An artificial intelligence-based bone age assessment model for Han and Tibetan children

Abstract

<b>Background:</b> Manual bone age assessment (BAA) is associated with longer interpretation time and higher cost and variability, thus posing challenges in areas with restricted medical facilities, such as the high-altitude Tibetan Plateau. The application of artificial intelligence (AI) for automating BAA could facilitate resolving this issue. This study aimed to develop an AI-based BAA model for Han and Tibetan children. <b>Methods:</b> A model named "EVG-BANet" was trained using three datasets, including the Radiology Society of North America (RSNA) dataset (training set <i>n</i> = 12611, validation set <i>n</i> = 1425, and test set <i>n</i> = 200), the Radiological Hand Pose Estimation (RHPE) dataset (training set <i>n</i> = 5491, validation set <i>n</i> = 713, and test set <i>n</i> = 79), and a self-established local dataset [training set <i>n</i> = 825 and test set <i>n</i> = 351 (Han <i>n</i> = 216 and Tibetan <i>n</i> = 135)]. An open-access state-of-the-art model BoNet was used for comparison. The accuracy and generalizability of the two models were evaluated using the abovementioned three test sets and an external test set (<i>n</i> = 256, all were Tibetan). Mean absolute difference (MAD) and accuracy within 1 year were used as indicators. Bias was evaluated by comparing the MAD between the demographic groups. <b>Results:</b> EVG-BANet outperformed BoNet in the MAD on the RHPE test set (0.52 vs. 0.63 years, <i>p</i> < 0.001), the local test set (0.47 vs. 0.62 years, <i>p</i> < 0.001), and the external test set (0.53 vs. 0.66 years, <i>p</i> < 0.001) and exhibited a comparable MAD on the RSNA test set (0.34 vs. 0.35 years, <i>p</i> = 0.934). EVG-BANet achieved accuracy within 1 year of 97.7% on the local test set (BoNet 90%, <i>p</i> < 0.001) and 89.5% on the external test set (BoNet 85.5%, <i>p</i> = 0.066). EVG-BANet showed no bias in the local test set but exhibited a bias related to chronological age in the external test set. <b>Conclusion:</b> EVG-BANet can accurately predict the bone age (BA) for both Han children and Tibetan children living in the Tibetan Plateau with limited healthcare facilities.

Ähnliche Arbeiten