Department of Biomedical Engineering, Asan Medical Institute of Convergence Science and Technology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, South Korea.
Department of Convergence Medicine, Asan Medical Institute of Convergence Science and Technology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, South Korea; Department of Computer Science and Engineering, Seoul National University, Seoul, South Korea.
Comput Biol Med. 2022 Sep;148:105914. doi: 10.1016/j.compbiomed.2022.105914. Epub 2022 Aug 7.
Landmark detection in flatfoot radiographs is crucial in analyzing foot deformity. Here, we evaluated the accuracy and efficiency of the automated identification of flatfoot landmarks using a newly developed cascade convolutional neural network (CNN) algorithm, Flatfoot Landmarks AnnoTating Network (FlatNet). A total of 1200 consecutive weight-bearing lateral radiographs of the foot were acquired. The first 1050 radiographs were used as the training and tuning, and the following 150 radiographs were used as the test sets, respectively. An expert orthopedic surgeon (A) manually labeled ground truths for twenty-five anatomical landmarks. Two orthopedic surgeons (A and B, each with eight years of clinical experience) and a general physician (GP) independently identified the landmarks of the test sets using the same method. After two weeks, observers B and GP independently identified the landmarks once again using the developed deep learning CNN model (DLm). The X- and Y-coordinates and the mean absolute distance were evaluated. The average differences (mm) from the ground truth were 0.60 ± 0.57, 1.37 ± 1.28, and 1.05 ± 1.23 for the X-coordinate, and 0.46 ± 0.59, 0.97 ± 0.98, and 0.73 ± 0.90 for the Y-coordinate in DLm, B, and GP, respectively. The average differences (mm) from the ground truth were 0.84 ± 0.73, 1.90 ± 1.34, and 1.42 ± 1.40 for the absolute distance in DLm, B, and GP, respectively. Under the guidance of the DLm, the overall differences (mm) from the ground truth were enhanced to 0.87 ± 1.21, 0.69 ± 0.74, and 1.24 ± 1.31 for the X-coordinate, Y-coordinate, and absolute distance, respectively, for observer B. The differences were also enhanced to 0.74 ± 0.73, 0.57 ± 0.63, and 1.04 ± 0.85 for observer GP. The newly developed FlatNet exhibited better accuracy and reliability than the observers. Furthermore, under the FlatNet guidance, the accuracy and reliability of the human observers generally improved.
在扁平足 X 光片中进行标志点检测对于分析足部畸形至关重要。在这里，我们评估了一种新开发的级联卷积神经网络（CNN）算法——扁平足标志点标注网络（FlatNet）自动识别扁平足标志点的准确性和效率。共获取了 1200 例连续的负重侧足部 X 光片。前 1050 张 X 光片用于训练和调整，后 150 张 X 光片用于测试集。一位骨科专家（A）手动标记了 25 个解剖标志点的真实位置。两位骨科医生（A 和 B，均具有 8 年临床经验）和一位全科医生（GP）分别使用相同的方法独立识别测试集的标志点。两周后，观察者 B 和 GP 再次使用开发的深度学习 CNN 模型（DLm）独立识别标志点。评估了 X、Y 坐标和平均绝对距离。从真实位置的平均差异（mm）为 X 坐标 0.60±0.57、1.37±1.28 和 1.05±1.23，Y 坐标 0.46±0.59、0.97±0.98 和 0.73±0.90，在 DLm、B 和 GP 中分别为 0.46±0.59、0.97±0.98 和 0.73±0.90。从真实位置的平均差异（mm）为绝对距离 0.84±0.73、1.90±1.34 和 1.42±1.40，在 DLm、B 和 GP 中分别为 0.84±0.73、1.90±1.34 和 1.42±1.40。在 DLm 的指导下，观察者 B 的 X 坐标、Y 坐标和绝对距离的总体差异（mm）分别提高到 0.87±1.21、0.69±0.74 和 1.24±1.31。对于观察者 GP，差异也提高到 0.74±0.73、0.57±0.63 和 1.04±0.85。新开发的 FlatNet 比观察者具有更高的准确性和可靠性。此外，在 FlatNet 的指导下，人类观察者的准确性和可靠性通常会提高。
