Portal de Periódicos da CAPES

Pilot study of ultrasound parotid imaging reporting and data system (PIRADS): Inter-observer agreement

2015; Elsevier BV; Volume: 84; Issue: 12 Linguagem: Inglês

10.1016/j.ejrad.2015.09.001

1872-7727

Ahmed Abdel Khalek Abdel Razek, Germeen Albair Ashmalla, Gada Gaballa, Nadia Nada,

AI in cancer detection

To establish proposal ultrasound parotid imaging reporting and data system (PIRADS) for classification and prediction of malignancy of parotid lesions and to assess the inter-observer agreement of this system.Retrospective analysis of ultrasound and power Duplex images of 142 patients with parotid lesions by two reviewers. Parotid focal lesions were classified into nine patterns and then categorized into five groups: PIRADS 1, definitively benign; PIRADS 2, probably benign; PIRADS 3, indeterminate; PIRADS 4, probably malignant; and PIRADS 5, highly suggestive malignant.There was excellent interobserver agreement of both reviewers for patterns and PIRADS (K=0.84, P=0.001) with 92% percent agreement. There was excellent agreement of PIRADS 1 (K=1.00, P=0.001), PIRADS 2 (K=0.97, P=0.001), PIRADS 3 (K=0.86, P=0.001) and PIRADS 5 (K=0.88, P=0.001) and good agreement of PIRADS 4 (K=0.67, P=0.001). The Odds ratio of PIRADS 3, 4 and 5 were 1.36 (95% CI=0.39-4.55), 7.11 (95% CI=3.02-11.15) and 8.27 (95% CI=3.49-10.27) respectively. The accuracy was 92% and 90%, sensitivity was 79% and 65%, specificity was 94% and 96% of PIRADS of both reviewers respectively.The proposed PIRADS is a reliable non-invasive imaging modality that can be used for categorizing parotid lesions and prediction of malignancy.