Optimal threshold for the diagnosis of anemia severity on unenhanced thoracic CT: A preliminary study

Objective To investigate and evaluate the value as accuracy of diagnosis different degrees of anemia through the unenhanced thoracic computed tomography (CT) scan. Materials and methods The CT attenuation of right ventricle (RV) cavity, left ventricular (LV) cavity, interventricular septum and difference between the interventricular septum and left ventricle cavity (IVS-LV) were retrospectively analyzed and measured in 317 patients with different degrees of anemia and the normal patients. The hemoglobin (Hb) level was estimated within 24 h. Statistical analysis was made to obtain the best diagnostic threshold of anemia levels by CT scan, measurement repeatability was assessed using the intra-class correlation coefficient (ICC) values. Results An obvious parallel correlation of hemoglobin concentration and CT attenuation of IVS-LV existed (the determination coefficient was 0.818; P < .001). Based on receiver operating characteristic (ROC) curves, the conclusion was that when the threshold of CT attenuation of IVS-LV at (5.5–9.5) HU in male and (4.5–8.5) HU in female, the specificity for diagnosing mild anemia was low, while the sensitivity was very high (87.1%, 100%, respectively), if the CT value was among (9.5–13.5) HU in male or (8.5–13.5) HU in female, the sensitivity and specificity for diagnosing moderate anemia were fine (92.9% and 74.8% in male; 93.9% and 60.0% in female), and once the CT value was more than 13.5 HU the sensitivity and specificity for diagnosing severe anemia in all gender were greatly good (94.7% and 83.6% in male; 82.4% and 84.6% in female). The intra-and inter-observer reliability and reproducibility were good (ICC were all more than 0.99). Conclusion The CT attenuation of IVS-LV could predict the severity of anemia with good sensitivity and specificity, which could add value to clinical practice and more importantly facilitating patient care as hematologic laboratory investigations are lacking.