Hsieh Chi-Wen, Liu Tzu-Chiang, Wang Jui-Kai, Jong Tai-Lang, Tiu Chui-Mei
Department of Electrical Engineering, National Chiayi University, Chiayi City, Taiwan.
Pediatr Int. 2011 Aug;53(4):567-75. doi: 10.1111/j.1442-200X.2011.03378.x.
The Tanner-Whitehouse III (TW3) method is popular for assessing children's bone age, but it is time-consuming in clinical settings; to simplify this, a grouped-TW algorithm (GTA) was developed.
A total of 534 left-hand roentgenograms of subjects aged 2-15 years, including 270 training and 264 testing datasets, were evaluated by a senior pediatrician. Next, GTA was used to choose the appropriate candidate of radius, ulna, and short bones and to classify the bones into three groups by data mining. Group 1 was composed of the maturity pattern of the radius and the middle phalange of the third and fifth digits and three weights were obtained by data mining, yielding a result similar to that of TW3. Subsequently, new bone-age assessment tables were constructed for boys and girls by linear regression and fuzzy logic. In addition, the Bland-Altman plot was utilized to compare accuracy between the GTA, the Greulich-Pyle (GP), and the TW3 method.
The relative accuracy between the GTA and the TW3 was 96.2% in boys and 95% in girls, with an error of 1 year, while that between the assessment results of the GP and TW3 was about 87%, with an error of 1 year. However, even if the three weights were not optimally processed, GTA yielded a marginal result with an accuracy of 78.2% in boys and 79.6% in girls.
GTA can efficiently simplify the complexity of the TW3 method, while maintaining almost the same accuracy. The relative accuracy between the assessment results of GTA and GP can also be marginal.
坦纳-怀特豪斯III（TW3）方法在评估儿童骨龄方面很受欢迎，但在临床环境中耗时较长；为简化此方法，开发了一种分组TW算法（GTA）。
由一位资深儿科医生评估了534张年龄在2至15岁受试者的左手X线片，包括270个训练数据集和264个测试数据集。接下来，使用GTA通过数据挖掘选择桡骨、尺骨和短骨的合适候选者，并将这些骨分为三组。第1组由桡骨以及第三和第五指中节指骨的成熟模式组成，通过数据挖掘获得三个权重，得出的结果与TW3相似。随后，通过线性回归和模糊逻辑为男孩和女孩构建了新的骨龄评估表。此外，利用布兰德-奥特曼图比较GTA、格鲁利希-派尔（GP）和TW3方法之间的准确性。
GTA与TW3之间在男孩中的相对准确率为96.2%，在女孩中为95%，误差为1年，而GP与TW3评估结果之间的相对准确率约为87%，误差为1年。然而，即使三个权重未得到最佳处理，GTA仍产生了边际结果，在男孩中的准确率为78.2%，在女孩中为79.6%。
GTA可以有效简化TW3方法的复杂性，同时保持几乎相同的准确性。GTA与GP评估结果之间的相对准确率也可能处于边际水平。
