我们病例-对照研究的目标是：1）开发一个自动化的三维卷积神经网络（CNN）用于诊断CT图像上胰腺导管腺瘤（PDA）的检测；2）在多机构的公共数据集上评估其普适性；3）作为一个潜在的筛查工具在高先验概率的模拟队列中的实用性；以及4）在预诊断CT图像上检测视觉隐藏的癌前病变的能力。 我们训练了一个三维CNN分类系统，使用算法生成的边界框和胰腺掩模，对一个精选数据集进行训练，该数据集包含696例含有PDA的门脉期诊断CT图像和1080例非肿瘤胰腺的对照病例。模型在以下方面进行了评估：（a）在院内保留的测试子集上（409例含有PDA的CT图像和829例对照图像）；（b）一个模拟队列，其中病例和对照的分布与糖尿病新发病例中PDA的风险和END-PAC评分≥3的患者类似；（c）多机构的公共数据集（194例含有PDA的CT图像和80例对照图像）；以及（d）一个包含100例预诊断CT图像（即在PDA临床诊断之前3-36个月偶然扫描获取的CT图像）且无局灶性肿块，以及134例对照图像。 院内测试子集中的大多数CT图像（n=798，64%）来自医院外部。该模型正确地对360例PDA的CT图像进行了分类（准确率为88%），对783例对照图像进行了分类（准确率为94%）[准确率（均值；95% CI）为0.92（0.91-0.94）；AUROC为0.97（0.96-0.98），敏感性为0.88（0.85-0.91），特异性为0.95（0.93-0.96）]。热图上的激活区域与大多数CT图像中的肿瘤重叠（350/360例CT图像，97%）。该模型的性能在不同肿瘤分期（敏感性分别为0.80、0.87、0.95和1.0，对应于T1至T4期）之间表现良好，对于低密度与等密度肿瘤的识别效果相当（敏感性：0.90 vs. 0.82），对于不同年龄、性别、CT切片厚度和供应商的CT图像也具有普适性（所有p >0.05）。模型在模拟队列[准确率0.95（0.94-0.95），AUROC 0.97（0.94-0.99）]和公共数据集[准确率0.86（0.82-0.90），AUROC 0.90（0.86-0.95）]上的表现也很好。尽管该模型仅在带有大肿瘤的诊断CT图像上进行了训练，但它能够在预诊断CT图像上检测到隐藏的PDA [准确率0.84（0.79-0.88），AUROC 0.91（0.86-0.94），敏感性0.75（0.67-0.84），特异性0.90（0.85-0.95）]，且在临床诊断前的中位数475天（范围：93-1082）。 该基于大规模多样化数据集训练的自动化人工智能模型在诊断CT图像上检测PDA以及预诊断CT图像上的隐藏PDA方面具有高准确率和普适性表现。需要进行基于血液生物标志物的前瞻性验证，以评估其在高风险人群中早期检测散发性PDA的潜力。 版权所有 © 2023 AGA Institute。Elsevier Inc.发表。保留所有权利。

The aims of our case-control study were - 1) to develop an automated 3D-Convolutional Neural Network (CNN) for detection of PDA on diagnostic CTs, 2) evaluate its generalizability on multi-institutional public datasets, 3) its utility as a potential screening tool using a simulated cohort with high pretest probability, and 4) its ability to detect visually occult pre-invasive cancer on pre-diagnostic CTs.A 3D-CNN classification system was trained using algorithmically generated bounding boxes and pancreatic masks on a curated dataset of 696 portal phase diagnostic CTs with PDA and 1080 controls with non-neoplastic pancreas. Model was evaluated on (a) an intramural hold-out test subset (409 CTs with PDA, 829 controls); (b) a simulated cohort with a case-control distribution that matched the risk of PDA in glycemically-defined new-onset diabetes and END-PAC score ≥3; (c) multi-institutional public datasets (194 CTs with PDA, 80 controls), and (d) a cohort of 100 pre-diagnostic CTs (i.e., CTs incidentally acquired 3-36 months before clinical diagnosis of PDA) without a focal mass, and 134 controls.Majority CTs (n=798; 64%) in intramural test subset were from outside hospitals. The model correctly classified 360 (88%) CTs with PDA and 783 (94%) controls [accuracy (mean; 95% CI) 0·92 (0·91-0·94); AUROC 0·97 (0·96-0·98), sensitivity 0·88 (0·85-0·91), specificity 0·95 (0·93-0·96)]. Activation areas on heat maps overlapped with the tumor in most CTs (350/360 CTs; 97%). Performance was high across tumor stages (sensitivity 0·80, 0·87, 0·95 and 1.0 on T1 through T4 stages, respectively), comparable for hypodense versus isodense tumors (sensitivity: 0·90 vs. 0·82), different age, sex, CT slice thicknesses & vendors (all p >0·05), and generalizable on both the simulated cohort [accuracy 0·95 (0·94-0·95), AUROC 0·97 (0·94-0·99)] and public datasets [accuracy 0·86 (0·82-0·90), AUROC 0·90 (0·86-0·95)]. Despite being exclusively trained on diagnostic CTs with larger tumors, the model could detect occult PDA on pre-diagnostic CTs [accuracy 0·84 (0·79-0·88), AUROC 0·91 (0·86-0·94), sensitivity 0·75 (0·67-0·84), specificity 0·90 (0·85-0·95)] at a median 475 days (range: 93-1082) prior to clinical diagnosis.Automated AI model trained on a large and diverse dataset shows high accuracy and generalizable performance for detection of PDA on diagnostic CTs as well as for visually occult PDA on pre-diagnostic CTs. Prospective validation with blood-based biomarkers is warranted to assess the potential for early detection of sporadic PDA in high-risk subjects.Copyright © 2023 AGA Institute. Published by Elsevier Inc. All rights reserved.