Department of Radiology, Mayo Clinic, 200 First Street SW, Charlton 1, Rochester, MN, 55905, USA.
Department of Radiology, Tata Medical Center, Kolkata, 700160, India.
Abdom Radiol (NY). 2022 Nov;47(11):3806-3816. doi: 10.1007/s00261-022-03668-1. Epub 2022 Sep 10.
To determine if pancreas radiomics-based AI model can detect the CT imaging signature of type 2 diabetes (T2D).
Total 107 radiomic features were extracted from volumetrically segmented normal pancreas in 422 T2D patients and 456 age-matched controls. Dataset was randomly split into training (300 T2D, 300 control CTs) and test subsets (122 T2D, 156 control CTs). An XGBoost model trained on 10 features selected through top-K-based selection method and optimized through threefold cross-validation on training subset was evaluated on test subset.
Model correctly classified 73 (60%) T2D patients and 96 (62%) controls yielding F1-score, sensitivity, specificity, precision, and AUC of 0.57, 0.62, 0.61, 0.55, and 0.65, respectively. Model's performance was equivalent across gender, CT slice thicknesses, and CT vendors (p values > 0.05). There was no difference between correctly classified versus misclassified patients in the mean (range) T2D duration [4.5 (0-15.4) versus 4.8 (0-15.7) years, p = 0.8], antidiabetic treatment [insulin (22% versus 18%), oral antidiabetics (10% versus 18%), both (41% versus 39%) (p > 0.05)], and treatment duration [5.4 (0-15) versus 5 (0-13) years, p = 0.4].
Pancreas radiomics-based AI model can detect the imaging signature of T2D. Further refinement and validation are needed to evaluate its potential for opportunistic T2D detection on millions of CTs that are performed annually.
确定基于胰腺放射组学的人工智能模型是否能够检测出 2 型糖尿病（T2D）的 CT 成像特征。
从 422 例 T2D 患者和 456 名年龄匹配的对照者的容积分割正常胰腺中提取了 107 个放射组学特征。数据集随机分为训练集（300 例 T2D、300 例对照 CT）和测试子集（122 例 T2D、156 例对照 CT）。通过基于 top-K 选择方法选择的 10 个特征和通过三折交叉验证在训练子集上进行优化的 XGBoost 模型在测试子集上进行评估。
该模型正确分类了 73 例（60%）T2D 患者和 96 例（62%）对照者，F1 评分、敏感性、特异性、精度和 AUC 分别为 0.57、0.62、0.61、0.55 和 0.65。模型在性别、CT 切片厚度和 CT 供应商方面的性能相当（p 值均>0.05）。在 T2D 持续时间[4.5（0-15.4）与 4.8（0-15.7）年，p=0.8]、抗糖尿病治疗[胰岛素（22%比 18%）、口服降糖药（10%比 18%）、两者（41%比 39%）（p>0.05）]和治疗持续时间[5.4（0-15）与 5（0-13）年，p=0.4]方面，正确分类与错误分类患者之间没有差异。
基于胰腺放射组学的人工智能模型可以检测出 T2D 的成像特征。需要进一步改进和验证，以评估其在每年进行数百万次 CT 检查时用于机会性 T2D 检测的潜力。
