RESUMO
Vertebral compression fractures (VCFs) are common and indicate a high future risk of additional osteoporotic fractures. However, many VCFs are unreported by radiologists, and even if reported, many patients do not receive treatment. The purpose of the study was to evaluate a new artificial intelligence (AI) algorithm for the detection of VCFs and to assess the prevalence of reported and unreported VCFs. This retrospective cohort study included patients over age 60 yr with an abdominal CT between January 18, 2019 and January 18, 2020. Images and radiology reports were reviewed to identify reported and unreported VCFs, and the images were processed by an AI algorithm. For reported VCFs, the electronic health records were reviewed regarding subsequent osteoporosis screening and treatment. Totally, 1112 patients were included. Of these, 187 patients (16.8%) had a VCF, of which 62 had an incident VCF and 49 had a previously unknown prevalent VCF. The radiologist reporting rate of these VCFs was 30% (33/111). For moderate and severe (grade 2-3) VCF, the AI algorithm had 85.2% sensitivity, 92.3% specificity, 57.8% positive predictive value, and 98.1% negative predictive value. Three of 30 patients with reported VCFs started osteoporosis treatment within a year. The AI algorithm had high accuracy for the detection of VCFs and could be very useful in increasing the detection rate of VCFs, as there was a substantial underdiagnosis of VCFs. However, as undertreatment in reported cases was substantial, to fully realize the potential of AI, changes to the management pathway outside of the radiology department are imperative.
Vertebral compression fractures (VCFs) are the most common osteoporotic fractures. However, they often go undetected leading to a high risk of further fractures. In this study, we tested a new artificial intelligence (AI) algorithm to detect VCFs in abdominal CT scans in patients over 60 yr of age and assessed how often VCFs were missed by radiologists. We found that VCFs were underreported, with only 30% being identified by the radiologists. The AI algorithm showed promising results and had high accuracy for detecting VCFs. However, many patients with a detected VCF still did not receive treatment. The results suggest that AI could increase the detection rate of VCFs, but also highlight the need for changes beyond radiology to ensure that patients with detected fractures are appropriately treated.
Assuntos
Algoritmos , Inteligência Artificial , Fraturas por Compressão , Fraturas da Coluna Vertebral , Tomografia Computadorizada por Raios X , Humanos , Feminino , Masculino , Fraturas por Compressão/diagnóstico por imagem , Fraturas por Compressão/epidemiologia , Idoso , Fraturas da Coluna Vertebral/diagnóstico por imagem , Fraturas da Coluna Vertebral/epidemiologia , Pessoa de Meia-Idade , Estudos Retrospectivos , Idoso de 80 Anos ou mais , SubtratamentoRESUMO
Objectives: Computed tomography pulmonary angiography (CTPA) is the gold standard diagnostic method for patients with suspected pulmonary embolism (PE), but it has its drawbacks, including exposure to ionizing radiation and iodinated contrast agent. The present study aims to evaluate the diagnostic performance of our in-house developed non-contrast MRI protocol for PE diagnosis in reference to CTPA. Methods: 107 patients were included, all of whom underwent MRI immediately before or within 36â¯hours after CTPA. Additional cases examined only with MRI and a negative result were added to reach a PE prevalence of approximately 20%. The protocol was a non-contrast 2D steady-state free precession (SSFP) sequence under free-breathing, without respiratory or cardiac gating, and repeated five times to capture the vessels at different breathing/cardiac phases. The MRIs were blinded and read by two radiologists and the results were compared to CTPA. Results: Of the 243 patients included, 47 were positive for PE. Readers 1 and 2 demonstrated 89% and 87% sensitivity, 100% specificity, 98% accuracy and Cohen's kappa of 0.88 on patient level. In the per embolus comparison, readers 1 and 2 detected, 60 and 59/61 (98, 97%) proximal, 101 and 94/113 (89, 83%) segmental, and 5 and 2/32 (16, 6%) subsegmental emboli, resulting in 81 and 75% sensitivity respectively. Conclusion: The repeated 2D SSFP can reliably be used for the diagnosis of acute PE at the proximal and segmental artery levels.