HUS Medical Imaging Center, Radiology, University of Helsinki and Helsinki University Hospital, P.O. Box 340 (Haartmaninkatu 4), FI-00290, Helsinki, Finland.
Department of Physics, University of Helsinki, P.O. Box 64, FI-00014, Helsinki, Finland.
Eur Radiol Exp. 2021 Sep 24;5(1):45. doi: 10.1186/s41747-021-00235-z.
Chronic pulmonary embolism (CPE) is a life-threatening disease easily misdiagnosed on computed tomography. We investigated a three-dimensional convolutional neural network (CNN) algorithm for detecting hypoperfusion in CPE from computed tomography pulmonary angiography (CTPA).
Preoperative CTPA of 25 patients with CPE and 25 without pulmonary embolism were selected. We applied a 48%-12%-40% training-validation-testing split (12 positive and 12 negative CTPA volumes for training, 3 positives and 3 negatives for validation, 10 positives and 10 negatives for testing). The median number of axial images per CTPA was 335 (min-max, 111-570). Expert manual segmentations were used as training and testing targets. The CNN output was compared to a method in which a Hounsfield unit (HU) threshold was used to detect hypoperfusion. Receiver operating characteristic area under the curve (AUC) and Matthew correlation coefficient (MCC) were calculated with their 95% confidence interval (CI).
The predicted segmentations of CNN showed AUC 0.87 (95% CI 0.82-0.91), those of HU-threshold method 0.79 (95% CI 0.74-0.84). The optimal global threshold values were CNN output probability ≥ 0.37 and ≤ -850 HU. Using these values, MCC was 0.46 (95% CI 0.29-0.59) for CNN and 0.35 (95% CI 0.18-0.48) for HU-threshold method (average difference in MCC in the bootstrap samples 0.11 (95% CI 0.05-0.16). A high CNN prediction probability was a strong predictor of CPE.
We proposed a deep learning method for detecting hypoperfusion in CPE from CTPA. This model may help evaluating disease extent and supporting treatment planning.
慢性肺栓塞（CPE）是一种危及生命的疾病，在计算机断层扫描（CT）上容易误诊。我们研究了一种用于从 CT 肺动脉造影（CTPA）中检测 CPE 灌注不足的三维卷积神经网络（CNN）算法。
选择 25 例 CPE 患者和 25 例无肺栓塞患者的术前 CTPA。我们应用了 48%-12%-40%的训练-验证-测试分割（12 例阳性和 12 例阴性 CTPA 容积用于训练，3 例阳性和 3 例阴性用于验证，10 例阳性和 10 例阴性用于测试）。每例 CTPA 的轴向图像中位数为 335 张（最小-最大，111-570 张）。专家手动分割用于训练和测试目标。将 CNN 的输出与使用 CT 值（HU）阈值检测灌注不足的方法进行比较。计算了接收者操作特征曲线（AUC）和马修相关系数（MCC）及其 95%置信区间（CI）。
CNN 预测的分割 AUC 为 0.87（95%CI 0.82-0.91），HU 阈值法为 0.79（95%CI 0.74-0.84）。最佳全局阈值为 CNN 输出概率≥0.37 且≤-850 HU。使用这些值，CNN 的 MCC 为 0.46（95%CI 0.29-0.59），HU 阈值法为 0.35（95%CI 0.18-0.48）（Bootstrap 样本中 MCC 的平均差异为 0.11（95%CI 0.05-0.16））。CNN 高预测概率是 CPE 的一个强有力预测因子。
我们提出了一种用于从 CTPA 中检测 CPE 灌注不足的深度学习方法。该模型可能有助于评估疾病范围并支持治疗计划。
