Helping radiologists identify structural sacroiliac joint lesions in CT scans acquired for other clinical indications.

RFMaple is an AI-powered add-on to the PACS that assists radiologists in detecting the structural bone lesion types erosion and ankylosis in the sacroiliac joint region, as incidental findings in CT examinations performed for other diagnostic questions. These lesions may be associated with inflammatory or non-inflammatory arthropathies.

RFMaple uses deep learning to analyse CT images for structural bone lesions. In a standalone performance study on 2,247 analysed anonymised CT studies, RFMaple achieved a patient-level sensitivity of 89% (95% CI 86–92%) and a specificity of 94% (95% CI 92–95%) for the detection of structural sacroiliac joint lesions, against a reference standard established by expert radiologist readers. Lesion-specific performance was 81% sensitivity and 95% specificity for erosion, and 90% sensitivity and 98% specificity for ankylosis.

RFMaple is designed as a cloud-native add-on to any type of PACS, to be offered as software-as-a-service on a pay-per-use model. Performance has been assessed in a standalone performance study and a retrospective reader comparison study; results are summarised in the clinical evaluation. The software is designed to support customers’ obligations under the GDPR, and to support HIPAA-covered entities in the United States.