Proceedings of SPIE(The international Society for Optical Engineering) (San Jose), Vol. 3013, pp149-156, 1997.

Development of a medical image filing system based on super high definition image and its functional evaluation

Hiroshi Takeda, Yasushi Matsumura, Takeo Okada, Shigeki Kuwata
(Dept. Medical Information Science, Osaka University Hospital, 2-15 Yamada-oka, Suita 565 Japan)
Minoru Wada, and Tsutomu Hashimoto
(Nara Research Center, Telecommunication Advancement Organization of Japan, 8916-19 Takayama, Ikoma 630-01 Japan)


ABSTRACT: Although many images are handled in the medical field, image monitoring devices and filing methods vary due to differences in@resolution and pixel depth. If a system enabling such images to be filed synthetically without losing its quality were to exist, and if retrieval of such images were made easily, such a system could be applied in various ways. In order to determine whether a super high definition (SHD) image system running at a series of 2048 resolution x 2048 line x 60 frame/sec was capable of such purposes, we established a filing system for medical images on this system. All images of various types produced from one case of cardiovascular disease were digitized and registered into this filing system. Images consisted of plain chest x-ray, electrocardiogram, ultrasound cardiogram, cardiac scintigram, coronary angiogram, left ventriculogram and so on. All images were animated and totaled a number of 243. We prepared a graphic user interface (GUI) for image retrieval based on the medical events and modalities. Twenty one cardiac specialists evaluated quality of the SHD images to be somewhat poor compared to the original pictures but sufficient for making diagnoses, and effective as a tool for teaching and case study group purposes because of its operability of the retrieval system. The system capability of simultaneously displaying several animated images was especially deemed effective in grasping comprehension of diagnosis . Efficient input methods,and creating a capacity of filing all produced images are future issue.

Keywords: super high definition (SHD) image, medical image filing system, cardiac disease

1. INTRODUCTION

In recent years, the importance of medical images in clinical diagnoses and research, as well as medical education continues to grow. However, there still exist some problems on archive, communication and display of the images which are obtained from various modalities. If these medical images were filed under an integrated digital environment and retrieval procedures were to be made easily and quickly, utility of medical images would become highly efficient. In this study, a "super high definition (SHD) image" technology (2000 pixels x 2000 line x 60 frame/sec. [1][2]) is implemented as a high quality integrated image filing system which meets the needs for clinical practice, education and research. Its objective is to be able to archive and display all image media, both still and motion pictures at its highest quality. In order to test whether the SHD image filing system sufficiently acquires and archives various types of medical images and to verify the clinical usefulness, we created a case file of ischemic heart disease patient which contains all types of medical images, and twenty-one physicians evaluated whether the system architecture and image quality fulfilled the clinical needs.

In this paper, we will describe the hardware architecture and software of this SHD medical image filing system, and will report some results on functional evaluation of the system.

2. SYSTEM OUTLINE

2.1 Hardware configuration

The SHD image format is defined as spatial resolution of 2048 x 2048 and 8 bit depth.The SHD filing system is structured as four parts of data acquisition, data processing and storage, display, and communication. Data acquisition and print out: the film scanner (Leafscan45, Leaf Systems, Inc) and the digital still camera (Leaf Digital Studio Camera, Leaf Systems) are connected to the personal computer (MAC 8100/80). The film digitizer (LD-4500, Konica) and the color printer (Fujix Pictograph 3000, Fuji Photo), are linked to the workstation (Sun SS20). The film digitizer and the digital camera which allows flexible input have spatial resolution of more than 3000 pixel x 2000 line at more than 12 bits depth. After the input image data were processed on the computers, connected to the input device, it was then transferred to the SHD image server via LAN.

Data processing and storage: image server (SHD-1000, Mitsubishi Electric Corporation) and an image database are installed. The server is able to contain 256 SHD images on localRAM that corresponds to 3 Gigabytes data volume. The SHD image data for each RGB signal channel is transmitted through 8 Gbps display bus, and converted to analog output with the rate of 357 mega-samples per second. Frame rate is achieved as progressive 60 frames per second. A video board for simple television meeting can also be connected to this server to digitize NTSC signals. The video signals input into the video board were used as a device to input non SHD images with low resolution.

Display: the SHD monitor (DDM2801,SONY) is used for the display which is able to demonstrate 2048 pixels x 2048 lines and temporal resolution is up to 60 frames per sec with non-interlace RGB 8 bits image. Communication: the SHD system provides an environment for multipoint SHD conference. The system contains the network interface for Ethernet, FDDI and ATM-LAN.

2.2 Multi-image filing

In constructing an image filing system which archives a variety of medical images, a cardiovascular disease case was selected as an example. Numerous medical images including motion pictures are accumulated in cardiovascular disease cases and these images should be integrated for diagnosis and treatment. The following is a description of the selected case and the medical images collected for this study.

(1) The case description

A 64 years old female patient complained of palpitation and chest oppressive feeling at exercise. The patient was hospitalized at Osaka University Hospital where she was diagnosed with anteroseptal myocardial infarction inveterata. Paries dyskinesia was admitted in a wide area of the left ventricle and believed to be the cause of cardiac insufficiency. PTCA (Percutaneous Transluminal Coronary Angioplasty) was performed on the left anterior descending coronary artery whereby symptoms of cardiac insufficiency improved and the patient discharged from the hospital. A month later, anginal attack reoccurred. The patient was rehospitalized and test results indicated reconstriction of the coronary artery in the treated area. A PTCA was reperformed. Both subjective symptoms and test results improved and the patient was discharged.

(2) Types of medical image

The following images were accumulated during the patient's hospitalization. (1)Plain chest X-ray (CX) (2)Rest ECG (RE) (3)Exercise loaded ECG by ergometer (EE) (4)Ultrasound echocardiogram (US) (5)Myocardial scintigram (MS) (6)Coronary angiogram (CA) (7)Left ventriculogram (LV)

This system accumulated a total of 248 SHD images, 243 medical images and 5 text images of the explanation of the history of the case. These images were obtained by conducting the same types of tests during the four time courses of this case (the point of primary hospitalization, after the first PTCA, at reconstriction and after the second PTCA). Details of these images are indicated in Table 1.

2.3 Input and display methods of image modality

The following methods were applied in inputing and displaying each medical image registered.

(1)Plain chest roentgenogram

The 14" x 17" roentgenogram film (CX) were digitized by the film digitizer at 2048 x 2048 space resolution and 8bit depth. All are displayed as 2048x2048 still image.

(2)Electrocardiogram

Both printouts of rest(RE) and ergometerloaded ECG (EE) were directly photographed by the digital camera. All images were 2048 x 2048 still format.

(3)Ultrasound echocardiogram (US)

For this experiment US images stored on videotapes in NTSC signal were converted to frames via video capture board. Each frame was digitized at 640 x 480 pixel depth 8 bits/pixel. The length of one scene was 2 heart beats/scene (60 frames in 2 seconds). A total of 12 images, three types of images, bi-directional section US and color doppler image (2 heart beats of animated images at 640 x 480 resolution/frame) in 4 time course of the disease history were displayed on one screen. Because these images were created by digitizing NTSC signals, each are of poor resolution and low quality. However, by taking advantage of the SHD image server's ample data quantity 12 animated scenes simultaneously were displayed and succeeded in making comparison of several animated images possible.

(4)Myocardial scintigram (MC)

MC is obtained by injecting radioisotopic substance into the patient and capturing the gamma rays emitted from the subjected organ from multiple directions. Generally MC is recorded on film and observed by a set of view box. Gray scale MC images recorded on clear film were brought in using a digitizer. These were 2048 x 2048 still images.

(5)Coronary angiogram (CA) and left ventriculogram (LV)

Contrast media is injected by catheterizing the artery and its flow is observed through as x-ray images. It is generally photographed on cinema film so that animated details can be observed. In this experiment cinema film is digitized frame by frame as SHD animated image for a minimum of one heart. As a result, CA image is 30 frames/scene and LV, 50 frames/scene. Regarding CA images used in this experiment, the three scenes capable of animation (each scene for one heart beat) are all 2048 x 2048 and can be repeatedly displayed in animation. LV images are reduced from its original image 2048 x 2048, to 1024 x 1024 in order to display bi-directional LV in the two time course in one screen. This as in the case of US has the objective of making comparison easy.

2.4 The retrieval program

In order to improve retrieval operability, a GUI (Graphic User Interface) displaying a matrix was developed. The vertical axis indicates the types of medical modalities and the horizontal axis, the time course. By clicking a button on the matrix, users can retrieve and instantly display the chosen modality image from the SHD image frame memory. For images which can be animated, users can repeatedly display for one heart beat period by clicking the image. Users can also choose to fast forward or rewind the animation at desired points. Furthermore by the click of the button, the image can be enlarged by two-fold at selected areas. Images can also be consecutively displayed in accordance to phases of the disease time course or test modalities.

3. SYSTEM EVALUATION

This system has made centralized and instantaneous retreival possible for a variety of high resolution medical images by combining a SHD input and output device with an effective GUI. In order to gain a medical perspective regarding this system, with cooperation from the First Department of Internal Medicine at Osaka, University Medical School, had twenty-one doctors from its medical staff evalaute its operability of GUI for the retrieval system,quality of the images and possibility of medical use.

3.1 Method of evaluation

An evaluation sheet in the form of a questionnaire was used. After operating the system, each evaluator filled out the sheet in a format asking them to evaluate the system in three areas 1) modality-specific operability, 2) image quality, and 3) its practical usefulness. The three areas then consisted of several more questions pertaining to certain details. In evaluating operability of GUI for the image retreival system, evaluators were also asked to grade the 1) operability when choosing a screen, 2) response time of screen display, 3) operability when comparing several images, and 4)functionability of displaying animated images. For each item a five scale evaluation choice was given and evaluators.

In evaluating the quality of this system, evaluators were first asked to grade the 1)image resolution, 2) brightness of image, and 3) contrast of image in comparison to the resolution of other systems. Evaluation remarks were selected from a four point scale. Four, the highest was labled as "equivalent" to the original images. Finally, regarding the first three questions, the evaluators were asked to give a grade between one to five for an overall evaluation of the operability and quality of the system.

The third part regarding evaluation was concerning about the practical use of the system . Evaluators were asked to grade the 1) possibility of use as an teaching system, 2) effectiveness in case conferences, and 3) use as a database for clinical research.

3.2 Results of evaluation

These results were obtained using the CCIR method taking the following steps. Each question had a selection of five (or four) grades. Each grade was scored from 1 to 5 (1 to 4) points. Scores given by all twenty-one evaluators were then added and an average point in each section was calculated.

3.2.1 Results of evaluation of each image modality

The evaluation results regarding operability of GUI for image retreival system are summarized in figure 1. The operability was graded as "rather good" for RE, EE and LV, and "average" for others. Figure 2 indicates the results of image quality when compared with other systems. Resolution was graded "equivalent" for CX, electrocardiogram (EC), CA, LV, and "rather poor" for US and MS.

Figure 3 summarizes the sufficiency of image quality in making diagnoses. EC and LV was graded "possible," CX, US, MS as "almost possible". Figure 4 indicates the comprehensive evaluation regarding the operability and image quality. LV was evaluated as "good," EC, US and CA as "rather good,"and "average" for MS.

3.2.2 Results of evaluation for practical use

The results of evaluation for practical use are indicated in figure 5. Use in education system and study groups were evaluated as "can be used" and use as clinical research database was evaluated as "can mostly be used."

4. DISCUSSION

The number and types of medical images produced during treatment of even one patient are massive. How efficiently these images can be utilized as tools for daily examinations and education is a critical issue in the medical field. This experiment examines whether a SHD image system enables various types of medical images to be integrated and utilized. A medical image filing system centered on the SHD system was constructed and all types of images produced during treatment of one cardiac disease case were registered. Furthermore, the system was evaluated focusing on simplicity of search monitoring and image quality. Doctors also shared opinions regarding the system's effectiveness in the medical scene.

All images are stored in frame memory and waiting time for the image to be displayed is extremely short. The interface which presents seven types of images in a matrix of four phases helps the user have command of the orientation of several images and easily select the desired image. The system is also capable of displaying animated images without losing its high quality. Doctors highly evaluated the system on the above points. When evaluating therapeutic effectiveness for patient care, physicians' need to compare two or more images of different phases on the same screen surfaces. Out of US and LV and other images which are generally poor in quality, two or more selected images with different phases were simultaneously displayed on one screen. Its convenience in observation was graded highly. In regard to images requiring high resolution, however, this function is not supported and this problem was pointed out. To resolve this problem additional functions which either reduces the size of each image and positions several images on one screen (although this would reduce image quality) or, cuts optional parts of each image and places them on one screen would be necessary.

As for resolution, CX, CA, LV received high evalution and US and MS received low evaluation. Original images of Us and MS have poor resolution to begin with and should have no problem being displayed as a SHD image. This however is contradicted by its evaluation. There may be a problem in the images' input methods, but the psychological mind desiring high resolution in the original image may have also been an influence. On the other hand, original images of CX, CA, LV have high resolution. Therefore, the fact that these were evaluated highly indicate that display of these images are possible by SHD format. Among the images used in this experiment CA required the most detailed distinction, but was evaluated to be suffucient enough for diagnosis on this system image. The above suggests that the SHD image format is sufficient in terms of resolution at least for the types of images handled in this experiment.

This system unfortunately did not receive much credit in terms of use as a clinical research database. In the case of handling medical images there are issues such as how to digitize the mass amounts of images and how to store that data besides the excellence of the monitoring device. In terms of this system however, the problem is considered to be that a basic solution on image input and accumulation has not yet been proposed. On the other hand, for educational systems which are fairly effective with a relatively limited number of images and diagnosis study groups the system was evaluated as functional. This suggests that its function as a printout device of SHD images is evaluated highly.

5. CONCLUSION

We have described the architecture and shown the evaluation results of a medical image filing system which utilizes the SHD image system. Medical images were obtained from a cardiac disease case. In this experiment we have illustrated that upon integrating various types of medical images including animated ones, the SHD image structure was sufficient and capable of establishing a useful medical image filing system. Furthermore, we have exemplified that the search GUI we prepared was extremely effective in this image filing system.

Although only medical images of a cardiac disease case were used in this experiment, the methods applied are considered to be effective for other cases. In future experiments, use of images which illustrates the brilliance of SHD image colors are being planned in addition to the gray scale images which were frequently used this time.

This experiment has illustrated the potential of the SHD image system to expand as an effective image handling/displaying technology in the medical field. In the future we plan on utilizing wide range ISDN for remote operation of this system, and continuing use of SHD image in medicine, including joint experiments from remote areas as well as evaluative experiments.

REFERENCES

1. S. Ono, "The super high definition image", TV Society Magazine, 47(9), pp.1204-1215. 1993

2. T. Murakami, "Super high definition image system', Telecom Forum, Oct., 1995