defined from RTI data acquired in vivo and a significant difference in attenuation is shown. However, accurate registration of the RT image with a corresponding photograph was crucial for this purpose and only possible when corresponding points could be reliably identified on both images. Combination of features from ultrasound and optical images may synergistically improve diagnostic accuracy and a larger study is warranted to investigate this.