人工智能系统在筛查性乳房 X 光摄影中检测间期乳腺癌的准确性。
Accuracy of an Artificial Intelligence System for Interval Breast Cancer Detection at Screening Mammography.
机构信息
From the Department of Radiology, School of Clinical Medicine, University of Cambridge, Box 218, Level 5, Cambridge Biomedical Campus, Cambridge CB2 0QQ, England (M.N., V.O.G., S.E.H., I.A., N.R.P., O.A., Y.H., A.N.P., F.J.G.); Department of Radiology, Addenbrooke's Hospital, Cambridge University Hospitals NHS Foundation Trust, Cambridge, England (M.N., I.A., R.B., A.N.P., F.J.G.); and Department of Radiology, The Royal London Hospital, Barts Health NHS Trust, London, England (S.E.H.).
出版信息
Radiology. 2024 Aug;312(2):e232303. doi: 10.1148/radiol.232303.
Background Artificial intelligence (AI) systems can be used to identify interval breast cancers, although the localizations are not always accurate. Purpose To evaluate AI localizations of interval cancers (ICs) on screening mammograms by IC category and histopathologic characteristics. Materials and Methods A screening mammography data set (median patient age, 57 years [IQR, 52-64 years]) that had been assessed by two human readers from January 2011 to December 2018 was retrospectively analyzed using a commercial AI system. The AI outputs were lesion locations (heatmaps) and the highest per-lesion risk score (range, 0-100) assigned to each case. AI heatmaps were considered false positive (FP) if they occurred on normal screening mammograms or on IC screening mammograms (ie, in patients subsequently diagnosed with IC) but outside the cancer boundary. A panel of consultant radiology experts classified ICs as normal or benign (true negative [TN]), uncertain (minimal signs of malignancy [MS]), or suspicious (false negative [FN]). Several specificity and sensitivity thresholds were applied. Mann-Whitney tests, Kruskal-Wallis tests, and χ tests were used to compare groups. Results A total of 2052 screening mammograms (514 ICs and 1548 normal mammograms) were included. The median AI risk score was 50 (IQR, 32-82) for TN ICs, 76 (IQR, 41-90) for ICs with MS, and 89 (IQR, 81-95) for FN ICs ( = .005). Higher median AI scores were observed for invasive tumors (62 [IQR, 39-88]) than for noninvasive tumors (33 [IQR, 20-55]; < .01) and for high-grade (grade 2-3) tumors (62 [IQR, 40-87]) than for low-grade (grade 0-1) tumors (45 [IQR, 26-81]; = .02). At the 96% specificity threshold, the AI algorithm flagged 121 of 514 (23.5%) ICs and correctly localized the IC in 93 of 121 (76.9%) cases, with 48 FP heatmaps on the mammograms for ICs (rate, 0.093 per case) and 74 FP heatmaps on normal mammograms (rate, 0.048 per case). The AI algorithm correctly localized a lower proportion of TN ICs (54 of 427; 12.6%) than ICs with MS (35 of 76; 46%) and FN ICs (four of eight; 50% [95% CI: 13, 88]; < .001). The AI algorithm localized a higher proportion of node-positive than node-negative cancers ( = .03). However, no evidence of a difference by cancer type ( = .09), grade ( = .27), or hormone receptor status ( = .12) was found. At 89.8% specificity and 79% sensitivity thresholds, AI detection increased to 181 (35.2%) and 256 (49.8%) of the 514 ICs, respectively, with FP heatmaps on 158 (10.2%) and 307 (19.8%) of the 1548 normal mammograms. Conclusion Use of a standalone AI system improved early cancer detection by correctly identifying some cancers missed by two human readers, with no differences based on histopathologic features except for node-positive cancers. © RSNA, 2024 .
背景 人工智能（AI）系统可用于识别间隔期乳腺癌，尽管定位并不总是准确。目的 评估 AI 对筛查性乳房 X 线照片中间隔期癌症（IC）的定位，根据 IC 类别和组织病理学特征进行分类。材料与方法 回顾性分析了 2011 年 1 月至 2018 年 12 月期间由两位人类读者评估的筛查性乳房 X 线摄影数据集（中位患者年龄为 57 岁[IQR，52-64 岁]），并使用商业 AI 系统进行分析。AI 的输出结果是病变位置（热图）和分配给每个病例的最高病变风险评分（范围，0-100）。如果 AI 热图出现在正常筛查性乳房 X 线照片或 IC 筛查性乳房 X 线照片（即随后诊断为 IC 的患者）上，但位于癌症边界之外，则认为是假阳性（FP）。一组顾问放射学专家将 IC 分为正常或良性（真阴性[TN]）、不确定（最小恶性迹象[MS]）或可疑（假阴性[FN]）。应用了几种特异性和敏感性阈值。采用 Mann-Whitney 检验、Kruskal-Wallis 检验和 χ 检验比较组间差异。结果 共纳入 2052 份筛查性乳房 X 线照片（514 例 IC 和 1548 例正常乳房 X 线照片）。TN IC 的中位数 AI 风险评分为 50（IQR，32-82），MS 的 IC 为 76（IQR，41-90），FN 的 IC 为 89（IQR，81-95）（ =.005）。与非浸润性肿瘤（33[IQR，20-55]； <.01）和低级别（0-1）肿瘤（45[IQR，26-81]； =.02）相比，浸润性肿瘤（62[IQR，39-88]）的中位数 AI 评分更高，高级别（2-3）肿瘤（62[IQR，40-87]）的中位数 AI 评分更高。在 96%特异性阈值下，AI 算法标记了 514 例 IC 中的 121 例（23.5%），并正确定位了 121 例中的 93 例（76.9%），在 IC 的乳房 X 线照片上有 48 个 FP 热图（每例发生率，0.093），在正常乳房 X 线照片上有 74 个 FP 热图（每例发生率，0.048）。AI 算法正确定位 TN IC 的比例较低（427 例中的 54 例；12.6%），低于 MS IC（76 例中的 35 例；46%）和 FN IC（8 例中的 4 例；50%[95%CI：13，88]； <.001）。AI 算法定位了更高比例的淋巴结阳性癌症（ =.03）。然而，未发现癌症类型（ =.09）、分级（ =.27）或激素受体状态（ =.12）存在差异。在 89.8%特异性和 79%敏感性阈值下，AI 检测增加到 514 例 IC 中的 181 例（35.2%）和 256 例（49.8%），在 1548 例正常乳房 X 线照片中 FP 热图分别为 158 例（10.2%）和 307 例（19.8%）。结论 使用独立的 AI 系统可通过正确识别两位人类读者漏诊的一些癌症，提高早期癌症检测的准确性，除了淋巴结阳性癌症外，在组织病理学特征方面无差异。
Zotero 插件