Semi-quantitative scoring of imaging modalities in detecting soft tissue foreign bodies: an in vitro study.

Abstract

Accurate identification of foreign bodies (FB) using medical imaging is essential for diagnosis and determining the suitable retrieval technique. To compare the sensitivity of different imaging modalities for detecting various FB materials in soft tissue and assess the reproducibility of a scoring system for grading the conspicuity of FBs. Five FB materials (plastic, wood, glass, aluminum, and copper) were embedded in a tissue-mimicking phantom. Computed radiography (CR), ultrasound, computed tomography (CT), and magnetic resonance imaging (MRI) were compared using a semi-quantitative 5-point Likert scale scoring system. The intra- and inter-reader reproducibility of four independent readers was analyzed using Kendall's coefficient of concordance (W). Glass was visible on all imaging modalities. Plastic was only visible in excellent detail using ultrasound. Wood was detected in excellent resolution using ultrasound and CT using the default window while plain X-ray failed to detect it. Ultrasound was the only modality that showed aluminum in excellent quality while CT showed it with good demarcation from the surroundings. Copper was detectable in excellent detail using CR, ultrasound, and CT. MRI performance was suboptimal, especially with the plastic FB. The scoring system showed excellent intra-reader (W = 0.91, P = 0.001) and inter-reader (W = 0.88, P < 0.001) reproducibility. Ultrasound can be used as the first line of investigation for wood, plastic, glass, and metallic FBs impacted at superficial depths in soft tissue. The semi-quantitative FB scoring system showed excellent within- and between-reader reliability, which can be used to score and compare the detection performance of new imaging techniques.