Finck Tom, Moosbauer Julia, Probst Monika, Schlaeger Sarah, Schuberth Madeleine, Schinz David, Yiğitsoy Mehmet, Byas Sebastian, Zimmer Claus, Pfister Franz, Wiestler Benedikt
Department of Diagnostic and Interventional Neuroradiology, Klinikum Rechts der Isar, Technische Universität München, Ismaninger Straße 22, 81675 Munich, Germany.
DeepC GmbH, Atelierstraße 29, 81671 Munich, Germany.
Diagnostics (Basel). 2022 Feb 10;12(2):452. doi: 10.3390/diagnostics12020452.
Most artificial intelligence (AI) systems are restricted to solving a pre-defined task, thus limiting their generalizability to unselected datasets. Anomaly detection relieves this shortfall by flagging all pathologies as deviations from a learned norm. Here, we investigate whether diagnostic accuracy and reporting times can be improved by an anomaly detection tool for head computed tomography (CT), tailored to provide patient-level triage and voxel-based highlighting of pathologies.
Four neuroradiologists with 1-10 years of experience each investigated a set of 80 routinely acquired head CTs containing 40 normal scans and 40 scans with common pathologies. In a random order, scans were investigated with and without AI-predictions. A 4-week wash-out period between runs was included to prevent a reminiscence effect. Performance metrics for identifying pathologies, reporting times, and subjectively assessed diagnostic confidence were determined for both runs.
AI-support significantly increased the share of correctly classified scans (normal/pathological) from 309/320 scans to 317/320 scans ( = 0.0045), with a corresponding sensitivity, specificity, negative- and positive- predictive value of 100%, 98.1%, 98.2% and 100%, respectively. Further, reporting was significantly accelerated with AI-support, as evidenced by the 15.7% reduction in reporting times (65.1 ± 8.9 s vs. 54.9 ± 7.1 s; < 0.0001). Diagnostic confidence was similar in both runs.
Our study shows that AI-based triage of CTs can improve the diagnostic accuracy and accelerate reporting for experienced and inexperienced radiologists alike. Through ad hoc identification of normal CTs, anomaly detection promises to guide clinicians towards scans requiring urgent attention.
大多数人工智能（AI）系统仅限于解决预定义任务，因此限制了它们对未选定数据集的通用性。异常检测通过将所有病变标记为与学习到的规范的偏差来弥补这一不足。在此，我们研究一种针对头部计算机断层扫描（CT）的异常检测工具能否提高诊断准确性和报告时间，该工具旨在提供患者级别的分诊和基于体素的病变突出显示。
四名分别具有1至10年经验的神经放射科医生研究了一组80例常规采集的头部CT，其中包括40例正常扫描和40例有常见病变的扫描。扫描按随机顺序在有和没有AI预测的情况下进行研究。两次运行之间包括4周的洗脱期，以防止记忆效应。确定了两次运行中识别病变的性能指标、报告时间以及主观评估的诊断置信度。
AI支持显著提高了正确分类扫描（正常/病变）的比例，从309/320次扫描提高到317/320次扫描（P = 0.0045），相应的灵敏度、特异度、阴性预测值和阳性预测值分别为100%、98.1%、98.2%和100%。此外，AI支持显著加快了报告速度，报告时间减少了15.7%（65.1±8.9秒对54.9±7.1秒；P < 0.0001）。两次运行中的诊断置信度相似。
我们的研究表明，基于AI的CT分诊可以提高经验丰富和经验不足的放射科医生的诊断准确性并加快报告速度。通过临时识别正常CT，异常检测有望引导临床医生关注需要紧急处理的扫描。
