PURPOSE To develop and validate a pachymetry-based machine learning index for differentiating keratoconus, keratoconus suspect, and normal corneas. DESIGN Development and validation of a machine learning diagnostic algorithm. METHODS This retrospective study included 349 eyes of 349 patients with normal, frank keratoconus (KC), and keratoconus suspect (KCS) corneas. KCS corneas included topographically/tomographically normal (TNF) and borderline fellow eyes (TBF) of patients with asymmetric keratoconus. Six parameters were derived from the Corneal Thickness Progression map on the Galilei Dual Scheimpflug-Placido system and fed into a machine-learning algorithm to create the Thickness Speed Progression Index (TSPI). The model was trained with 5-fold cross-validation using a random search over 7 different machine learning algorithms, and the best model and hyperparameters were selected. RESULTS 133 normal eyes, 141 KC eyes, and 75 KCS eyes, subdivided into 34 TNF and 41 TBF eyes, were included. In experiment 1 (normal and KC), the best model (Random Forest) achieved an accuracy of 100% and AUROC of 1.00 for both normal and KC groups. In experiment 2 (normal, KCS, and KC), the model achieved an overall accuracy of 91%, and AUROC curves of 0.93, 0.83, and 0.99 in detecting normal, KCS, and KC corneas respectively. In experiment 3 (normal, TNF, TBF, and KC), the model achieved an accuracy of 87% with AUROC curves of 0.91, 0.60, 0.77, and 0.94 for normal, TNF, TBF, and KC corneas, respectively. CONCLUSION Using data solely based on pachymetry, machine learning algorithms such as the TSPI are able to discriminate normal corneas from keratoconus and keratoconus suspects corneas with reasonable accuracy.

中文翻译：

---

厚度速度进展指数 （TSPI）：圆锥角膜检测的机器学习方法：用于 KC 检测的基于厚度的机器学习模型。


目的 开发和验证基于厚实度的机器学习指数，用于区分圆锥角膜、疑似圆锥角膜和正常角膜。设计 机器学习诊断算法的开发和验证。方法 本回顾性研究包括 349 例正常、明显的圆锥角膜 （KC） 和疑似圆锥角膜 （KCS） 患者的 349 只眼睛。KCS 角膜包括不对称圆锥角膜患者的地形图/断层扫描正常 （TNF） 和交界对侧眼 （TBF）。六个参数来自 Galilei Dual Scheimpflug-Placido 系统上的角膜厚度进展图，并输入到机器学习算法中以创建厚度速度进展指数 （TSPI）。该模型使用对 7 种不同的机器学习算法进行随机搜索，通过 5 倍交叉验证进行训练，并选择了最佳模型和超参数。结果 共纳入 133 只正常眼、141 只 KC 眼和 75 只 KCS 眼，细分为 34 只 TNF 眼和 41 只 TBF 眼。在实验 1 （正常和 KC） 中，最佳模型 （随机森林） 在正常组和 KC 组中均达到 100% 的准确率和 1.00 的 AUROC。在实验 2 （正常、KCS 和 KC） 中，该模型在检测正常、KCS 和 KC 角膜方面实现了 91% 的总体准确率，AUROC 曲线分别为 0.93、0.83 和 0.99。在实验 3 （正常、TNF、TBF 和 KC） 中，该模型实现了 87% 的准确率，正常、TNF、TBF 和 KC 角膜的 AUROC 曲线分别为 0.91、0.60、0.77 和 0.94。结论 仅使用基于厚实测量的数据，TSPI 等机器学习算法能够以合理的准确性区分正常角膜和圆锥角膜和疑似圆锥角膜。