1.AO_935_CAM_Apport d’un modèle de compte rendu structuré LI-RADS sur la qualité des rapports diagnostiques du CHC en TDM

Aim: The objective of this study was to evaluate the concordance between LI-RADS categories assigned in free-text reports and those derived from the structured LI-RADS template, and to assess their impact on the diagnostic accuracy for hepatocellular carcinoma (HCC) on computed tomography. Methods: We carried out a cross-sectional study from November 1, 2021 to August 31, 2022 at the Yaoundé General Hospital and at the Cathedral Medical Center. We reviewed abdominal and thoraco-abdomino-pelvic CT reports indicated for suspicion of HCC. Sociodemographic, biological and computed tomography characteristics were recorded. For each liver lesion detected, we noted the following elements: the number, the unequivocal description of the major LI-RADS criteria, the unequivocal diagnosis of HCC, the mention or not of a LI-RADS category and the concordance of the categories between free-text reports and reports structured according to the LI-RADS model, after proofreading by 2 radiologists. Cohen's kappa coefficient was used to quantify the concordance between the two types of report. Results: We enrolled 76 reports from 76 patients, i.e. 45 men and 31 women with a male/female sex ratio of 1.4. The mean age of the patients was 54.1 ±17.2 years. Chronic viral B liver disease (53.9%) was the main reported risk factor. Only 33 of the 76 CT examinations (43.4%) were performed using an optimal acquisition protocol. Twenty (20) different terms were used in the conclusion sections, leading to diagnostic ambiguity. LI-RADS categorization was almost not used in free text reports (6.6%). After proofreading and recoding, there was poor agreement (k = 17.5%) between the free-text reports and the LI-RADS structured model. Conclusion: HCC occurs more frequently in adult male patients, with chronic viral B liver disease as the main risk factor. Agreement between free-text reports and structured LI-RADS reports is poor. The use of unstructured reports potentially increases the risk of diagnostic error by ambiguity unlike those structured LI-RADS.