不同品牌3MP影像诊断显示器对孤立性肺结节识读影响的ROC评价
|
摘要
【背景与目的】
随着图像存储与传输系统(PACS)以及图像后处理与分析技术的发展,以计算机和显示器为主的“软拷贝”阅读模式,已经越来越普及,逐步取代“硬拷贝”的阅读模式。“软拷贝”阅读及诊断方式大大加速了图像的贮存、传输及诊断过程,已成为影像诊断科室最终发展方向。但医用影像诊断专业显示器应用于“软拷贝”识读影像,对诊断所带来的准确性和可靠性的影响,目前还在研究评估中,也是相关领域的热门课题。
当前,国外采用受试者操作特性(ROC)曲线分析方法对数字影像进行了大量的探讨,国内对于ROC曲线分析方法的应用研究也有一定的进展,在数字化影像中对肺部小结节等细微结构的显示、识读是放射学研究中的一个关键因素,也是一个难点。国内外学者已经在不同分辨率的显示器对于肺部小结节识读影响的方面做出许多有益的探讨及研究,已经普遍认同了软阅读与硬阅读具有同样的诊断效能,高分辨率的显示器在识读及诊断方面明显优于低分辨率显示器。在以上研究基础上,本课题的研究方向是①通过PACS影像诊断工作站所应用的不同品牌的影像诊断显示器(相同分辨率)对胸部数字化X线摄影(DR)影像上肺结节检出准确率及结节性质的比较,评价品牌之间显示器的性价比(诊断效能)。②针对国内外所研究的影像诊断显示器对肺结节的识别全部基于不同品牌、不同分辨率,并且很多比较研究都是在液晶显示器(LCD)及阴极射线管显示器(CRT)中进行比较,本研究的范畴进一步规范化,显示器全部应用不同品牌的3百万象素(MP)的LCD,基于同一平台,对肺结节的识读进行对比,比较其差异性。并应用ROC曲线对3MP的显示器在肺结节诊断精确度上进行初步探讨。目的是进一步探讨不同品牌医用影像诊断显示器选用的必要性及可行性,给各医疗单位在显示器的选用方面提供初步指导意见,使医院合理利用资源,合理配备显示器,达到较高的性价比。
【材料与方法】
1、病例的选取
由2位未参加本试验研究的高年资(＞15年)放射诊断医师从本院PACS数据库中搜寻2005年3月～2008年12月的胸部DR,发现或者怀疑有结节者行CT检查,对资料进行筛选,在CT上发现单发、直径≦30mm的结节,列为阳性组;而CT证实无结节者列入阴性组。按上述方法选取90例DR胸片资料,其中64例DR胸片经CT证实有结节者为阳性组,按结节直径分为1～10mm组(微结节组)24例和10～30mm组(结节组)40例;26例DR胸片经CT证实无结节者为阴性组。
2、实验器材及方法
所采用的三种不同品牌显示器分别为台湾的奇菱CHILIN MDM2130-3NC(品牌A)、日本的艺卓EIZO RadiForce G31(品牌B)、美国的平达PLANARDome E3n(品牌C),分辨率为(1536×2048)20.8英寸医用竖屏单色LCD;5位医师分别在3种显示器上进行2次独立阅片,每次间隔时间为2周,每位医师每次均可根据需要使用窗宽、窗位调节功能、放大及不放大功能,阅片室环境控制在120lux以下。
3、实验一
90例肺结节,应用不同品牌3MP显示器,由5位放射诊断医生(5年＜工作年限＜8年)在PACS上分别对胸片上是否有肺结节进行判断,并记录结节的具体位置,以CT诊断为金标准,评价的标准采用5分法:肯定有、可能有、不确定、可能无、肯定无。
4、实验二
选取结节组的肺结节,从8个方面对结节征象及周边征象进行了对比,主要分为①结节边缘是否清晰,②结节密度是否均匀,③分叶有无及深浅,④毛刺有无及长短粗细,⑤有无钙化,⑥有无胸膜尾征,⑦有无空洞,⑧有无空腔等征象;由2位高年资放射诊断医师在CT影像上作判断并统一意见为相对金标准。5位放射诊断医师对病灶周边的征象进一步加以对比,综合意见同相对金标准比较,判定各品牌显示器诊断的符合率。并利用显示器做出结节组肺结节伴随征象的数据进一步对肺部结节进行初步定性诊断,以手术病理为金标准加以判断,判定各品牌显示器对结节定性诊断的符合率。
5、统计方法
对于微结节组的结节,数据结果利用SPSS 13.0软件绘制ROC曲线并进行统计分析,计算和比较每种显示系统的诊断精确度(用曲线下面积-Az值来表示)及各品牌之间有无统计学差异。并对3种显示器工作站下的Az值及Sx(Az)值进行配伍组方差分析(P＜0.05表示差异有统计学意义,P＜0.01表示差异有显著统计学意义),比较三种品牌显示器的诊断效能有无差异。对于结节组的结节,仔细核对每位医师的诊断结果,并与试验前所记录的“相对金标准”对每一结节的8个方面及定性诊断进行对照,对肺结节周边伴随征象及结节定性的对比采用Kappa分析(多类别多评估者分析)进行一致性检验,符合率、K值越接近1,一致性越好,符合率也越高;得出各显示器符合率,最后进行综合比较。
【结果】
1、3MP的不同品牌显示器对于≤10mm的孤立性肺部结节的识读,品牌A、B、C显示器的Az值的总平均值((?)+Sx)分别为(0.877±0.053,0.895±0.050,0.887±0.051),结果显示,显示器B的诊断效能最高,显示器A、C的诊断效能稍低于B,但三个品牌显示器曲线下面积都接近0.9。经过配伍组方差分析得出,F=6.686,P=0.020＜0.05,三台显示器对小于10mm的孤立性肺结节识读有差异。多重比较P_(AB)=0.006＜0.05;P_(BC)=0.135,P_(AC)=0.082,均大于0.5,可以得出,不同品牌相同分辨率医用单色显示器对于10mm以下的孤立性肺结节检出以品牌B诊断效能较高,品牌B、C间,品牌A、C差异无统计学意义,但品牌C诊断效能高于品牌A;而品牌A、B之间差异有统计学意义的。
2、对于10～30mm的肺部结节,品牌A、B、C显示器的Az值的总平均值((?)±Sx)分别为(0.891±0.045,0.901±0.044,0.892±0.045);结果显示,显示器B的诊断效能是最高的,显示器A、C的诊断效能稍低于B,但三个品牌的显示器曲线下面积都接近0.9。由配伍组方差分析得出,F=6.198,P=0.024＜0.05,三个品牌显示器对于10～30mm的肺部结节亦有统计学差异;多重比较得出P_(AB)=0.012,PBC=0.022,前两者P＜0.05,则表示品牌B同品牌A、C的差异有统计学意义;P_(AC)=0.720＞0.05,表示品牌A、C之间差异无统计学意义。因此,虽然品牌间的P值有所不同,部分具有统计学差异,但从三种品牌曲线下面积分别等于0.891、0.901、0.892,均接近于0.9;可以得出3MP的显示器在诊断孤立性肺结节的诊断精度总体来说是较高。
3、将5位医师对于10～30mm的孤立性肺结节的周边伴随征象经讨论统一意见,在3个品牌的显示器中分别同2位高年资诊断医师的结论(相对金标准)进行比对显示,各显示器对结节周边伴随征象识读同金标准有统计学差异,仔细观察每种品牌对孤立性肺结节伴随征象的识读,并认真记录所得数据,利用kappa分析,品牌B的K值有5项超过了0.40,品牌A、C的K值各有2项超过了0.40;从8种分类比较得出,3MP的显示器对于钙化、胸膜尾征及空洞空腔的符合率均达到了0.8以上,对于这些征象可以说DR有较高的识读能力;而对于边缘、密度、分叶及毛刺征,三种显示器的符合率明显降低,最低达到了0.575。综合来说,3MP的显示器对于结节周围主要征象的识读效能偏低,本研究所涉及的显示器中,品牌B的总体符合率及K值较其他两个显示器稍高,说明品牌B在孤立性肺结节周边伴随征象的识读方面稍强于其他两个品牌。
4、根据孤立肺结节的大小,形态及周边伴随征象的识读,对10～30mm的肺结节进行初步的定性诊断,结果同CT及病理对照比较,利用kappa分析,我们得到在结节性质的判定方面,品牌A、B、C的符合率分别为0.35、0.475、0.4,可以看出,DR胸片对孤立性肺结节定性诊断符合率较低,只能对肺部结节的良、恶性肿瘤或肿瘤样病变进行初步的筛选,必要的时候需要结合临床资料及进一步检查,最终的定性诊断还是靠CT和病理。
【结论】
1、本研究应用的分辨率相同(3MP)三种显示器中,其ROC曲线下的面积Az是有差别的,经过统计学检验,P值也有所不同,但在对肺结节的识读方面,不论是≤10mm的微结节,还是10～30mm的小结节,曲线下面积都接近于0.9,说明3MP的LCD在孤立性肺结节的诊断效能较高。因此,高分辨率影像诊断显示器能提高放射诊断医师的诊断准确率,减少由于假阳性带来的误诊,并在一定程度上减低诊断医师个体差别所造成的影响。
2、对于10～30mm的结节周边伴随征象的对比研究,结合文献报道,结果表明,3MP的显示器对于钙化、胸膜尾征及空洞空腔的识读符合率均达到了0.8以上,对于这些征象可以说DR有较高的识读能力;而对于边缘、密度、分叶及毛刺征,三种显示器的符合率明显降低,最低达到了0.575。所以DR胸片对于肺部结节病灶伴随征象及定性准确率不高。即使采用高分辨率的显示器,还是只能对肺部结节的定性进行初步判定。
3、由于医用专业LCD同普通LCD、CRT显示器对软读片室光照度要求不同,诊断效能也有差别,结合前面学者的文献,建议尽量选用高分辨率的专业LCD。但医用专业的LCD价格昂贵。因此可以根据医院的实际投入能力及各个科室、各个部门、各个岗位对影像质量的实际需求,进行合理配置。如乳房钼靶摄片图像要求5MP或5MP以上,CR、DR胸片要求3MP,而CT、MR、DSA、数字胃肠要求2MP,技术组图像处理要求2MP,相关显示及文字识别则可选用1MP的彩色显示器。不同分辨率显示器的合理配置,保证诊断准确率的前提下可获得较好的性价比。
4、对于普通放射诊断科室,完全可以采用医用显示器与普通显示器相结合组成诊断工作站系统,专业显示器用于影像识读,普通显示器主要用于放射信息系统(RIS)报告和资料查询,同时也可用于辅助阅片。由于不同品牌的医用专业显示器价格差别很大,建议以低成本高性价比的显示器配置诊断工作站系统,达到高诊断效能,充分发挥数字影像的优势并有效节约成本,促进PACS普及应用。
[Background and Objectives]
With the development of picture archiving and communication system (PACS) and image post-processing and analysis technology, mainly computers and monitors to the "soft copy" reading mode, has been becoming increasingly popular, and gradually replace the "hard copy" reading mode. "Soft copy" means reading and greatly accelerated the diagnosis of image storage, transmission and diagnosis of the process of diagnostic imaging has become the development direction of the final section. However, the diagnosis of medical images used in professional displays of "soft copy" reading images, the diagnosis brought about by the impact of accuracy and reliability, the current assessment is still under study, but also a hot topic related fields.
At present, Lots of research on using the Receiver Operating Characteristic (ROC) curve method to analysis digital images has been done abroad. There is also progress in the research of applying ROC curve domestically. Displaying and reading the minute structure in digital image, such as the small pulmonary nodules in chest image, is a key factor and a challenge in radiology research. Domestic and foreign scholars have been at different resolution displays for small lung nodules has made reading the impact of many useful and Research, has been generally accepted "soft reading" and "hard reading" with the same diagnostic performance, It's also believed that display with high resolution are much better than monitors with low resolution.
Research on the basis of the above, the issue is the direction of the research diagnostic imaging①Through PACS workstation to application of different diagnostic imaging brands display (same resolution) on the chest Digital Radiography (DR) images on the accuracy of pulmonary nodule detection and the nature of nodules compared to evaluate the brand value between the display (diagnostic performance).②Studied at home and abroad, the display of the image diagnosis of pulmonary nodules based on the identification of all the different brands, different resolution, and a lot of comparative studies have been conducted in liquid crystal display (LCD) and cathode ray tube(CRT) display in comparison, the study to further standardize and monitor all the different brands of 3 million pixels (MP) of the LCD as a fundamental, based on the same platform, the reading of pulmonary nodules compared to compare their differences. And the application of ROC curves on display in 3MP diagnostic accuracy of pulmonary nodules on the preliminary study. The purpose is to further explore the different brands of medical diagnostic imaging and the need to monitor the feasibility of selection, to the medical unit of selection in the display to provide initial guidance to make rational use of hospital resources, with a reasonable display, to achieve a higher price.
[Materials and Methods]
1. The selected cases
By the two did not participate in this pilot study of senior-owned (> 15 years) doctors of diagnostic radiology from our search in March 2005-December 2008 in PACS database of the chest DR,and find or doubt nodules to CT screening examination, was found in a single CT diameter <30mm of nodules, and before two weeks to CT examination of chest examination was DR experts as positive; Also searched suspected chest nodules, and CT confirmed these nodules were not included in the negative group. According to nodule diameter,these cases were divided to 3 groups, 1～10mm group 24 cases and 10～30mm 40 cases; 26 cases of chest DR found but CT examining was not found as the negative group.
2. Experimental equipment and methods
Used three different brands of monitors ,Taiwan's CHILIN MDM2130-3NC (brand A), Japan EIZO RadiForce G31 (brand B), and the United States PLANAR Dome E3n (brand C), resolution of (1536×2048),they are all 20.8 inch monochrome screen erected Medical LCD; Five Diagnostic Radiology physicians at 3 brands of display separately on a second independent access to films, Each doctor can use the tools of the window width, level adjustment,enlarge and zoom access and the films room environmental control in the following 1201ux.
3. Experimental one
There are 90 cases of pulmonary nodules, using different brands 3MP monitors, diagnosed respectively by the five doctors (5 yearsand the nodule determined by CT scaning as the gold standard.Recording the specific location of nodules, the evaluation criterion used by the 5 Method: definitely positive, probably positive, indeterminate, probably negative and definitely negative.
4. Experimental two
Select 10～30mm groups of pulmonary nodules, from eight aspects of nodular surrounding signs and symptoms were compared, mainly divided into①whether the nodular edge clear,②whether uniform density nodule,③the depth of leaf availability,④the availability and the length of burr thickness,⑤with calcification or not,⑥with pleural tail sign or not,⑦wih empty or not,⑧with cavity or not. For the surrounding signs and symptoms of pulmonary nodules ,CT scans of the same patient served to two senior physician-owned diagnostic as the gold standard. 5 Diagnostic Radiology physicians around for signs of lesions further comparison, the comprehensive views of the same relative standard of comparison, to determine the display of the brand in line with the rate of diagnosis. And using the display to further pulmonary nodules on the group N the data around the signs of a preliminary characterization of pulmonary nodule diagnosis, surgery and pathology in order to be judged as a gold standard to determine the brand of the nodules display in line with the rate of diagnosis.
5. Statistical methods
For the nodules of group 1～10mm, the interpreting results were analyzed by the SPSS 13.0 software and the receiver operating characteristiccurve(ROC) was painted, calculating and comparing each display system diagnostic accuracy (area under the curve with values expressed as-Az) and the availability of statistics between the different brands. And analysising the the Az and Sx (Az) value(P <0.05,that the difference was significant) under the there kinds of workstations, comparing the diffennece between the three brands displays in diagnostic performance. Nodules for group 10 ~ 30mm, carefully check every the datas of every doctor and the records before trial, "the relative gold standard" for each of the eight aspects of nodules and to control the qualitative diagnosis of peripheral pulmonary nodules with signs and qualitative comparison of the use of nodular Kappa analysis in line with the rate of the display, and finally have a comprehensive comparison.
[Results]
1. The 3MP different brands display for the isolation of < 10mm pulmonary nodules reading from the area under the curve can be seen, brand A, B, C displays thetotal average Az value (X±Sx) are (0.877±0.053,0.895±0.050,0.887±0.051), the display of the diagnostic performance of B is the highest, A and C is slightly lower, but three brands display area under the curve are near 0.9.Through combined analysis of variance, F = 6.686, P = 0.020 <0.05, less than 10mm of three displays of solitary pulmonary nodules are reading differently. Multiple comparisons, the PAB = 0.006 <0.05; PBC = 0.135, PAC = 0.082, both larger than 0.5.From that, we can see different brand of medical-resolution monochrome display the same 10mm below the solitary pulmonary nodule detection in diagnosis of the effectiveness is different. The brand B is higher than brand B and C, the difference between the brand A and C was not statistically significant. But the diagnostic performance of the brand A is higher than C, and brand A and B were statistically significant different.
2. For lung nodules of 10～30mm, brand A, B, C display the total average Azvalue (X±Sx) was (0.891±0.045,0.901±0.044,0.892±0.045); The results showed the Diagnosis of brand B is the highest performance, display A and C is slightly lower than the brand B, but three brands display area under the curve are all close to 0.9. By the combined group analysis of variance, F = 6.198, P = 0.024 <0.05, the three brands for the display of the 10～30mm are also significant difference between pulmonary nodule; Multiple comparisons drawn PAB = 0.012, PBC = 0.022,P <0.05, while that brand B with brand A and C were statistically significant differences; PAC = 0.720> 0.05, that brand A and C the difference was not between statistical significance. Thus, although the P value between brands vary, some with a significant difference, but in three separate brand names mean area under the curve can be drawn 0.891,0.901,0.892, are nearly equal to 0.9. So 3MP display s in the diagnosis diagnostic accuracy of pulmonary nodules is higher in general.
3. The five doctors disscuss and unify the solitary pulmonary nodules of 10～30mm surrounding signs and symptoms, and the three brands with the display of two senior physician-owned diagnostic conclusions (as relative to gold standard) to carry out than the right. The results showed that the display of signs associated with peripheral nodular read with the significant difference between the standards, careful observation of each brand of solitary pulmonary nodules with signs reading, and carefully recorded data, Using kappa analysis (multi-class analysis of multi-evaluators) conformance testing, if in line with the rate, the K value is more close to 1, consistency as possible, consistent with the higher rate. The K value of brand B exceeds 0.40, K value of brand A and C exceeds 0.40. Classification from eight kinds of comparison, 3MP display for calcification, pleural tail sign and the empty cavity in line with the rate reached more than 0.8, it can be said that these signs with a high reading DR capability; As for the edge density, leaf and burr levy, in line with the three displays was significantly reduced, reaching a minimum 0.575. In summary, 3MP display nodular surrounding signs and symptoms for the low reading performance, the institute involved in the display, brand B in line with the overall rate and the K value slightly higher than the other two monitors, on brand B in the isolated peripheral pulmonary nodules with signs reading a little in regard to the other two brands.
4. Solitary pulmonary nodules based on the size, shape and perimeter with signs reading, the 10～30mm of pulmonary nodules as a preliminary diagnosis, the results with CT and pathology compared using kappa analysis, we have the nature of the nodules determine the context, brand A, B and C were consistent with the 0.35,0.475,0.4. The results showed, DR of solitary pulmonary nodules consistent with the low rate of diagnosis can only be good for the lung nodules, malignant tumors or tumor-like lesions in a preliminary screening. If necessary, when combined with clinical data and the need for further examination , the final diagnosis by CT and pathological.
[Conclusion]
1.In this study, the same resolution (3MP) in three types of display, the areas under the ROC curve Az are different. After statistical testing, the P values are different, but the reading of the pulmonary nodules, Whether a micro-nodules of < 10mm or 10～30mm of small nodules, the area under the curve is close to 0.9, solitary pulmonary nodule diagnosis of high performance on the 3MP LCD. Therefore, the diagnosis of high-resolution image display diagnostic radiology physicians can improve diagnostic accuracy and reduce false positives arising due to misdiagnosis, and to some extent to reduce individual differences in physician diagnosis of the impact.
2.10～30mm for the perimeter of the nodules with the comparative study of signs, with review of the literature reported that the results show that, 3MP display for calcification, pleural tail sign and the hollow cavity of the reading in line with the rate has reached more than 0.8, it can be said that DR have a higher reading ability of these signs. For the edge density, leaf and burr levy, in line with the three displays were significantly reduced, reaching a minimum 0.575. Therefore, DR for signs of pulmonary nodule and characterization of lesions without high accuracy. Even with high-resolution display, or only for pulmonary nodule characterization of a preliminary determination.
3.LCD as a result of medical professionals with ordinary LCD, CRT display piece for the soft reading room illumination requirements are different, there are differences in diagnostic performance.It is recommended as far as possible the professional choice of high-resolution LCD. But the medical profession the high cost of LCD and high brightness, long-term use cause eyestrain. Can be put into actual use in accordance with the capacity of hospitals and various departments, all departments in various positions on the image quality of the actual demand for a reasonable configuration. Such as breast mammography 5MP or 5MP image above requirements, CR and DR chest requirements 3MP, and CT, MR, DSA, the number of gastrointestinal request 2MP, image processing requirements of the Technical Group 2MP, related to display and can use character recognition 1MP color display. Reasonable configurations of the display of different resolution, to ensure that diagnostic accuracy will be better under the premise of cost-effective.
4. For general radiodiagnoses department, it is definitely to combine the diagnosis workstation system with commercially available and professional monitors completely, the professional displays used for image reading and the commercially available displays mainly for general indicator of RIS reports and information enquirying. As different brands of medical professionals monitor the prices vary greatly, it is recommended to display cost-effective low-cost diagnostic workstation system configuration to achieve high diagnostic performance, and give full play to the advantages of digital imaging and effectively reduce costs and promote wider application of PACS.
随着图像存储与传输系统(PACS)以及图像后处理与分析技术的发展,以计算机和显示器为主的“软拷贝”阅读模式,已经越来越普及,逐步取代“硬拷贝”的阅读模式。“软拷贝”阅读及诊断方式大大加速了图像的贮存、传输及诊断过程,已成为影像诊断科室最终发展方向。但医用影像诊断专业显示器应用于“软拷贝”识读影像,对诊断所带来的准确性和可靠性的影响,目前还在研究评估中,也是相关领域的热门课题。
当前,国外采用受试者操作特性(ROC)曲线分析方法对数字影像进行了大量的探讨,国内对于ROC曲线分析方法的应用研究也有一定的进展,在数字化影像中对肺部小结节等细微结构的显示、识读是放射学研究中的一个关键因素,也是一个难点。国内外学者已经在不同分辨率的显示器对于肺部小结节识读影响的方面做出许多有益的探讨及研究,已经普遍认同了软阅读与硬阅读具有同样的诊断效能,高分辨率的显示器在识读及诊断方面明显优于低分辨率显示器。在以上研究基础上,本课题的研究方向是①通过PACS影像诊断工作站所应用的不同品牌的影像诊断显示器(相同分辨率)对胸部数字化X线摄影(DR)影像上肺结节检出准确率及结节性质的比较,评价品牌之间显示器的性价比(诊断效能)。②针对国内外所研究的影像诊断显示器对肺结节的识别全部基于不同品牌、不同分辨率,并且很多比较研究都是在液晶显示器(LCD)及阴极射线管显示器(CRT)中进行比较,本研究的范畴进一步规范化,显示器全部应用不同品牌的3百万象素(MP)的LCD,基于同一平台,对肺结节的识读进行对比,比较其差异性。并应用ROC曲线对3MP的显示器在肺结节诊断精确度上进行初步探讨。目的是进一步探讨不同品牌医用影像诊断显示器选用的必要性及可行性,给各医疗单位在显示器的选用方面提供初步指导意见,使医院合理利用资源,合理配备显示器,达到较高的性价比。
【材料与方法】
1、病例的选取
由2位未参加本试验研究的高年资(＞15年)放射诊断医师从本院PACS数据库中搜寻2005年3月～2008年12月的胸部DR,发现或者怀疑有结节者行CT检查,对资料进行筛选,在CT上发现单发、直径≦30mm的结节,列为阳性组;而CT证实无结节者列入阴性组。按上述方法选取90例DR胸片资料,其中64例DR胸片经CT证实有结节者为阳性组,按结节直径分为1～10mm组(微结节组)24例和10～30mm组(结节组)40例;26例DR胸片经CT证实无结节者为阴性组。
2、实验器材及方法
所采用的三种不同品牌显示器分别为台湾的奇菱CHILIN MDM2130-3NC(品牌A)、日本的艺卓EIZO RadiForce G31(品牌B)、美国的平达PLANARDome E3n(品牌C),分辨率为(1536×2048)20.8英寸医用竖屏单色LCD;5位医师分别在3种显示器上进行2次独立阅片,每次间隔时间为2周,每位医师每次均可根据需要使用窗宽、窗位调节功能、放大及不放大功能,阅片室环境控制在120lux以下。
3、实验一
90例肺结节,应用不同品牌3MP显示器,由5位放射诊断医生(5年＜工作年限＜8年)在PACS上分别对胸片上是否有肺结节进行判断,并记录结节的具体位置,以CT诊断为金标准,评价的标准采用5分法:肯定有、可能有、不确定、可能无、肯定无。
4、实验二
选取结节组的肺结节,从8个方面对结节征象及周边征象进行了对比,主要分为①结节边缘是否清晰,②结节密度是否均匀,③分叶有无及深浅,④毛刺有无及长短粗细,⑤有无钙化,⑥有无胸膜尾征,⑦有无空洞,⑧有无空腔等征象;由2位高年资放射诊断医师在CT影像上作判断并统一意见为相对金标准。5位放射诊断医师对病灶周边的征象进一步加以对比,综合意见同相对金标准比较,判定各品牌显示器诊断的符合率。并利用显示器做出结节组肺结节伴随征象的数据进一步对肺部结节进行初步定性诊断,以手术病理为金标准加以判断,判定各品牌显示器对结节定性诊断的符合率。
5、统计方法
对于微结节组的结节,数据结果利用SPSS 13.0软件绘制ROC曲线并进行统计分析,计算和比较每种显示系统的诊断精确度(用曲线下面积-Az值来表示)及各品牌之间有无统计学差异。并对3种显示器工作站下的Az值及Sx(Az)值进行配伍组方差分析(P＜0.05表示差异有统计学意义,P＜0.01表示差异有显著统计学意义),比较三种品牌显示器的诊断效能有无差异。对于结节组的结节,仔细核对每位医师的诊断结果,并与试验前所记录的“相对金标准”对每一结节的8个方面及定性诊断进行对照,对肺结节周边伴随征象及结节定性的对比采用Kappa分析(多类别多评估者分析)进行一致性检验,符合率、K值越接近1,一致性越好,符合率也越高;得出各显示器符合率,最后进行综合比较。
【结果】
1、3MP的不同品牌显示器对于≤10mm的孤立性肺部结节的识读,品牌A、B、C显示器的Az值的总平均值((?)+Sx)分别为(0.877±0.053,0.895±0.050,0.887±0.051),结果显示,显示器B的诊断效能最高,显示器A、C的诊断效能稍低于B,但三个品牌显示器曲线下面积都接近0.9。经过配伍组方差分析得出,F=6.686,P=0.020＜0.05,三台显示器对小于10mm的孤立性肺结节识读有差异。多重比较P_(AB)=0.006＜0.05;P_(BC)=0.135,P_(AC)=0.082,均大于0.5,可以得出,不同品牌相同分辨率医用单色显示器对于10mm以下的孤立性肺结节检出以品牌B诊断效能较高,品牌B、C间,品牌A、C差异无统计学意义,但品牌C诊断效能高于品牌A;而品牌A、B之间差异有统计学意义的。
2、对于10～30mm的肺部结节,品牌A、B、C显示器的Az值的总平均值((?)±Sx)分别为(0.891±0.045,0.901±0.044,0.892±0.045);结果显示,显示器B的诊断效能是最高的,显示器A、C的诊断效能稍低于B,但三个品牌的显示器曲线下面积都接近0.9。由配伍组方差分析得出,F=6.198,P=0.024＜0.05,三个品牌显示器对于10～30mm的肺部结节亦有统计学差异;多重比较得出P_(AB)=0.012,PBC=0.022,前两者P＜0.05,则表示品牌B同品牌A、C的差异有统计学意义;P_(AC)=0.720＞0.05,表示品牌A、C之间差异无统计学意义。因此,虽然品牌间的P值有所不同,部分具有统计学差异,但从三种品牌曲线下面积分别等于0.891、0.901、0.892,均接近于0.9;可以得出3MP的显示器在诊断孤立性肺结节的诊断精度总体来说是较高。
3、将5位医师对于10～30mm的孤立性肺结节的周边伴随征象经讨论统一意见,在3个品牌的显示器中分别同2位高年资诊断医师的结论(相对金标准)进行比对显示,各显示器对结节周边伴随征象识读同金标准有统计学差异,仔细观察每种品牌对孤立性肺结节伴随征象的识读,并认真记录所得数据,利用kappa分析,品牌B的K值有5项超过了0.40,品牌A、C的K值各有2项超过了0.40;从8种分类比较得出,3MP的显示器对于钙化、胸膜尾征及空洞空腔的符合率均达到了0.8以上,对于这些征象可以说DR有较高的识读能力;而对于边缘、密度、分叶及毛刺征,三种显示器的符合率明显降低,最低达到了0.575。综合来说,3MP的显示器对于结节周围主要征象的识读效能偏低,本研究所涉及的显示器中,品牌B的总体符合率及K值较其他两个显示器稍高,说明品牌B在孤立性肺结节周边伴随征象的识读方面稍强于其他两个品牌。
4、根据孤立肺结节的大小,形态及周边伴随征象的识读,对10～30mm的肺结节进行初步的定性诊断,结果同CT及病理对照比较,利用kappa分析,我们得到在结节性质的判定方面,品牌A、B、C的符合率分别为0.35、0.475、0.4,可以看出,DR胸片对孤立性肺结节定性诊断符合率较低,只能对肺部结节的良、恶性肿瘤或肿瘤样病变进行初步的筛选,必要的时候需要结合临床资料及进一步检查,最终的定性诊断还是靠CT和病理。
【结论】
1、本研究应用的分辨率相同(3MP)三种显示器中,其ROC曲线下的面积Az是有差别的,经过统计学检验,P值也有所不同,但在对肺结节的识读方面,不论是≤10mm的微结节,还是10～30mm的小结节,曲线下面积都接近于0.9,说明3MP的LCD在孤立性肺结节的诊断效能较高。因此,高分辨率影像诊断显示器能提高放射诊断医师的诊断准确率,减少由于假阳性带来的误诊,并在一定程度上减低诊断医师个体差别所造成的影响。
2、对于10～30mm的结节周边伴随征象的对比研究,结合文献报道,结果表明,3MP的显示器对于钙化、胸膜尾征及空洞空腔的识读符合率均达到了0.8以上,对于这些征象可以说DR有较高的识读能力;而对于边缘、密度、分叶及毛刺征,三种显示器的符合率明显降低,最低达到了0.575。所以DR胸片对于肺部结节病灶伴随征象及定性准确率不高。即使采用高分辨率的显示器,还是只能对肺部结节的定性进行初步判定。
3、由于医用专业LCD同普通LCD、CRT显示器对软读片室光照度要求不同,诊断效能也有差别,结合前面学者的文献,建议尽量选用高分辨率的专业LCD。但医用专业的LCD价格昂贵。因此可以根据医院的实际投入能力及各个科室、各个部门、各个岗位对影像质量的实际需求,进行合理配置。如乳房钼靶摄片图像要求5MP或5MP以上,CR、DR胸片要求3MP,而CT、MR、DSA、数字胃肠要求2MP,技术组图像处理要求2MP,相关显示及文字识别则可选用1MP的彩色显示器。不同分辨率显示器的合理配置,保证诊断准确率的前提下可获得较好的性价比。
4、对于普通放射诊断科室,完全可以采用医用显示器与普通显示器相结合组成诊断工作站系统,专业显示器用于影像识读,普通显示器主要用于放射信息系统(RIS)报告和资料查询,同时也可用于辅助阅片。由于不同品牌的医用专业显示器价格差别很大,建议以低成本高性价比的显示器配置诊断工作站系统,达到高诊断效能,充分发挥数字影像的优势并有效节约成本,促进PACS普及应用。
[Background and Objectives]
With the development of picture archiving and communication system (PACS) and image post-processing and analysis technology, mainly computers and monitors to the "soft copy" reading mode, has been becoming increasingly popular, and gradually replace the "hard copy" reading mode. "Soft copy" means reading and greatly accelerated the diagnosis of image storage, transmission and diagnosis of the process of diagnostic imaging has become the development direction of the final section. However, the diagnosis of medical images used in professional displays of "soft copy" reading images, the diagnosis brought about by the impact of accuracy and reliability, the current assessment is still under study, but also a hot topic related fields.
At present, Lots of research on using the Receiver Operating Characteristic (ROC) curve method to analysis digital images has been done abroad. There is also progress in the research of applying ROC curve domestically. Displaying and reading the minute structure in digital image, such as the small pulmonary nodules in chest image, is a key factor and a challenge in radiology research. Domestic and foreign scholars have been at different resolution displays for small lung nodules has made reading the impact of many useful and Research, has been generally accepted "soft reading" and "hard reading" with the same diagnostic performance, It's also believed that display with high resolution are much better than monitors with low resolution.
Research on the basis of the above, the issue is the direction of the research diagnostic imaging①Through PACS workstation to application of different diagnostic imaging brands display (same resolution) on the chest Digital Radiography (DR) images on the accuracy of pulmonary nodule detection and the nature of nodules compared to evaluate the brand value between the display (diagnostic performance).②Studied at home and abroad, the display of the image diagnosis of pulmonary nodules based on the identification of all the different brands, different resolution, and a lot of comparative studies have been conducted in liquid crystal display (LCD) and cathode ray tube(CRT) display in comparison, the study to further standardize and monitor all the different brands of 3 million pixels (MP) of the LCD as a fundamental, based on the same platform, the reading of pulmonary nodules compared to compare their differences. And the application of ROC curves on display in 3MP diagnostic accuracy of pulmonary nodules on the preliminary study. The purpose is to further explore the different brands of medical diagnostic imaging and the need to monitor the feasibility of selection, to the medical unit of selection in the display to provide initial guidance to make rational use of hospital resources, with a reasonable display, to achieve a higher price.
[Materials and Methods]
1. The selected cases
By the two did not participate in this pilot study of senior-owned (> 15 years) doctors of diagnostic radiology from our search in March 2005-December 2008 in PACS database of the chest DR,and find or doubt nodules to CT screening examination, was found in a single CT diameter <30mm of nodules, and before two weeks to CT examination of chest examination was DR experts as positive; Also searched suspected chest nodules, and CT confirmed these nodules were not included in the negative group. According to nodule diameter,these cases were divided to 3 groups, 1～10mm group 24 cases and 10～30mm 40 cases; 26 cases of chest DR found but CT examining was not found as the negative group.
2. Experimental equipment and methods
Used three different brands of monitors ,Taiwan's CHILIN MDM2130-3NC (brand A), Japan EIZO RadiForce G31 (brand B), and the United States PLANAR Dome E3n (brand C), resolution of (1536×2048),they are all 20.8 inch monochrome screen erected Medical LCD; Five Diagnostic Radiology physicians at 3 brands of display separately on a second independent access to films, Each doctor can use the tools of the window width, level adjustment,enlarge and zoom access and the films room environmental control in the following 1201ux.
3. Experimental one
There are 90 cases of pulmonary nodules, using different brands 3MP monitors, diagnosed respectively by the five doctors (5 years
4. Experimental two
Select 10～30mm groups of pulmonary nodules, from eight aspects of nodular surrounding signs and symptoms were compared, mainly divided into①whether the nodular edge clear,②whether uniform density nodule,③the depth of leaf availability,④the availability and the length of burr thickness,⑤with calcification or not,⑥with pleural tail sign or not,⑦wih empty or not,⑧with cavity or not. For the surrounding signs and symptoms of pulmonary nodules ,CT scans of the same patient served to two senior physician-owned diagnostic as the gold standard. 5 Diagnostic Radiology physicians around for signs of lesions further comparison, the comprehensive views of the same relative standard of comparison, to determine the display of the brand in line with the rate of diagnosis. And using the display to further pulmonary nodules on the group N the data around the signs of a preliminary characterization of pulmonary nodule diagnosis, surgery and pathology in order to be judged as a gold standard to determine the brand of the nodules display in line with the rate of diagnosis.
5. Statistical methods
For the nodules of group 1～10mm, the interpreting results were analyzed by the SPSS 13.0 software and the receiver operating characteristiccurve(ROC) was painted, calculating and comparing each display system diagnostic accuracy (area under the curve with values expressed as-Az) and the availability of statistics between the different brands. And analysising the the Az and Sx (Az) value(P <0.05,that the difference was significant) under the there kinds of workstations, comparing the diffennece between the three brands displays in diagnostic performance. Nodules for group 10 ~ 30mm, carefully check every the datas of every doctor and the records before trial, "the relative gold standard" for each of the eight aspects of nodules and to control the qualitative diagnosis of peripheral pulmonary nodules with signs and qualitative comparison of the use of nodular Kappa analysis in line with the rate of the display, and finally have a comprehensive comparison.
[Results]
1. The 3MP different brands display for the isolation of < 10mm pulmonary nodules reading from the area under the curve can be seen, brand A, B, C displays thetotal average Az value (X±Sx) are (0.877±0.053,0.895±0.050,0.887±0.051), the display of the diagnostic performance of B is the highest, A and C is slightly lower, but three brands display area under the curve are near 0.9.Through combined analysis of variance, F = 6.686, P = 0.020 <0.05, less than 10mm of three displays of solitary pulmonary nodules are reading differently. Multiple comparisons, the PAB = 0.006 <0.05; PBC = 0.135, PAC = 0.082, both larger than 0.5.From that, we can see different brand of medical-resolution monochrome display the same 10mm below the solitary pulmonary nodule detection in diagnosis of the effectiveness is different. The brand B is higher than brand B and C, the difference between the brand A and C was not statistically significant. But the diagnostic performance of the brand A is higher than C, and brand A and B were statistically significant different.
2. For lung nodules of 10～30mm, brand A, B, C display the total average Azvalue (X±Sx) was (0.891±0.045,0.901±0.044,0.892±0.045); The results showed the Diagnosis of brand B is the highest performance, display A and C is slightly lower than the brand B, but three brands display area under the curve are all close to 0.9. By the combined group analysis of variance, F = 6.198, P = 0.024 <0.05, the three brands for the display of the 10～30mm are also significant difference between pulmonary nodule; Multiple comparisons drawn PAB = 0.012, PBC = 0.022,P <0.05, while that brand B with brand A and C were statistically significant differences; PAC = 0.720> 0.05, that brand A and C the difference was not between statistical significance. Thus, although the P value between brands vary, some with a significant difference, but in three separate brand names mean area under the curve can be drawn 0.891,0.901,0.892, are nearly equal to 0.9. So 3MP display s in the diagnosis diagnostic accuracy of pulmonary nodules is higher in general.
3. The five doctors disscuss and unify the solitary pulmonary nodules of 10～30mm surrounding signs and symptoms, and the three brands with the display of two senior physician-owned diagnostic conclusions (as relative to gold standard) to carry out than the right. The results showed that the display of signs associated with peripheral nodular read with the significant difference between the standards, careful observation of each brand of solitary pulmonary nodules with signs reading, and carefully recorded data, Using kappa analysis (multi-class analysis of multi-evaluators) conformance testing, if in line with the rate, the K value is more close to 1, consistency as possible, consistent with the higher rate. The K value of brand B exceeds 0.40, K value of brand A and C exceeds 0.40. Classification from eight kinds of comparison, 3MP display for calcification, pleural tail sign and the empty cavity in line with the rate reached more than 0.8, it can be said that these signs with a high reading DR capability; As for the edge density, leaf and burr levy, in line with the three displays was significantly reduced, reaching a minimum 0.575. In summary, 3MP display nodular surrounding signs and symptoms for the low reading performance, the institute involved in the display, brand B in line with the overall rate and the K value slightly higher than the other two monitors, on brand B in the isolated peripheral pulmonary nodules with signs reading a little in regard to the other two brands.
4. Solitary pulmonary nodules based on the size, shape and perimeter with signs reading, the 10～30mm of pulmonary nodules as a preliminary diagnosis, the results with CT and pathology compared using kappa analysis, we have the nature of the nodules determine the context, brand A, B and C were consistent with the 0.35,0.475,0.4. The results showed, DR of solitary pulmonary nodules consistent with the low rate of diagnosis can only be good for the lung nodules, malignant tumors or tumor-like lesions in a preliminary screening. If necessary, when combined with clinical data and the need for further examination , the final diagnosis by CT and pathological.
[Conclusion]
1.In this study, the same resolution (3MP) in three types of display, the areas under the ROC curve Az are different. After statistical testing, the P values are different, but the reading of the pulmonary nodules, Whether a micro-nodules of < 10mm or 10～30mm of small nodules, the area under the curve is close to 0.9, solitary pulmonary nodule diagnosis of high performance on the 3MP LCD. Therefore, the diagnosis of high-resolution image display diagnostic radiology physicians can improve diagnostic accuracy and reduce false positives arising due to misdiagnosis, and to some extent to reduce individual differences in physician diagnosis of the impact.
2.10～30mm for the perimeter of the nodules with the comparative study of signs, with review of the literature reported that the results show that, 3MP display for calcification, pleural tail sign and the hollow cavity of the reading in line with the rate has reached more than 0.8, it can be said that DR have a higher reading ability of these signs. For the edge density, leaf and burr levy, in line with the three displays were significantly reduced, reaching a minimum 0.575. Therefore, DR for signs of pulmonary nodule and characterization of lesions without high accuracy. Even with high-resolution display, or only for pulmonary nodule characterization of a preliminary determination.
3.LCD as a result of medical professionals with ordinary LCD, CRT display piece for the soft reading room illumination requirements are different, there are differences in diagnostic performance.It is recommended as far as possible the professional choice of high-resolution LCD. But the medical profession the high cost of LCD and high brightness, long-term use cause eyestrain. Can be put into actual use in accordance with the capacity of hospitals and various departments, all departments in various positions on the image quality of the actual demand for a reasonable configuration. Such as breast mammography 5MP or 5MP image above requirements, CR and DR chest requirements 3MP, and CT, MR, DSA, the number of gastrointestinal request 2MP, image processing requirements of the Technical Group 2MP, related to display and can use character recognition 1MP color display. Reasonable configurations of the display of different resolution, to ensure that diagnostic accuracy will be better under the premise of cost-effective.
4. For general radiodiagnoses department, it is definitely to combine the diagnosis workstation system with commercially available and professional monitors completely, the professional displays used for image reading and the commercially available displays mainly for general indicator of RIS reports and information enquirying. As different brands of medical professionals monitor the prices vary greatly, it is recommended to display cost-effective low-cost diagnostic workstation system configuration to achieve high diagnostic performance, and give full play to the advantages of digital imaging and effectively reduce costs and promote wider application of PACS.
引文
[1]袁聿德,谢晋东,王昌元.加强对显示器诊断影像评价的研究[J].中华放射学杂志,2004,38(4):343-344.
[2]Hyun WGoo,Hyun JK,Koun-Sik S,et al.Using Edge Enhancement to Identify Subtle Findings on Soft-Copy Neonatal Chest Radiographs[J].AJR 2001,177(2):437-440.
[3]Michael J,Flynn,Ehsan Samei.Experimental comparison of noise and resolution for 2k and 4k storage phosphor radiography systems[J].Medical Physics,1999,26(8):1612-1623.
[4]陈自谦,张碧云,赵政文,等.数字化X线胸片计算机辅助检测在肺结节诊断中的应用价值[J].医学研究生学报,2005,18(7):629-631.
[5]Arenson RL,Chakraborty DP,Seshadri SB,et al.The digital imaging workstation[J].Radiology,1990,176(2):303-315.
[6]Erich Sorantin.Soft-copy display and reading:what the radiologist should know in the digital era.Pediatr Radiol,2008,38:1276-1284.
[7]Oschatz EA,Prokop MA,Scharitzer MA,et al.Comparison of liquid crystal versus cathode ray tube display for the detection of simulated chest lesions[J].Eur Radiol,2005,15:1472-1476.
[8]J.M.Goo,J.-Y.Choi,J.-G.Im,et al.Effect of Monitor Luminance and Ambient Light on Observer Performance in Soft-Copy Reading of Digital Chest Radiographs[J].Radiology,2004,232(3):762-766.
[9]Ueda K,Iwasaki S,Nagasawa M,et al.Hard-copy versus soft-copy image reading for detection of ureteral stones on abdominal radiography[J].Radiat Med,2003,21(5):210-213.
[10]M Uffmann,M Prokop,W Kupper,et al.Soft-Copy Reading of Digital Chest Radiographs:Effect of Ambient Light and Automatic Optimization of Monitor Luminance[J].Invest Radiol,2005,40(3):180-185.
[11]高培毅,林燕,张红梅,等.颅内恶性胶质瘤术后早期MR、CT对残存肿瘤检出 的受试者操作特征分析.中华放射学杂志,2000,34:240-243.
[12]Hadjiisld L,Chan H-P,Sahiner B,et al.Quasi-continuous and discrete confidence rating scales for observer performance studies:effects on ROC analysis.Acad Radiol.2007,14:38-48.
[13]Pesce LL,Metz CE.Reliable and computationally efficient maximum-likelihood estimation of "proper" binormal ROC curves.Acad Radiol.2007,14:814-29.
[14]李惠民,肖湘生.肺结节CT影像评价.中国医学计算机成像杂志,2001,7:30-41.
[15]M Ikeda,T Ishigaki,K Shimamoto,et al.Influence of Monitor Luminance Change on Observer Performance for Detection of Abnormalities Depicted on Chest Radiographs[J].Invest Radiol,2003,38(1):57-63.
[16]Ishihara S,Shimamoto K,Ikeda M,et al.CRT diagnosis of pulmonary disease:influence of monitor brightness and room illuminance on observer performance[J].Comput Med Imaging Graph,2002,26(3):181-185.
[17]吴杰,王学建,王波,等.不同显示矩阵影像诊断工作站识读肺部小结节的初步研究[J].中华放射学杂志,2004,38(3):295-300.
[18]梁志刚,李坤成,杨延辉,等.PACS中显示设备的选择对于肺部孤立小结节诊断影响的研究[J].临床放射学杂志,2005,24(7):637-639.
[19]程勇,陈卫国,张雪林,等.影像诊断工作站显示器的选择对孤立性肺结节识读影响的研究[J].中国医学影像技术,2007,23(2):284-287.
[20]Ikeda R,Katsuragawa S,Shimonobou T,et al.Comparison of LCD and CRT monitors for detection of pulmonary nodules and interstitial lung diseases on digital chest radiographs by using receiver operating characteristic analysis[J].Nippon Hoshasen Gijutsu Gakkai Zasshi,2005,61(7):974-5.
[21]Park CM,Lee HJ,Goo JM et al.Comparison of observer performance on soft-copy reading of digital chest radiographs:high resolution liquid-crystal display monitors versus cathode-ray tube monitors.Eur J Radiol,2008,66:13-18.
[22]Egan JP.Signal detection theory and ROC analysis.New York:Academic,1975.
[23]Lusted LB.Logical analysis in roentgen diagnosis(Review). Radiology, 1960,74 :178-193.
[24]Lusted LB.Observer error, signal detectability, medical decision making.In: Jacquez JA, editor.Computer diagnosis and diagnostic methods.Springfield:Thomas; 1972.p.29^4.
[25]Lusted LB.General problems in medical decision making, with comments on ROC analysis.SeminNucl Med.1978;8:299-306.
[26]Goodenough DJ, Rossmann K, Lusted LB.Radiographic applications of receiver operating characteristic (ROC) analysis.Radiology.1974;110:89-95.
[27]Metz CE, Starr SJ, Lusted LB.Quantitative evaluation of visual detection performance in medicine: ROC analysis and determination of diagnostic benefit.In: Hay GA, editor.Medical images: formation, perception and measurement.London: Wiley; 1977.p.220-240.
[28]Metz CE.Receiver operating characteristic (ROC) analysis: a tool for quantitative evaluation of observer performance and imaging systems.JACR.2006;3:413-22.
[29]Wagner RF, Metz CE, Campbell GAssessment of medial imaging systems and computer aids: a tutorial review.Acad Radiol.2007;14:723-48.
[30]Krupinski EA, Jiang Y.Evaluation of medical imaging systems.Med Phys.2008 (in press).
[31]孙军,陈峰,郑凯尔．ROC曲线分析在放射学中的应用[J]．中华放射学杂志,2001,35(8)：574-579．
[32]Armitage P, Colton T.Encyclopedia of biostatistics.New York:John , 1998, 3738-3744.
[33]Metz CE.ROC Methodology in radiologic imaging.Invest Radiol, 1986,21:720-733.
[34]Swensson RGUnified measurement of observer performanc in detecting and localizing target objects on images.Med Phys, 1996,23:1709-1725.
[35]Brismar J .Understanding receiver operating characteristic curves : agraphic approach.AJR, 1991 ,157: 1119-1121.
[36]Aldrich JE, Rutledge JD.Assessment of PACS display systems.J Digit Imaging,2005,18:287-295.
[37]Fukushima H, Ikeda M, Ishigaki T et al.Influence of liquid crystal display monitors on observer performance for detection of diffuse pulmonary disease on chest radiographs.Radiat Med (2007) 25:211-217
[38]Krupinski EA,Johnson J,Roehrig H,et al.Use of a Human Visual System Model to Predict Observer Performance with CRT vs LCD Display of Images[J].J Digit Imaging,2004,17(4):258-263.
[39]Balassy C, Prokop M, Weber M et al.Flat-panel display(LCD) versus high-resolution gray-scale display (CRT) for chest radiography: an observer preference study.AJR,2005,184:752-756.
[1]H.Usami,M.Ikeda,T.Ishigaki.et al.The influence of liquid crystal display(LCD)monitors on observer performance for the detection of nodular lesions on chest radiographs.Eur Radiol,2006,16:726-732.
[2]Scharitzer M,Prokop M,Weber M,et al.Detectability of catheters on bedside chest radiographs:comparison between liquid crystal display and high-resolution cathoderay tube monitors.Radiology.2005(234):611-616.
[3]Su Young Kim,Yoon Joon Hwang,Yoon Hee Han,et al.An ROC Study of Chest Radiographs:2K Versus 4K High-Resolution Soft-Copy Images,Journal of Digital Imaging,2007,Vol 20,No 4(December):347-351.
[4]Langer S,Fetterly K,Mandrekar J,et al.ROC study of four LCD displays under typical medical center lighting conditions.J Digit Imaging,2006,19:30-40.
[5]辛海燕,陈长忠,韩文英.医用显示器的选择不容忽视,医疗设备信息,2006,21(8):42-43.
[6]Hakan Geijer,Mats Geijer,Lillemor Forsberg,et al.Comparison of Color LCD and Medical-grade Monochrome LCD Displays in Diagnostic Radiology[J].Journal of Digital Imaging,2007,Vol 20,No 2(June):114-121.
[7]曹军,魏渝清,王学建,等.医学数字化影像显示终端空间分辨率的测试[J].中华放射学杂志,2005,39(10):1105-1106.
[8]严涛,陈钧,王冰.医用专业显示器应用与选型.中国数字医学,2007,8:35-40.
[2]Hyun WGoo,Hyun JK,Koun-Sik S,et al.Using Edge Enhancement to Identify Subtle Findings on Soft-Copy Neonatal Chest Radiographs[J].AJR 2001,177(2):437-440.
[3]Michael J,Flynn,Ehsan Samei.Experimental comparison of noise and resolution for 2k and 4k storage phosphor radiography systems[J].Medical Physics,1999,26(8):1612-1623.
[4]陈自谦,张碧云,赵政文,等.数字化X线胸片计算机辅助检测在肺结节诊断中的应用价值[J].医学研究生学报,2005,18(7):629-631.
[5]Arenson RL,Chakraborty DP,Seshadri SB,et al.The digital imaging workstation[J].Radiology,1990,176(2):303-315.
[6]Erich Sorantin.Soft-copy display and reading:what the radiologist should know in the digital era.Pediatr Radiol,2008,38:1276-1284.
[7]Oschatz EA,Prokop MA,Scharitzer MA,et al.Comparison of liquid crystal versus cathode ray tube display for the detection of simulated chest lesions[J].Eur Radiol,2005,15:1472-1476.
[8]J.M.Goo,J.-Y.Choi,J.-G.Im,et al.Effect of Monitor Luminance and Ambient Light on Observer Performance in Soft-Copy Reading of Digital Chest Radiographs[J].Radiology,2004,232(3):762-766.
[9]Ueda K,Iwasaki S,Nagasawa M,et al.Hard-copy versus soft-copy image reading for detection of ureteral stones on abdominal radiography[J].Radiat Med,2003,21(5):210-213.
[10]M Uffmann,M Prokop,W Kupper,et al.Soft-Copy Reading of Digital Chest Radiographs:Effect of Ambient Light and Automatic Optimization of Monitor Luminance[J].Invest Radiol,2005,40(3):180-185.
[11]高培毅,林燕,张红梅,等.颅内恶性胶质瘤术后早期MR、CT对残存肿瘤检出 的受试者操作特征分析.中华放射学杂志,2000,34:240-243.
[12]Hadjiisld L,Chan H-P,Sahiner B,et al.Quasi-continuous and discrete confidence rating scales for observer performance studies:effects on ROC analysis.Acad Radiol.2007,14:38-48.
[13]Pesce LL,Metz CE.Reliable and computationally efficient maximum-likelihood estimation of "proper" binormal ROC curves.Acad Radiol.2007,14:814-29.
[14]李惠民,肖湘生.肺结节CT影像评价.中国医学计算机成像杂志,2001,7:30-41.
[15]M Ikeda,T Ishigaki,K Shimamoto,et al.Influence of Monitor Luminance Change on Observer Performance for Detection of Abnormalities Depicted on Chest Radiographs[J].Invest Radiol,2003,38(1):57-63.
[16]Ishihara S,Shimamoto K,Ikeda M,et al.CRT diagnosis of pulmonary disease:influence of monitor brightness and room illuminance on observer performance[J].Comput Med Imaging Graph,2002,26(3):181-185.
[17]吴杰,王学建,王波,等.不同显示矩阵影像诊断工作站识读肺部小结节的初步研究[J].中华放射学杂志,2004,38(3):295-300.
[18]梁志刚,李坤成,杨延辉,等.PACS中显示设备的选择对于肺部孤立小结节诊断影响的研究[J].临床放射学杂志,2005,24(7):637-639.
[19]程勇,陈卫国,张雪林,等.影像诊断工作站显示器的选择对孤立性肺结节识读影响的研究[J].中国医学影像技术,2007,23(2):284-287.
[20]Ikeda R,Katsuragawa S,Shimonobou T,et al.Comparison of LCD and CRT monitors for detection of pulmonary nodules and interstitial lung diseases on digital chest radiographs by using receiver operating characteristic analysis[J].Nippon Hoshasen Gijutsu Gakkai Zasshi,2005,61(7):974-5.
[21]Park CM,Lee HJ,Goo JM et al.Comparison of observer performance on soft-copy reading of digital chest radiographs:high resolution liquid-crystal display monitors versus cathode-ray tube monitors.Eur J Radiol,2008,66:13-18.
[22]Egan JP.Signal detection theory and ROC analysis.New York:Academic,1975.
[23]Lusted LB.Logical analysis in roentgen diagnosis(Review). Radiology, 1960,74 :178-193.
[24]Lusted LB.Observer error, signal detectability, medical decision making.In: Jacquez JA, editor.Computer diagnosis and diagnostic methods.Springfield:Thomas; 1972.p.29^4.
[25]Lusted LB.General problems in medical decision making, with comments on ROC analysis.SeminNucl Med.1978;8:299-306.
[26]Goodenough DJ, Rossmann K, Lusted LB.Radiographic applications of receiver operating characteristic (ROC) analysis.Radiology.1974;110:89-95.
[27]Metz CE, Starr SJ, Lusted LB.Quantitative evaluation of visual detection performance in medicine: ROC analysis and determination of diagnostic benefit.In: Hay GA, editor.Medical images: formation, perception and measurement.London: Wiley; 1977.p.220-240.
[28]Metz CE.Receiver operating characteristic (ROC) analysis: a tool for quantitative evaluation of observer performance and imaging systems.JACR.2006;3:413-22.
[29]Wagner RF, Metz CE, Campbell GAssessment of medial imaging systems and computer aids: a tutorial review.Acad Radiol.2007;14:723-48.
[30]Krupinski EA, Jiang Y.Evaluation of medical imaging systems.Med Phys.2008 (in press).
[31]孙军,陈峰,郑凯尔．ROC曲线分析在放射学中的应用[J]．中华放射学杂志,2001,35(8)：574-579．
[32]Armitage P, Colton T.Encyclopedia of biostatistics.New York:John , 1998, 3738-3744.
[33]Metz CE.ROC Methodology in radiologic imaging.Invest Radiol, 1986,21:720-733.
[34]Swensson RGUnified measurement of observer performanc in detecting and localizing target objects on images.Med Phys, 1996,23:1709-1725.
[35]Brismar J .Understanding receiver operating characteristic curves : agraphic approach.AJR, 1991 ,157: 1119-1121.
[36]Aldrich JE, Rutledge JD.Assessment of PACS display systems.J Digit Imaging,2005,18:287-295.
[37]Fukushima H, Ikeda M, Ishigaki T et al.Influence of liquid crystal display monitors on observer performance for detection of diffuse pulmonary disease on chest radiographs.Radiat Med (2007) 25:211-217
[38]Krupinski EA,Johnson J,Roehrig H,et al.Use of a Human Visual System Model to Predict Observer Performance with CRT vs LCD Display of Images[J].J Digit Imaging,2004,17(4):258-263.
[39]Balassy C, Prokop M, Weber M et al.Flat-panel display(LCD) versus high-resolution gray-scale display (CRT) for chest radiography: an observer preference study.AJR,2005,184:752-756.
[1]H.Usami,M.Ikeda,T.Ishigaki.et al.The influence of liquid crystal display(LCD)monitors on observer performance for the detection of nodular lesions on chest radiographs.Eur Radiol,2006,16:726-732.
[2]Scharitzer M,Prokop M,Weber M,et al.Detectability of catheters on bedside chest radiographs:comparison between liquid crystal display and high-resolution cathoderay tube monitors.Radiology.2005(234):611-616.
[3]Su Young Kim,Yoon Joon Hwang,Yoon Hee Han,et al.An ROC Study of Chest Radiographs:2K Versus 4K High-Resolution Soft-Copy Images,Journal of Digital Imaging,2007,Vol 20,No 4(December):347-351.
[4]Langer S,Fetterly K,Mandrekar J,et al.ROC study of four LCD displays under typical medical center lighting conditions.J Digit Imaging,2006,19:30-40.
[5]辛海燕,陈长忠,韩文英.医用显示器的选择不容忽视,医疗设备信息,2006,21(8):42-43.
[6]Hakan Geijer,Mats Geijer,Lillemor Forsberg,et al.Comparison of Color LCD and Medical-grade Monochrome LCD Displays in Diagnostic Radiology[J].Journal of Digital Imaging,2007,Vol 20,No 2(June):114-121.
[7]曹军,魏渝清,王学建,等.医学数字化影像显示终端空间分辨率的测试[J].中华放射学杂志,2005,39(10):1105-1106.
[8]严涛,陈钧,王冰.医用专业显示器应用与选型.中国数字医学,2007,8:35-40.