Edwards Lindsay A, Yang Christina, Sharma Surbhi, Chen Zih-Hua, Gorantla Lahari, Joshi Sanika A, Longhi Nicolas J, Worku Nahom, Yang Jamie S, Martinez Di Pietro Brandy, Armenian Saro, Bhat Aarti, Border William, Buddhe Sujatha, Blythe Nancy, Stratton Kayla, Leger Kasey J, Leisenring Wendy M, Meacham Lillian R, Nathan Paul C, Narasimhan Shanti, Sachdeva Ritu, Sadak Karim, Chow Eric J, Boyle Patrick M
Department of Pediatrics, Division of Cardiology, Duke University Medical Center, DUMC Box 3090, Durham, NC, 27710, USA.
Department of Pediatrics, University of Washington, Seattle, WA, USA.
Cardiooncology. 2024 Oct 9;10(1):66. doi: 10.1186/s40959-024-00268-4.
Despite routine echocardiographic surveillance for childhood cancer survivors, the ability to predict cardiomyopathy risk in individual patients is limited. We explored the feasibility and optimal processes for machine learning-enhanced cardiomyopathy prediction in survivors using serial echocardiograms from five centers.
We designed a series of deep convolutional neural networks (DCNNs) for prediction of cardiomyopathy (shortening fraction ≤ 28% or ejection fraction ≤ 50% on two occasions) for at-risk survivors ≥ 1-year post initial cancer therapy. We built DCNNs with four subsets of echocardiographic data differing in timing relative to case (survivor who developed cardiomyopathy) index diagnosis and two input formats (montages) with differing image selections. We used holdout subsets in a 10-fold cross-validation framework and standard metrics to assess model performance (e.g., F1-score, area under the precision-recall curve [AUPRC]). Performance of the input formats was compared using a combined 5 × 2 cross-validation F-test.
The dataset included 542 pairs of montages: 171 montage pairs from 45 cases at time of cardiomyopathy diagnosis or pre-diagnosis and 371 pairs from 70 at-risk survivors who didn't develop cardiomyopathy during follow-up (non-case). The DCNN trained to distinguish between non-case and time of cardiomyopathy diagnosis or pre-diagnosis case montages achieved an AUROC of 0.89 ± 0.02, AUPRC 0.83 ± 0.03, and F1-score: 0.76 ± 0.04. When limited to smaller subsets of case data (e.g., ≥ 1 or 2 years pre-diagnosis), performance worsened. Model input format did not impact performance accuracy across models.
This methodology is a promising first step toward development of a DCNN capable of accurately differentiating pre-diagnosis versus non-case echocardiograms to predict survivors more likely to develop cardiomyopathy.
尽管对儿童癌症幸存者进行了常规超声心动图监测，但预测个体患者患心肌病风险的能力有限。我们利用来自五个中心的系列超声心动图，探讨了机器学习增强的心肌病预测在幸存者中的可行性和最佳流程。
我们设计了一系列深度卷积神经网络（DCNN），用于预测初始癌症治疗后≥1年的高危幸存者患心肌病的情况（缩短分数≤28%或射血分数≤50%，两次测量结果）。我们构建了具有四个超声心动图数据子集的DCNN，这些子集相对于病例（患心肌病的幸存者）索引诊断的时间不同，以及两种具有不同图像选择的输入格式（蒙太奇）。我们在10折交叉验证框架中使用留出子集和标准指标来评估模型性能（例如，F1分数、精确召回曲线下面积[AUPRC]）。使用组合的5×2交叉验证F检验比较输入格式的性能。
数据集包括542对蒙太奇：171对蒙太奇来自45例心肌病诊断时或诊断前的病例，371对来自70例在随访期间未患心肌病的高危幸存者（非病例）。训练用于区分非病例与心肌病诊断时或诊断前病例蒙太奇的DCNN的曲线下面积（AUROC）为0.89±0.02，AUPRC为0.83±0.03，F1分数为0.76±0.04。当限于病例数据的较小子集时（例如，诊断前≥1或2年），性能会变差。模型输入格式不会影响各模型的性能准确性。
这种方法是朝着开发一种能够准确区分诊断前与非病例超声心动图以预测更可能患心肌病的幸存者方向迈出的有希望的第一步。
