04_学術01.indd

Size: px

Start display at page:

Download "04_学術01.indd"

とよみのたけ
5 years ago
Views:

1 Arts and Sciences CR X Principles and clinical application of multi-resolution image processing in computed radiography Key words: screen film system, computed radiography, unsharp mask processing, multi-resolution image processing summary Medical X-ray imaging creates processed images with subjective contrast. In screen film (SF) systems it is difficult to depict images if the subjective contrast is very high. For example, the thorax contains lungs (high radiolucency) and bone (low radiolucency) and therefore has very high subjective contrast, which results in missing information due to the performance limits of SF systems. On the other hand, computed radiography (CR) systems can work more flexibly than SF systems by using frequency processing for images with very high levels of contrast. To solve this problem, CR systems employ sharpening processing in order to depict a wide range of contrast with a low gradation, using high pass filters that suppress fluctuation of the direct current (DC) component between local images. Before development of CR systems, two output methods were used which preserved the DC component to approximate SF images and sharpened images with high pass filters. However, this style of processing was not universally adopted due to a feeling of degraded reliability created by resampling images with a coarse reading pitch and by using an artificial signal in imaging processing. Therefore, even though improvements in reading the sampling pitch and obtaining actual-sized images have been made, the present mainstream method is just an iteration of former methods that were used to approximate SF. This paper proposes that images that are mainly composed of an alternating current (AC) component are valid for clinical detection of abnormalities also seen in X-ray images that use a fluctuating DC component, such as in X-ray images of the thorax. 4 X screen film system : SF X 1 XX computed radiography : CR CR CRCR SF SF X X X CR XX SF SF CR SF SF 2 3 CR CR Tabuchi Motohiro Depatment of Radiology, Dojinkai Konko Hospital

CR CR SF Imaging plate : IP 4 SF IP CR CR X X 2 3 1 CR 6

b 5 2 Fig.1 2 B m 1.2 1 3 5 7 n f = 1 n 1, 1,.

2 CR CR SF Imaging plate : IP 4 SF IP CR CR X X CR S = S org + w(s org S b ) 1 S S org S b w 1 2 CR 1 2 Fig.1 workflow of multi-resolution image processing S = S org + M w m (S bm-1 S bm ) m Mm w m m= S bm S b 5 2 Fig.1 2 B m n f = 1 n 1, 1,..., 1F F = 1 (n 1)/2 n (1 + 2 k=1 coskω) vol.61 no.744

3 CR X Arts and Sciences amplitude.5 n = 3 n = 5 n = 7 amplitude F(n=1) - F(n=3) F(n=3) - F(n=5) F(n=5) - F(n=7) frequency π -1 frequency π Fig.2 frequency properties with support size of the moving average filter Fig.4 frequency properties of the band pass filter with multi-resolution image processing 4.4 amplitude 1.5 n = 3 n = 5 n = 7 amplitude frequency π frequency π Fig.3 frequency properties with support size of the sharpening filter Fig.5 frequency properties of the synthesis filter with multi-resolution image processing Fig.2 n Fig Fig Fig.4 Fig Fig.4 3 Fig.3n=3 Fig.5 Fig.4 w={.2,.5,.2} 1.3 Fig.6 lena Fig.7 Fig.3 n=3 Fig.8 Fig.5 Fig.5 Fig.8 Fig

Fig.6 lena Fig.7 unsharp mask processing Fig.8 multi-resolution image processing 2 X X 2 1 3 2.1 Fig.9Fig.1 Fig.1 X X Table 1 X XX 7kV X 2mA X 1msec 6 1 CRX 1 m 1bit 6 3 Fig.11 Fig.

4 Fig.6 lena Fig.7 unsharp mask processing Fig.8 multi-resolution image processing 2 X X Fig.9Fig.1 Fig.1 X X Table 1 X XX 7kV X 2mA X 1msec 6 1 CRX 1 m 1bit 6 3 Fig.11 Fig.13 % 1% 1% Fig.9 image Table 1 imaging machinery and materials for image and conditions Fig.1 conventional X-ray image for reading ribs in thorax X X X X X FFD CR KXO-5G DRX-3724HD 7kV 2mA 1msec 1cm 6:1 FCR PROFECT CS 1 m 1bit 214pixels * 176pixels vol.61 no.744

11 Fig.13 3 Fig.11 4 Fig.11 AC component + %component of the image Fig.13 AC component + 1%component of the image Fig.

5 CR X Arts and Sciences 2 1 w={.2,.5,.5, 1., 2., 1.} Fig.11 Fig.14Fig.11 Fig.17 7 Fig.11 Fig.131 Fig.17 Fig.11 Fig.13 Fig.11 Fig.12 Fig.14Fig.15 Fig.16 Fig AFROC alternative free-response receiver operating characteristic AFROC Fig.11 Fig.13 3 Fig.11 4 Fig.11 AC component + %component of the image Fig.13 AC component + 1%component of the image Fig.12 AC component + 1%component of the image Fig.14 subtracted image between the image and the multi-resolution processing image fig

16 subtracted image between the image and the multi-resolution processing image fig.13 Fig.18 trimmed image of the image fig.9 1% Fig.12

6 Fig.15 subtracted image between the image and the multi-resolution processing image fig.12 Fig.17 one-dimensional averaged frequency properties in the horizontal direction of each multi-resolution image, fig.11 fig.13, and its image Fig.16 subtracted image between the image and the multi-resolution processing image fig.13 Fig.18 trimmed image of the image fig.9 1% Fig.12 1% Fig AFROC AFROC 3 4 AFROC Nuclear Associates American College of Radiology ACR X Fig.19Fig.2 ACR X Table 2 conditions of scanning ACR phantom X-ray image Fig.2 ES-85 3dpi 8bit 945pixels * 945pixels ACR X Table 2 ACR X X X ES-85 3dpi 8bit 214pixels * 176pixels ACR X 945pixels * 945pixels ACR vol.61 no.744

CR X Arts and Sciences Fig.19 X-ray image of the mammographic phantom 18-22 with ACR accreditation Fig.

7 CR X Arts and Sciences Fig.19 X-ray image of the mammographic phantom with ACR accreditation Fig.21 sample image for the recognition test of abnormal signals Table 3 conditions of the AFROC experiment 4 CCL 254i2 377lx Fig.2 schema of the mammographic phantom with ACR accreditation XACR X ACR18-22 X 9 8 Fig.18Fig.21 AFROC Table 3 377lx CCL 254i Fig

8 1 AFROC AUC area under the curve 16 d 4 s s n n 3 4 Table 4 Table 5 d = µ s µ n (σ s + σ n )/2 Fig.13 Fig.22Table 6 41 AFROC AUC N N1N1 1%1% 1% N1 Table 7 t p p.5 Table 7 Table 4 necessity of the image for reading thorax Table 8 Fig.23 Fig.24 Table 5 preference of the images for reading thorax Fig.11 1% Fig.12 1% Fig.13 true positive fraction Table 6 AUC + N + N1 + N1 +N +N1 +N false positive fraction Fig.22 AFROC curves AUC vol.61 no.744

9 CR X Arts and Sciences Table 7 itemized list of the number of the recognition of abnormal signals and p value by the t-test + N + N1 + N p probability score +N +N1 +N Table 8 distance between the both probably density functions + N + N1 + N1 4 Fig.13 SF AFROC AUC.5 Fig.23 Fig.24Table 8 X Fig.11 d Fig.23 probably density function of the true positive fraction probability Fig.11 AFROC Fig N +N1 +N score Fig.24 probably density function of the false negative fraction CR X X

10 6 7 Fig.1 Fig.2 Fig.3 Fig.4 Fig.5 Fig.6 Fig.7 Fig.8 Fig.9 Fig.1 Fig.11 % Fig.12 1% Fig.13 1% Fig.14 Fig.11 Fig.15 Fig.12 Fig.16 Fig.13 Fig.17 Fig.11 Fig.13 1 Fig.18 Fig.9 Fig.19 ACR18-22 X Fig.2 ACR18-22 Fig.21 Fig.22 AFROC Fig.23 Fig.24 Table 1 Table 2 ACR X Table 3 AFROC Table 4 Table 5 Table 6 AUC Table 7 t p Table 8 1. p.431,, pp.36-pp.39,, Visual C#.NET & Visual Basic.NET - -. pp.137-pp.161, CQ, CR -., No Michael P. Ekstrom Digital Image Processing Techniques. p.16, Academic Press, FCR., Azriel Rosenfeld and Avinash C.Kak Digital Picture Processing Volume 1 Second Edition (Computer Science and Applied Mathematics). pp.365-pp.369, Academic Press, X --,, 57, pp , ,, 66, pp , Dev P.Chakraborty Comparison of free-response and ROC analyses in modality comparison studies involving lesion localization, Medical Imaging 22: Image Perception Observer Performance and Technology Assessment, Proc. SPIE, vol.4686, pp , DC Edwards et al. Maximum likelihood fitting of FROC curves under an initial-detection-andcandidate-analysis model. Medical physics, pp , Dev P.Chakraborty Recent advances in observer performance methodology: jackknife free-response ROC (JAFROC). Radiation Protection Dosimetry, vol.114, pp.26-31, Extensions to Conventional ROC Methodology: LROC, FROC and AFROC. Journal of the ICRU 8 (1), pp.31-35, Khurd P et al. Ideal AFROC and FROC observers, IEEE Transactions of Medical Imaging. vol.29, pp , F.Zanca et al. Consistency of methods for analysing location-specific data. Radiation Protection Dosimetry, vol.139, pp.52-56, ,. pp ,, vol.61 no.744

06_学術_関節単純X線画像における_1c_梅木様.indd

06_学術_関節単純X線画像における_1c_梅木様.indd Arts and Sciences X The formulation of femoral heard measurement corrected enlargement ratio using hip joints X-ray Imaging 1 2 1 1 1 2 Key words: Bipolar Hip Arthroplasty (BHA) Preoperative planning Enlargement