斯坦福2012ASA研究报告:用Phone4扫描拭血纱垫的方法估计术中出血量
美国斯坦福大学医学院的Siddarth Satish等使用iPhone4扫描拭血纱垫的方法估计术中出血量，以期对输血决策提供更便捷有效的依据。这种方法使用OpenCV图像数据库对扫描图像的颜色与像素进行分析，研究结果发表于2012年美国麻醉医师年会上。A809
October 14, 2012
1:00:00 PM - 4:00:00 PM
Room Hall C-Area K
Novel Method for Real-Time Measurement of Surgical Blood Loss
Siddarth Satish, M.S., Charles L. Zhao, M.S., Amer G. Abdulla, M.S., Kevin Miller, B.S., David M. Rempel, M.D., Mark L. Gonzalgo, M.D.,Ph.D., Michael H. Hsieh, M.D.,Ph.D.
Stanford University School of Medicine, Stanford, California, United States
Introduction: Measuring perioperative blood loss is an integral quality indicator for any surgical procedure and is an important factor in monitoring hemodynamic stability of the patient and determining the need for infusion of crystalloid, colloid or blood products. Visual estimation of blood loss, although standard, remains highly error-prone. To this end, a computer-vision algorithm for measurement of blood loss via optical scans of surgical surfaces was developed. The aim of this study was to validate the performance of the algorithm in benchtop and clinical settings.
Methods: A non-parametric regression algorithm was implemented to analyze visible color data from the active pixel sensor of an iPhone 4. OpenCV, a cross-platform library focused on real-time computer vision applications, was employed using the C++ programming language. The algorithm was calibrated for prediction of blood loss on surgical sponges via training and testing on outdated human blood. Standardized samples of surgical blood-absorbing media were employed, including Ray-Tec X-Ray-Detectable Sponges (Cardinal Health) and Lap Sponges (NovaPlus). An iPhone 4 was mounted at a distance of 16.5 cm +/- 0.3 cm above the sponge surface. Simulated blood spills were generated and Images of the soaked sponges were captured and processed via the prediction algorithm. EBL was then compared to ABL using RMS error methods to assess quality of algorithm training. To further validate the algorithm in a clinical setting, scans of blood soaked sponges were collected in operating rooms at Stanford Hospital & Clinics and at Lucille Packard Children’s Hospital following IRB approval. All gauze pads used to collect blood during these procedures were scanned for post-procedural EBL calculation via the trained algorithm. Soaked gauze pad weights were subtracted from dry weight to obtain estimates for actual blood volume (ABL). Visual EBL values estimated by the staff (per specimen) were also recorded for comparison. Accuracies of visual and algorithmic estimates were assessed using RMS error methods.
Results: Relative to gravimetric blood loss estimates, visual estimations by staff varied widely and a systematic overestimate was present. The algorithm outperformed the staff at high blood volumes (Figure 1). Although a systematic underestimation by the algorithm (relative to gravimetric determination) was noted, staff estimate variability was significantly higher than algorithm variability. These results suggest a high degree of precision in the algorithm that may ensure a high degree of reliability in measurement. It is important to note that a contributor to the observed underestimate was the presence of saline in weighed sponges, which confounds the gravimetric assessment of ABL.
Conclusions: Inaccurate estimation of blood loss may contribute to inappropriate transfusions resulting in increased morbidity and mortality to patients. Results of this pilot study demonstrate the promise of computer vision approaches for measurement of blood loss via optical scanning. The development of a standardized method to quantify blood loss could lead to (1) improved determination of the need for intraoperative transfusions (2) quality management tools to evaluate EBL by surgical procedure as part of peer-review/performance processes, and (3) point of care testing to evaluate when to initiate therapeutic interventions for bleeding (pharmacologic interventions, blood component therapy). 大师给翻译一下最好了 求翻译啊，iPhone在中国市场占有率还是不小的，能为我们做点实事那就最好了。
页:
[1]