肋骨骨折深度学习辅助诊断系统对急诊老年患者的应用价值

Application of Computer-aided Diagnostic System for Rib Fractures in Emergency Elderly Patients

探讨急诊老年患者与中青年患者新鲜肋骨骨折的诊断效能,并评估基于深度学习(DL)的肋骨骨折计算机辅助诊断(CAD)系统在老年患者中的应用效果。选取809例因胸部外伤行胸部CT急诊检查患者,分为中青年组和老年组,6名医师分别独立阅片,并在CAD系统辅助下再次阅片,记录每次阅片的结果。2名具有15年以上胸部CT诊断经验的放射科高年资医师对患者初诊及复诊CT独立阅片,结果不一致时以2人协商一致的意见作为诊断金标准。统计并比较采用2种阅片方式对2组不同年龄患者行新鲜肋骨骨折诊断的效能和阅片时间。结果显示,医师对老年组独立阅片的敏感度显著低于中青年组(P<0.05),但在CAD系统辅助下,敏感度显著提高,与中青年组无显著差异。医师对老年组独立阅片的假阳性率显著高于中青年组(P<0.05),但在CAD系统辅助下,假阳性率显著降低,与中青年组无显著差异。医师对老年组独立阅片时间显著长于中青年组(P<0.05),但在CAD系统辅助下,阅片时间显著缩短。综上所述,医师对急诊老年患者新鲜肋骨骨折独立阅片的诊断效能低于中青年患者;采用基于深度学习的肋骨骨折CAD系统辅助医师阅片老年患者的CT图像,可在减少阅片时间的同时提高诊断效能,达到和医师对中青年患者独立阅片相同的诊断效能。

The objectives of this study were to investigate the diagnostic efficacy of fresh rib fractures in emergency older and middle-aged and young patients and to evaluate the value of a deep learning (DL)-based computer-aided diagnostic (CAD) system for rib fractures in older patients. A total of 809 patients who underwent emergency chest CT examination due to chest trauma were selected and divided into two aged groups: middle-aged and young group, and older group. CT images were initially read independently by six radiologists, and subsequently reread by the same six radiologists with the assistance of the CAD system. The locations and types of the fresh rib fractures as well as each reading time were recorded by themselves. The initial and follow-up CT images were independently read by two senior radiologists, each possessing 15 years of experience in thoracic CT diagnosis, who represented the reference standard, including the locations and types of the fresh rib fractures. If the findings were ambiguous, the findings of two senior radiologists’ final discussion were set as the reference standard. The sensitivity, false-positives rate (FPR) and reading time were compared for the two groups and for the two reading methods. The sensitivity for detecting fresh fractures using radiologist independently reading was lower in the older than in the middle-aged and young group. With the assistance of the CAD system, the sensitivity significantly increased in the older group to the same level as in the middle-aged and young group using radiologist independently reading. The FPR of fresh fractures with radiologist independently reading was higher in the older than in the middle-aged and young group. With the assistance of the CAD system, the FPR in the older group decreased to the same level as in the middle-aged and young group when using radiologist independently reading. The reading time of fresh fractures when using radiologist independently reading was longer in the older than in the middle-aged and young group. With the assistance of the CAD system, the reading time in the older group was significantly reduced. In conclusion, the efficacy of radiologists in diagnosing fresh rib fractures by radiologist independently reading in emergency older patients was shown to be lower than that in middle-aged and young patients. When radiologists are assisted by a deep learning-based CAD system, the diagnostic efficacy of identifying fresh fractures in older patients improves to the same level as independent reading by radiologists in middle-aged and young patients, while reducing the reading time.