JP2002330951A - Image encoding/decoding device and method, computer program and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To discriminate a concerned area in a medical image from the other areas and to reproduce the area distinctly. SOLUTION: The medical image such as an X-ray image inputted from an image input part 1 is diagnosed by a support diagnosis processing part, and when there is found an affected part, its positional and dimensional information is extracted as of the concerned area. This information is supplied to an encoding part 3. The part 3 converts the image to the data of sub-band component of each frequency and separates the component data in the concerned area from the data outside of the concerned area for encoding. Then, the information at least identifying the concerned area from the other areas is added and outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an image processing apparatus and method for encoding and decoding medical images such as X-ray images and a storage medium storing the method.

[0002]

2. Description of the Related Art Radiation (X-ray, α-ray,
(β-rays, γ-rays, electron beams, ultraviolet rays, etc.), a part of this radiation energy is accumulated in the phosphor, and when this phosphor is irradiated with excitation light such as visible light, It is known that phosphors exhibit stimulated emission. Phosphors exhibiting such properties are called stimulable phosphors (stimulable phosphors). Utilizing this stimulable phosphor,
The radiation image information of a subject such as a human body is temporarily recorded on a sheet of a stimulable phosphor, and the stimulable phosphor sheet is scanned with excitation light such as a laser beam to generate stimulating light, and the obtained stimulating light is emitted. A radiation image information recording / reproducing system which photoelectrically reads emitted light to obtain an image signal, and outputs a radiation image of a subject as a visible image to a recording material such as a photographic photosensitive material or a display device such as a CRT based on the image signal is provided. It has already been proposed by the present applicant (JP-A-55-12429, JP-A-56-11395, etc.).

[0003] In recent years, an apparatus for similarly photographing an X-ray image using a semiconductor sensor has been developed. These systems have the practical advantage of being able to record images over a very wide radiation exposure area compared to conventional radiographic systems using silver halide photography. That is, an X-ray having a very wide dynamic range is read by a photoelectric conversion unit and converted into an electric signal, and the electric signal is used as a visible image on a recording material such as a photographic photosensitive material or a display device such as a CRT using the electric signal. By outputting, a radiation image which is not affected by the fluctuation of the radiation exposure amount can be obtained.

[0004] The radiation image thus obtained is output from the imaging device as a digital image signal, stored, transmitted or displayed and used for diagnosis. On the other hand, with the spread of such digitalization of medical images, computer-assisted diagnosis (Co
Research and commercialization of mputer Aided Diagnosis (hereinafter CAD) are being actively conducted. The CAD is expected to have an effect of preventing an operator from overlooking an abnormal shadow caused by a fatigue of an interpreter or a difference in interpretation ability when a large amount of images must be interpreted as in a mass examination.

[0005]

However, a medical image, particularly an X-ray image, has a very high resolution and a pixel precision of 12 bits or more per pixel. However, there is a problem that the cost is increased to store or transmit the data. In addition, in the above-described example of the group examination, since a large number of images are to be interpreted, it is necessary to rapidly perform a series of processes related to the region of interest extraction and image display by CAD. The present invention also provides a technique for automatically generating a region of interest based on text information such as a doctor's diagnosis result (primary diagnosis report) or a medical history. Here, the diagnostic information is not limited to text, but also includes audio information.

According to the present invention, when a medical image having a region of interest is reproduced, the image encoding device, the decoding device, and the method enable to reproduce the region of interest and the other region separately on the reproduction side. And a computer program and a storage medium.

[0007]

In order to solve this problem, for example, an image coding apparatus according to the present invention has the following arrangement. That is, an image coding apparatus that sets a region of interest in a given medical image, or encodes an image having information on the region of interest by using a unit that detects the region of interest, comprising: Conversion means for converting the data into sub-band component data for each component; and a code for separating and coding the component data in the region of interest and the component data outside the region of interest in the component data obtained by the conversion means. And an output unit for adding at least information for identifying a region of interest and an area outside the region of interest to the encoded data encoded by the encoding unit, and outputting the encoded data.

According to a preferred embodiment of the present invention,
An image processing apparatus according to the present invention includes: an image input unit that inputs a medical image; a support diagnosis unit that performs a support diagnosis analysis that analyzes the input image and extracts a candidate region of a lesion; and encodes the image. An encoding unit that performs processing and outputs a code sequence related to the image; a storage unit that stores the code sequence by the encoding unit; and an output unit that outputs the stored code sequence. Is a region of interest output by the support diagnosis unit, and configures an output code string based on the information of the region of interest.The output unit converts the code string stored in the storage unit based on a predetermined external input. It is characterized by outputting.

Further, the image processing apparatus according to the present invention, when decoding the code sequence and displaying an image, code input means for inputting the code sequence generated by the coding means, and stores the code sequence. Storage means, decoding means for decoding the code string and outputting a decoded image, extraction means for extracting position information of the region of interest from the code string, and displaying the decoded image based on the position information. It is characterized by comprising display means for adding a predetermined marker to the decoded image and displaying the decoded image.

[0010]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings.

<First Embodiment> FIGS. 1 and 2 show an image processing system (FIG. 1 shows an apparatus on the encoding side and FIG. 2 shows an apparatus on the decoding side) according to the present embodiment. .

In FIG. 1, an image input unit 1 includes one or more X-ray (medical) images (not limited to a simple X-ray image) taken by a predetermined method, an X-ray tomographic image by X-ray CT, an MRI
May be input in sequence, and output to the subsequent support diagnosis processing unit 2. The image input unit 1 may be a predetermined imaging device that generates a digital image signal, or a reading device for a recording medium such as an MO that stores an image captured in advance by such an imaging device.

[0013] The support diagnosis processing unit 2
A region considered to be an abnormal shadow is extracted as a region of interest (hereinafter, ROI: Region Of Interest) by a predetermined method.
Any algorithm can be used for the extraction, and details thereof are omitted. For example, “Computer-Aided Diagnosis Sch
eme for Detecting Pulmonary Nodules Using WaveletT
ransform ", Proc SPIE2434: Medial Imaging, pp.
621-626, San Diego, February, 1995. Also in the present embodiment, description will be made assuming that this is adopted.

The support diagnosis processing unit 2 outputs information on the image to be detected and the position and size of the detected ROI to the subsequent encoding unit 3.

The encoding unit 3 compresses and encodes an input image based on the input information (position and size) relating to the ROI, and outputs a code string. In the present embodiment, the encoding method is ISO / IEC15444 (hereinafter, JPEG2
000), but the details of the encoding procedure are detailed in the JPEG2000 standard document. Therefore, only the outline and the necessary items will be described here.

FIG. 3 shows the configuration of the encoding unit 3. The image signal is transformed by the discrete wavelet transform unit 301, and the transform coefficient is quantized by the quantization unit 302, and the quantization index is output. However, quantization is not performed when the encoding mode is lossless encoding, and the transform coefficient is output as it is to a subsequent process as a quantization index.

On the other hand, the ROI extracted by the support diagnosis processing unit 2 is input to the ROI setting unit 303. The ROI setting unit 303 determines, based on the input information about the ROI, R in the input data supplied from the quantization unit 302.
The quantization index included in the OI area is shifted up by a predetermined number of bits.

FIG. 4 shows the state of the quantization index at this time. In FIG. 7A, a portion surrounded by a circle is a quantization index in the ROI, and belongs to an area specified by the support diagnosis processing unit 2 in the preceding stage. This quantization index is shifted up to a level where the bit plane does not overlap with the surrounding quantization index as shown in FIG. 7B, and is output to the entropy encoder 304 following. For example, in the quantization process, when the value of the maximum quantization value of each index is 4 bits, the index in the ROI is shifted upward by 4 bits.

The entropy coding section 304 performs bit plane coding on the input quantization index, and outputs the coded data to the subsequent code string forming section 305. The code sequence forming unit 305 forms the input coded data into a final code sequence and outputs the result. Since the index in the ROI is shifted up by 4 bits in the case of FIG. 4B, there are eight bit planes (bit 7 plane, bit 6 plane,... Bit 0 plane). Become. Each of the bit planes is encoded.

In JPEG2000, five progressive forms are possible as a method of forming a code string.
In the present embodiment, the Layer-Resolution level-component-pos
ition. Therefore, the encoded data of the quantization index belonging to the ROI in the code string is always located before the encoded data of the quantization index outside the ROI.

The code string generated in this manner is output to the storage unit 4 and is temporarily stored. The above operation is executed for each of a series of images output from the image input unit 1,
A code string for each image is stored in the storage unit 4. Note that R
If the OI exists, the number of bits shifted (ROI shift amount s) is stored in the RGN marker.

Next, the operation of transmitting, decoding and displaying the code string stored as described above will be described.

FIG. 2 shows the configuration of a display system for decoding and displaying an image. Prior to image interpretation, transmission of a code string is instructed by instruction means (for example, a keyboard or a mouse) (not shown), and the input control unit 8 requests the output control unit 5 to output a code via the transmission path 7. The output control unit 5 instructs the output unit 6 to read and transmit the code string from the storage unit 4. Thereby, a series of previously stored code strings is transmitted to the code input unit 9 via the transmission unit 7. Here, the transmission unit 7 is a network such as a LAN or the Internet.

In the above description, when there are a plurality of images, all the code strings may be transmitted in response to one transmission instruction, or the processing for one code string described later is terminated. After that, the next code string may be transmitted. However, in the following description, it is assumed that a plurality of code strings are transmitted at a time and temporarily stored in the storage unit 10.

The decoding unit 11 reads out the previously transmitted code string from the storage unit 10 and decodes the code string by an instruction means (not shown). FIG. 5 is a diagram illustrating the configuration of the decoding unit 11. The input code string is output to a subsequent entropy decoding unit 1102 via a code string analysis unit 1101. The code sequence analysis unit 1101 scans the input code sequence and outputs information necessary for the decoding process to the control unit 1105.

Here, the decoding process is performed by a method standardized by JPEG2000.
Outputs a ROI shift amount s attached to the RGN marker to the control unit 1105 when detecting the maximum number of bit planes Mb of each subband and an RGN marker indicating the existence of the ROI when scanning the code string.

On the other hand, when decoding the input code sequence, the entropy coding unit 1102 appropriately shifts the quantization index located in the ROI downward from the Mb and the number of decoded bit planes (by 4 bits in the above example). (Shift) and outputs the result to the inverse quantization unit 1103. The position of the quantization index at this time is output to the control unit 1105 at the same time. The control unit 1105 stores the position of the quantization index to which the ROI thus obtained belongs.

The quantization index output from the entropy decoding unit 1102 is inversely quantized by the inverse quantization unit 1103, and the discrete wavelet transform coefficient is restored. The discrete wavelet transform coefficient is subjected to an inverse transform in an inverse discrete wavelet transform unit 1104, and an image signal is restored and output to the display unit 13. On the other hand, the position of the quantization index belonging to the ROI is output to the display control unit 12.

The display control unit 12 controls the display unit 13 to overlay and display a marker indicating the position of the ROI on the image signal input from the decoding unit 11. At this time, if there is no ROI in the input code string (RG
(Eg, when the ROI shift amount s associated with the N marker is 0)
Does not add any marks. Thus, the display unit 1
The image displayed in 3 is used for image interpretation.

A specific application example of the above-described embodiment will be described with reference to FIG.

In the figure, 100 corresponds to the encoding device shown in FIG. 1, and 200 corresponds to the decoding device. The apparatus 100 is provided, for example, in a medical car that performs X-ray imaging used in a group examination or the like, and the apparatus 200 is assumed to be installed in a medical institution (an institution with a doctor, such as a hospital). Easy to understand.

In the apparatus 100, a medical X-ray image or an X-ray imaging apparatus 102 stored in a storage medium 101 in advance.
Has a CAD 106 for detecting or setting a region of interest ROI from the X-ray image (digital image) of the subject obtained in step (1). Although 103 and 105 shown in the figure are an encoding unit and a decoding unit, they are provided in consideration of the transfer efficiency until the image is transferred to the CAD 106, and are not necessarily essential. The CAD 106 employs a technique of automatically detecting a candidate of a region of interest as described above,
Alternatively, a technician or doctor sets the region of interest candidate while looking at the monitor. The number of regions of interest to be detected or set is not limited to one, and a plurality of regions may be present.

This result was obtained by using JPEG2000,
Encoding is performed by the encoding unit 107 in the above embodiment,
The result is temporarily stored as a JPEG2000 compression-encoded file with ROI. Then, when a certain number of sheets (or the amount of data) are prepared, the data is transferred to the apparatus 200 (for example, a medical institution) on the side of actually examining.

The apparatus 200 temporarily stores the plurality of compressed and encoded files 201 transferred as described above in a storage device such as a hard disk. The device 200 includes:
As described above, a unit (decoding unit 202) for realizing the decoding process shown in FIG. 2 is provided, and the example shown in the drawing is implemented in a web browser by a program as a plug-in. With such a configuration, the decoding unit 202 reads the file specified by the operator, decodes the file, and generates display image data so that the ROI can be distinguished. The display control unit 203 performs a process of displaying the generated image on the display unit 204.

The system configuration shown in FIG. 12 can be similarly applied to the embodiments described below.

<Second Embodiment> In the first embodiment, encoding is performed by shifting up the quantization index of the ROI part.
It is also possible to separate the I part from other parts. Less than,
A description will be given as a second embodiment.

In the second embodiment, the region of interest ROI extracted by the support diagnosis processing unit 2 is included in a predetermined layer in a hierarchized code string. FIG. 6 illustrates a schematic configuration of a code string according to the second embodiment. As shown in the figure, the code sequence starts with a main header MH including information on the size of an image to be coded, components, and the like, and coded data BS is arranged following the tile part header TH. In the second embodiment, since the image is not divided into tiles (since the entire image is one tile), the tile part header TH and the encoded data BS are composed of one.

In the second embodiment, the progressive form is the same as that of the first embodiment, and the coded data BS is arranged with the upper layer Layer0 at the top as shown in FIG. In this arrangement, Layer 0 is configured for coded data relating to the region of interest. Furthermore, the encoded data of each layer includes a packet header P including information on packets PK relating to a plurality of resolutions and code blocks included in the packets PK.
H. For these configurations, please refer to JPE
Since the details are described in the G2000 standard document, the description is omitted here.

FIG. 7 is a diagram for explaining such a layer configuration method. In FIG. 7A, a portion surrounded by a line is a region of interest extracted by the support diagnosis processing unit 2 (FIG. 7B).
(The part indicated by ○). FIG. 2C shows the subband configuration and code blocks when the discrete wavelet transform is performed on the same image, and the shaded code blocks include a region of interest.

When constructing a code string, the encoding unit 3
As shown in (A), the uppermost Layer 0 includes only the coded data of the code block including the region of interest shown in FIG. On the other hand, if the region of interest does not exist in the target image, Layer 0 does not actually include the encoded data of the code block, and
0 is set in the zero length packet field.

The operation at the time of decoding when the code string is configured as described above will be described below.

The code string analyzer 1101 scans the packet header PH of the coded data for Layer 0 to determine whether or not the image includes a region of interest. When the zero length packet field of the packet header PH is all 0, it is determined that the image does not include the region of interest, and normal decoding processing is performed, and the image is displayed.

On the other hand, when the encoded data is included in Layer 0, the subsequent operation is controlled on the assumption that the image includes the region of interest.

The decoding process is performed sequentially from the upper layer to the lower layer.
When decoding is performed at r0, the position of a code block belonging to a predetermined resolution level of the layer is read from the packet header PH and output to the control unit 1105. Here, the predetermined resolution level may be any level. For example, by selecting the highest resolution sub-band in FIG. 7C, the position of the region of interest in the image can be specified more accurately. Can be.

For example, in FIG. 7C, B1, B
The encoded data of the three code blocks of 2, 3 is Laye
It is included in r0, and the position of the code block can be calculated from the image size and the code block size included in the main header MH.

Each layer is subjected to normal decoding processing, and the restored image signal is output to the display unit 13. At this time,
The control unit 1105 instructs the display control unit 12 to overlay display a marker on a portion corresponding to the position of the code block including the region of interest, generates a marker on the display device 13, and displays an image to which the marker is added. Output.

<Third Embodiment> In the above-described embodiment, the code sequence is arranged so that the coded data relating to the region of interest is arranged in a specific portion of the code sequence. May be included in the column.

FIG. 8 is a diagram showing additional components in this case, where "1" is assigned as a pixel value to a portion corresponding to the region of interest, and "0" is assigned to the other portions as pixel values. What is necessary is just one bit per pixel.

By adding such components to the code string, the number of components becomes 2 when the image is monochrome gray scale. When the region of interest is not included, the number of components is 1, and the presence or absence of a region of interest can be checked by checking the number of components described in the main header MH at the time of decoding.

When a component of one bit per pixel is coded as described above, the decomposition level of the discrete wavelet transform is set to 0 in the coding to prevent an increase in the dynamic range due to the conversion. Efficiency reduction can be suppressed.

At the time of decoding, the position of the region of interest may be specified by examining the value of the pixel in the additional component, and a marker may be added to the decoded image and displayed.

<Fourth Embodiment> In the above-described embodiment, the addition of information relating to a region of interest is realized as a method of forming encoded data. In addition to this, necessary information is added to a code string. May be.

FIG. 9 shows a format of data stored in the storage unit 4 in this case. In the figure, a predetermined header BH is added to the code string generated in the previous embodiment, and a description about the region of interest is stored in the same file as additional information DSC.

Here, the additional information DSC is information such as parameters and detection accuracy used in the support diagnosis analysis, which are extracted by the decoding unit 11 at the time of interpretation, and are used as reference information by the display control unit in the display control unit. It is additionally displayed on the display unit 13 in the same manner as the mark addition.

<Fifth Embodiment> In the above-described code string configuration, the coded data related to the portion where the region of interest exists is arranged in the first half of the layered code string.
The encoded data to be transmitted may be partially omitted based on the result of the support diagnosis.

FIG. 10 shows the structure of a code string at this time. FIG. 10A shows a code string when a region of interest is extracted by the support diagnosis processing unit 2, and FIG. This shows a code string when no area exists.

Normally, when a compression-encoded image is used in interpretation such as a group examination, irreversible encoding to the extent that there is no problem in interpretation may be permitted. In this case, all encoded data is transmitted. No need. The code string in FIG. 7B does not include the lowest layer, and can obtain a higher compression rate. The encoded data included in the lowest layer includes, for example, a code of a code block or a tile belonging to a part that is not directly related to interpretation such as an abdominal part, an axillary part, or a blank area in a chest radiograph. I have.

On the other hand, in the code sequence of the image in which the region of interest is present as shown in FIG. 9A, all the encoded data are transmitted, so that in the interpretation, the image obtained by adding the marker to the completely restored image is read. Can be used.

As a method of changing the compression ratio according to the presence or absence of the region of interest, the image may be divided into tiles and the quantization step may be changed for each tile.

At the time of encoding, the number of layers is included in the header, and the decoding side determines the presence or absence of a region of interest based on the number of layers, and performs decoding and image reproduction.

<Sixth Embodiment> In the above embodiment, the code string as a unit of transmission is obtained by encoding one image. However, the present invention is not limited to this. The present invention can be applied even when an image sequence composed of a plurality of images is used as a unit.

FIG. 11 shows the structure of a file stored in the storage unit 4 when the present invention is applied to an image sequence composed of a plurality of images. In the present embodiment, the images included in the storage file are chest X-ray images of the same subject, and are obtained by photographing the same part at different times.

The image input unit 1 sequentially reads the plurality of images, and outputs the current image to the support diagnosis processing unit 2 and outputs the other images directly to the encoding unit 3. Thus, the support diagnosis analysis process is performed on the newest image, and the region of interest is extracted by the above-described method to form a code string. The method of constructing the code string at this time may be any of the methods described above.

If the region of interest is present in the current image, its position is stored in a storage area (not shown), and when the past image is coded, the position is coded as the region of interest. You. The code string obtained in this way is collected as one storage file, stored in the storage unit 4, and transmitted.

At the time of image interpretation, these code strings are sequentially decoded by the above-described method, and when displayed on the display unit 13, a marker is added to a region of interest of each image.
In the present embodiment, the support diagnosis processing is not performed on the past image, but the processing may be performed on a predetermined image or all target images. Or,
As the support diagnosis processing, the latest image may be compared with one or more past images.

In the present embodiment, the chest X-ray image is taken as an example. However, even when a sequence using a plurality of other images as a unit, for example, a plurality of tomographic images by CT is used, each tomographic image can be obtained. The present invention can be applied by performing the support diagnosis processing and the encoding on the image in the same manner.

<Seventh Embodiment> A seventh embodiment will be described. In the following, a description will be given including the processing procedure for determining the ROI, but it is needless to say that the method for determining the ROI may be applied to each of the above embodiments.

FIG. 13 is a block diagram of a device on the encoding side and FIG. 14 is a block diagram of a device on the decoding side in this embodiment.

The image input means 501 may be an input from a network or an image from a digital X-ray imaging apparatus using a stimulable phosphor or a semiconductor sensor. The input here assumes a data array as original image data, and encoding such as compression is not performed.

On the other hand, diagnostic information for the input image is input to the diagnostic information input means 504. Although it is considered that the diagnostic information is attached to the diagnostic image as text information, the diagnostic information exists in a device connected via a network such as a hospital information system, a radiation information system, and a patient diagnostic information server (all not shown). It is also possible. The diagnostic information input unit 504 fetches the diagnostic information by interpreting the image data stream when the diagnostic information is attached to the image, and using the network search unit when the diagnostic information is present on the network.

In the present embodiment, it is assumed that the image is a digital image of a chest front X-ray, and that the diagnostic information (diagnosis report) is that "calcification is recognized at the upper right of the right lung". The input diagnostic information is sent to the diagnostic information interpreting means 505.

FIG. 15 is a functional block diagram of the diagnostic information interpreting means 505, which interprets diagnostic information and outputs an interpretation result 526. The interpretation result includes a segment number and area division information. What is a segment number?
This is a predetermined anatomical number as shown in FIG.

In the case of the front of the chest, the right lung, the left lung, and the mediastinum are assigned to segment numbers 1, 2, and 3, respectively. As shown in FIG. 25, the area division code is information for specifying each anatomical site as an area (location) inside the site.

The interpretation result 526 is sent to the region of interest determining means 503 shown in FIG. In this example, the interpretation information is segment number 1 (right lung), and the area information is the upper right. On the other hand, an image is also supplied from the image input unit 501 to the region of interest determination unit 503. The region of interest is determined by the region of interest determination means based on the interpretation result of the image information and the diagnosis information.

FIG. 24 shows a process of determining the region of interest means.
The region of interest determination means 503 has a different data processing process between the learning process and the use process, but the determination process indicates the use process. The image is segmented according to the segment numbers shown in FIG. 22 (FIG. 22A), and the right lung corresponding to the segment number 1 is labeled from the segments (FIG. 22B). Elliptic approximation (Fig. (C)) and centroid calculation (Fig. (D)) are applied to divide the area of the labeled image. The result is applied to the labeling image (FIG. 13B), and the upper right of the right lung is determined as the region of interest.

By providing the region of interest determined above to the encoding means 502 of FIG. 13, encoding is performed with the region of interest set.

The encoding means 502 preferentially encodes the region of interest by bit-shifting, as in the embodiment described above. In this case, in the JPEG2000, it is possible to first selectively display only the region of interest by applying the method called the MAX shift method described in the first embodiment. As the second means for realizing the region of interest, the tile corresponding to the region of interest or the CODE
The coding layer (Laye
r) There is a way to set.

The coding layer will be described. The code strings are arranged in order from the upper layer to the lower layer,
Each layer includes a code related to each sub-band subjected to DWT, and a code related to each sub-band includes a packet header PH including a code block included in the sub-band and designation of an encoding path, and a code of each code block. It is included.

FIG. 26 illustrates the relationship between layers and compression ratios. In the figure, the code string is composed of 10 layers, and when all the layers from the highest layer 9 to the lowest layer 0 are decoded, lossless decoding is possible.
On the other hand, the code amount related to the uppermost layer 9 corresponds to 1/20 of the data amount of the input image, and the code amount related to layers 9 to 7 corresponds to 1/10. Encoding section 502
Constitutes a layer such that the code amount corresponding to the predetermined compression ratio corresponds to the boundary of the predetermined layer.

According to the Max shift method described above, the DWT coefficient of the region of interest is always coded as a bit plane higher than the DWT coefficient of the non-region of interest. Layer if
9) is performed. Further, in addition to preferential encoding of the region of interest, it is conceivable to add the determined outline (outline) of the region of interest to the image as overlay information. The data encoded as described above is transferred to an image server on a network or a diagnostic image display device (for example, the device 200 in FIG. 12) via an image output unit.

The region of interest determined above is encoded by the encoding means 50.
By outputting the data to 2, the coding is performed with the region of interest set. In the present embodiment, the encoding unit 502
According to IEC15444 (hereinafter JPEG2000), the present invention is not limited to this.
In O / IEC 10918 (JPEG) or the like, a method of setting a low compression ratio for the region of interest may be used.

FIG. 27 shows a basic configuration of a JPEG2000 encoder. An image signal input by an image input unit 401 is transformed by a discrete wavelet transform transform unit 402, and And a quantization index is generated. The quantization index is bit-plane coded in code block units in the entropy coding unit 404, and a code sequence in a predetermined progressive form is configured in the code sequence configuration unit 405. Since a series of these processes is detailed in the JPEG2000 standard document, detailed description is omitted.

In the present embodiment, the image including the region of interest obtained by the above-described method is the RO of JPEG2000.
It is encoded using I. The ROI is realized by shifting up the quantization index by a predetermined amount before performing entropy coding in the system of FIG.

FIG. 28 shows the relationship between the ROI at this time and the quantization index belonging to outside the ROI. The quantization index in the ROI is located in the upper bit plane with respect to the quantization index outside the ROI. become.

Here, the region of interest obtained by the above-described method is encoded as an ROI, and in the code sequence generated by the code sequence forming unit 405, the coded data relating to the region of interest is arranged in the first half of the code sequence. Is done. In such a configuration, the encoding process is terminated at a predetermined bit plane lower than the bit plane including the ROI, or all the bit planes are encoded. (Excluding such bit planes), the ROI on the receiving side
It becomes possible to restore an image in which only parts are high definition.

In JPEG2000, the above-mentioned region of interest can be preferentially encoded by a layered encoded data structure called a layer. FIG. 29 shows the state of the layer configuration at this time. In FIG. 29, code blocks 3 and 4 are included in the region of interest.
The code sequence forming unit 405 determines only the code of the code block included in the region of interest based on the input information of the region of interest.
A code string is included in yer9.

Here, assuming that the progressive form is Layer-Resolution level-Component-Position in JPEG2000, the layout of the layers in the entire code string is as shown in FIG. Only encoded data is included.

If quantization is not performed in the quantization process of the encoding process, as shown in FIG. 26, lossless encoding can be performed as a whole while making the compression ratio correspond to each layer. It is possible. Here, the top layer (La
Since yer9) includes only the region of interest, if the size of the region is different, the encoded data length is different.

The data thus encoded is transmitted to an image server on a network or a diagnostic image display device (not shown) via the image output means. Here, by arranging the coded data relating to the region of interest in the first half of the code sequence and transmitting a predetermined amount from the beginning of the code sequence to the display device side, an image restored with high image quality is obtained for the region of interest. be able to.

On the decoding side, a code string including a region of interest can be decoded and displayed as follows.

First, when the region of interest is coded as an ROI, the generated code string contains an RGN marker indicating the existence of the ROI, so that the code string contains the region of interest prior to image display. I understand. Regarding the decoding process of the coded data including the ROI, the coefficient related to the ROI can be specified, the portion can be decoded with high image quality, and a marker can be displayed by a method defined in the JPEG2000 standard.

On the other hand, when the coded data of the region of interest is included only in the uppermost layer, the decoding side analyzes the header of the first packet (a unit of the code string configuration based on the coded data of the layer) and analyzes the packet. Is zero length pa
If the packet is not a cket (a packet in which no encoded data exists), it is interpreted that the code string includes a region of interest, and a region occupied by a code block included in the packet is set as a region of interest, and a marker is attached to the decoded image. indicate.

Here, as described above, since the first layer contains all the encoded data of the code block belonging to the region of interest, the image of the portion can be reversibly reproduced.

Since the size of the code block is the same at each resolution level, it is assumed that the resolution level to which the code block specifying the size of the region of interest belongs is predetermined.

According to the method described above, it is possible to arrange the coded data relating to the region of interest in the first half of the code string, display the portion with high image quality, and attach a marker to the display image. Further, if the inverse discrete wavelet transform at the time of decoding does not perform all level number inverse transforms, it is possible to reproduce and display the reduced image.

Further, the input image may be divided into rectangular areas (tiles) of a predetermined size, and whether or not each tile is a region of interest may be encoded. In this case, the encoding method
A method other than JPEG2000 can be used. Further, in addition to preferential encoding of the region of interest, the determined outer shape (outline) of the region of interest may be formed as another component as overlay information and added to the code string.

Next, the decoding side will be described with reference to FIG. The image input means 511 is based on encoded image data. In the following description, it is assumed that a DWT-converted image is converted into a coded stream and the image is input. However, the technology is not limited to the DWT conversion, and a series of DCT, KL conversion, etc. An application example of encoding a converted image can also be considered.

[0098] The encoded image data is transferred from an image server (not shown) on a network (including the Internet), but the data input by the image input means 511 is transferred to the decoding means 512. The decoding means for decoding the DWT image will be described later. The output image 513 of the decoding means 512 is displayed via the image display means 515, and is provided to the region of interest determination means 503, or the region of interest determination means 503 takes in the image 515. The diagnostic information input means 504 and the diagnostic information interpreting means 505 leading to the region of interest determining means 503 are the same as in the first embodiment.

It should be noted here that it is not necessary to read all the converted data streams in order to obtain an image having a resolution and image quality enough to be used for the region of interest determination means. A data amount of about 1 is sufficient.

In the following description, it is assumed that an image sufficient for reduced display is transferred with a data amount of about 1/10 of the data for the target image. When about one tenth of the image is input, the region-of-interest determination unit 503 determines the region of interest also using the diagnostic information, as described above. The determined region of interest is overlaid on the reduced image displayed on the image display unit 515 as outline information, and provided to the data input control unit 514.

The data input control means 514 uses the same technique as that described in the section of coding, and uses a tile area for a region of interest,
Alternatively, the image server is designated to preferentially transfer data by designating the CODE BLOCK area. The selective data request to the image server is performed via the image input unit. However, depending on the configuration of the image server, all the codes of LOSSLESS may be stored, and the image server may be dynamically requested in response to the request from the image input unit. It is also possible to configure a stream such as a layer (dynamically). This example will be described with reference to FIG.

FIG. 30 is a diagram showing a configuration in a case where a code string is dynamically generated between the image input means 511 and the image server SV. In the figure, the image input means 511 transmits the position of the region of interest in the image and the transmission request of the relevant portion to the image server SV, and the image server SV converts the code string stored in the external storage device ST in advance. The necessary part is transmitted to the image input means 511. Hereinafter, an example of the output form of the code string related to the region of interest in this configuration will be described.

FIG. 31A shows code string transmission when an image is divided into tiles and encoded. In the figure, an image is divided into tiles that are rectangular areas, and each tile is independently encoded and stored in the external storage device ST. The transmission request from the image input unit 511 is input to the parser PS in the image server SV. The parser PS reads the code string of the tile corresponding to the region from the input position information of the region of interest from the external storage device ST and outputs the code string.

Here, the code sequence of each tile is composed of a plurality of layers as described above, and the code sequence of the output tile may be only the encoded data of the necessary layer, or may be the encoded data of all the layers. There may be. The output code strings are combined into one, and a marker code representing the start of the tile and the tile number is inserted at the head of the encoded data of each tile.

In FIG. 31A, for example, the fifth tile out of 16 (hereinafter referred to as tile 5) is a region of interest.
If no code string has been transmitted before that time, the parser constructs and outputs a code string by combining the encoded data of each tile from tile 1 to tile 16, but outputs the tile 5
For all tiles, the number of layers to be transmitted for the other tiles is limited to a necessary minimum.

In this manner, if the encoding is performed reversibly, the decoded image is reversible with respect to the portion of the tile 5 which is the region of interest, and the other portions are sufficient for observation. The image is reproduced with the image quality of.

On the other hand, it is also possible to adopt a different code string construction method, and this will be described below with reference to FIG.

In JPEG2000, coefficients or quantization indices of subbands generated by DWT are bit-plane coded in units of code blocks. In this example, the encoded data of each code block is stored in the external storage device ST in a form that is not formed into a code string.

Referring to FIG. 29, the parser PS generates a code string in which coded data of a code block relating to a region of interest is arranged in an upper layer as shown in FIG. 29 in response to a transmission request from the image input means 511. I do. In this way, the coded data in the intermediate state before the code string configuration is stored in the external storage device ST.
, And dynamically generate an optimal code string in response to an external request, so that even if the above-mentioned tile division is not performed, the region of interest can be displayed in high quality in response to the request on the image display side. Can be generated and output.

Next, the diagnostic information interpreting means 505 (FIG. 14) in the embodiment will be described with reference to FIG.

It is assumed that the provided diagnostic information 521 is “calcification is found at the upper right of the right lung”. The diagnosis information (diagnosis report) may be a case where characters are written in a voice signal or a schema (FIG.) In addition to the text as described above.
In the case of the figure, only the character portion is converted into text information by the character recognition means 523. The following description will be made on the assumption that the text information converted into text is “calcification is found in the upper right of the right lung”.

The keyword determination means 524 extracts a keyword from the input diagnosis information using the keyword dictionary 525. This keyword is registered for each imaging region, and there is a keyword that does not depend on the region. In the case of frontal chest imaging, the right lung, the left lung, the right lung, the left lung, the mediastinum, the mediastinum, and the like are anatomical keywords.
There is the center. The result of extracting the registered keywords is “right lung”, “upper right”, and “calcification”.
In the present embodiment, the pathological information “calcification” is not used, but a layer is formed at the time of encoding depending on the difference between “tumor” and “calcification” and the size of each. It is also conceivable to control the SN ratio of the image. Generally, “tumor” tends to be found at a lower resolution than “calcification”, but “calcification” can be expressed at a lower density resolution in terms of density resolution. As described above, it is conceivable to adjust the coding layer depending on the type of the pathological disease. Similarly, at the time of decoding, it is also possible to control the readout SN ratio using the pathological information.

The extracted keywords are sent to the meaning determining means 52.
7 is sent. The meaning determining means 527 is provided for the meaning dictionary 52.
Using the contents of No. 8, the segment number and the area division information corresponding to the imaging region (eg, the front of the chest, the side of the cervical vertebra) are determined.

[0114] The segment number refers to each image (image area)
Is the number corresponding to the anatomical segment assigned to. For example, as shown in FIG. 22, in the chest front image, three anatomical segments are set for the right lung, the left lung, and the mediastinum. The number of anatomical segments corresponding to each part is dependent on imaging, and in head imaging, there is only one head segment.

The semantic dictionary 528 is composed of a segment number table for the radiographed part shown in FIG. 22 and a definition word table of the area division shown in FIG. In the semantic determination means, the keyword matching the keyword from the segment number table for the imaging region,
Alternatively, an approximate segment number is selected, and at the same time, a matching region division definition word (region definition word) is selected from the region division table. In the present embodiment, segment number 1 and
The region definition word “upper right” is selected and provided to the region of interest determination means 503.

The region of interest determination means will be described with reference to FIG. The image of interest 531 and the interpretation result 5
26 is input. Although the input image is a two-dimensional image, it is generally not necessary to preserve the resolution and image quality of the original image, and a pixel size of about 1-2 mm is sufficient. The input image is subjected to segmentation corresponding to the imaging region by the segmentation means 532.

The number of segmentations is determined depending on the region to be imaged as shown in FIG. For example, a chest front image is decomposed into three resolution parts. FIG. 21 shows an example of the segmentation result image. The anatomical classification number is written inside, and this value is used for labeling. Segmentation means 532
The image segmented by the
3 is converted to the final region of interest. An example of the conversion process is as shown in FIG.

Next, the segmentation means 532 will be described. The structure of the segmentation means differs depending on the learning phase and the use phase. Since the segmentation means of the present embodiment is configured by a neural network, a phase of inputting learning data and forming a coefficient inside the neural network is set as a learning phase, and a phase of performing segmentation on input image data is defined as a phase. Called the use phase. In the following phases, segmentation is performed in pixel units. However, the segmentation method is not limited to pixel unit, and segmentation can be performed by a method of tracking an outline from an image. For contour tracking type segmentation, see Ts
ujii, MT Freedman, and SK Mun, "Lung contour
detection in chest radiographs using 1-D convolut
ion neural networks, "Electronic Imaging, 8 (1), p
See pages 46-53, January 1999.

The learning phase will be described with reference to FIG. The feature amount is calculated by the feature extracting unit 542 for each pixel of the input image. The calculated feature amount includes a value calculated based on a pixel value, a value calculated based on a texture, a value calculated based on a relative address from an anatomical structure, and the like. See the paper (O. Tsujii, MT
Freedman, and SK Mun, "Automated Segmentation
of Anatomic Regions in Chest Radiographs using Ada
ptive-Sized Hybrid Neural Network, "Med. Phys., 25
(6), pp. 998-1007, June 1998.).

The features used are not limited to the features described in the above examples and papers.
MF McNitt-Gray, HK Huang and JW Sayre, "F
eature Selection in the Pattern Classification Pro
blem of Digital Chest Radiograph Segmentation, "IE
Probabilistic features described in EE Trans. Med. Imag., 14, 537-547 (1995).

Next, the neural network means 543 will be described. Various neural networks have been developed. An example is the Feed Forward type backpropagation neural network developed by Rammelhart (references: DERumelhart and JL McC)
elland, Parallel Distributed Processing: Explorati
ons in the Microstructure of Cognition, Vol. 1: Fo
undation. Cambridge: The MIT Press, 1986.), Radial
Basis Function Neural Network (Easy RBF
-NN) (Reference: C. Bishop, "Improving the General
ization Properties of Radial Basis Function Neural
Networks, "Neural Comp., Vol. 3, pp. 579-588, 199
1.)

In the present embodiment, RBF-NN is used. Structurally adopting Feed Forward type, input layer,
It has a three-layer structure of one intermediate layer and an output layer. In the input layer, input nodes corresponding to the number of feature amounts extracted by the feature amount extraction unit 542 are arranged. The RBF neuron arranged in the hidden layer has an output characteristic that has a Gaussian distribution as a nonlinear element. The number of RBF neurons depends on the number of learning cases and the complexity of the problem, but about 5000 is appropriate for setting the calculation time appropriately. The number of outputs of the neural network corresponds to the number of segments for each part. For example, in the case of an image of the front of the chest, there are three anatomical segments as shown in FIG.
NODE is prepared.

In the present embodiment, the neural network is prepared for each imaging part. However, since the neural network is constituted by software, the coefficients for each part are actually stored, and when used, the coefficients for each part are stored. Set the coefficient.

In the learning phase, an answer example as a segmentation result is presented to the neural network 543 together with the feature amount for each pixel. The teacher output image for the input image is provided from the manual segmentation means 541. Input chest image (Fig. 1
FIG. 19 shows a teacher image corresponding to 8). A teacher image is created so as to have an output value of 1-3 in the teacher image so as to correspond to the segment classification corresponding to the chest front image. Teacher image can be done with graphic software that can specify the area in the image,
It is also possible to use hop (made by US Adobe). In the case of RBF-NN, neural network learning can be obtained analytically by using the least squares method that minimizes output errors, and internal coefficients can be calculated in a short time.

The use phase will be described with reference to FIG. The extraction of the feature amount for each pixel in the image is performed in the same manner as in the learning phase. As the internal coefficients for the neural network, the coefficients for the target imaging region are loaded before use. The feature amount for each pixel is presented to the neural network, and the network outputs an output value to an output NODE. In case of chest front image, output NODE
Are three, and the input pixels are classified into segments corresponding to the NODEs having the output closest to YES among the output NODEs. FIG. 21 shows an example of a segment image in which all pixels in the image have been classified in this manner.

Generally, the segment image output by the neural network 551 contains noise. For example, a lung field region having a small area may be generated separately from the lung field region. Such noise-like small regions are removed in the post-processing process. Regarding this technique, see the aforementioned literature: Tsujii, MT Freedman, and SK Mun, "Lungc
ontour detection in chest radiographs using 1-D co
nvolution neural networks, "Electronic Imaging, 8
(1), pp. 46-53, January 1999.

Next, the area dividing means 533 will be described with reference to FIG.

The segment image 561 output from the segmentation output means 551 is input to a labeling means 562. The labeling means 562 is generally a process of assigning a unique numerical value to a distinctive individual region. In this case, a labeling unit 562 assigns a unique numerical value to an image to which a numerical value has already been assigned as shown in FIG. A process for extracting only a region having a target value is performed. More specifically, for an input as shown in FIG. 24A, a process of extracting only a target segment number (in this case, 1 for the right lung) is performed. As a processing algorithm, a Look-Up-Table in which a pixel value is set to a value other than 1 to zero may be applied. An example of a labeling result image is as shown in FIG.

Next, an ellipse approximation process 563 is applied to the labeled image. The elliptic approximation can be obtained by performing principal component analysis while considering the pixels constituting the right lung as a data distribution, but can also be obtained by the following calculation formula. The reason for using the elliptical approximation is that anatomical parts are generally taken at an angle, and the spine,
This is because, when the bones and organs are approximated by ellipses, their locations are generally easy to define. FIG. 24 shows an example of an ellipse approximated to the right lung.
Shown in (C). Long axis of object a = (4 / π) ^1/4 (Imax ³ / Imin) ^1/8 Short axis of object b = (4 / π) ^1/4 (Imin ³ / Imax) ^1/8 The center of gravity of the object's X-axis is Xc, the center of gravity of the Y-axis is Yc, the central moment of the (p, q) order of the object is Mpq, and the inclination of the object θ = (1/2) tan ^-1 (2M ₁₁ / (M ₂₀ -M ₀₂ )), the minimum moment of the object Imin = ΣΣ ((X-Xc) cosθ- (Y-Yc)
sin [theta) maximum moment Imax = ΣΣ ² object ((X-Xc) sinθ + (Y-Yc)
cos θ) ² .

The center of gravity calculating means 564 is simultaneously applied to the segmented anatomical site. The center of gravity is calculated by calculating average coordinates for each of the X and Y coordinates of all the pixels constituting the right lung. The division determining means 565 can separate the right lung into four regions by passing the major axis and the minor axis determined by the ellipse approximating means 563 so as to pass through the center of gravity determined in this way. is there. In the example of the chest front image, the upper right is extracted. The result is shown in FIG.
It should be noted that the medical image is generally displayed so that the doctor faces the patient, so that the left and right are reversed. In the region division definition word shown in FIG. 25, there is a region defined as the center. In this case, an ellipse whose center is the center of the center of gravity and whose major and minor axes are half each is determined as the central region. The area of interest determined in this way is
This is the output of the region of interest determination means 503.

As described above, according to the seventh embodiment, a region of interest is automatically determined using diagnosis information for a diagnosed image without specifying a region of interest by a doctor. Thus, the region of interest can be efficiently encoded.

Further, when an image in which the region of interest is not specified is displayed, the region of interest is automatically formed using the diagnostic information for the image, and the extracted region of interest is given high quality or high priority. The image can be displayed at the resolution, and rapid image diagnosis can be performed.

In addition, when an image in which a region of interest is not displayed is displayed, the region of interest is automatically generated by using diagnostic information for the image, so that the region of interest can be specified on the image. That is, a doctor who diagnoses an image can easily perceive the region of interest.

<Other Embodiments> Further, the present invention is not limited to only the apparatus and method for realizing the above-described embodiment, and the computer (CPU or MPU) in the above-mentioned system or apparatus is provided with the above-mentioned computer. The scope of the present invention also includes a case where a program code of software for realizing the embodiment is provided, and the computer of the system or the apparatus operates the various devices according to the program code to realize the embodiment. included.

In this case, the software program code itself implements the functions of the above-described embodiment, and the program code itself and means for supplying the program code to the computer, specifically, A storage medium storing the above program code is included in the scope of the present invention.

As storage media for storing such program codes, for example, floppy (registered trademark) disks, hard disks, optical disks, magneto-optical disks, CDs
-A ROM, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

In addition to the case where the computer controls various devices in accordance with only the supplied program code to realize the functions of the above-described embodiment, the program code operates on the computer. Such a program code is included in the scope of the present invention even when the above-described embodiment is realized in cooperation with an OS (Operating System) or other application software.

Further, after the supplied program code is stored in a memory provided in a function expansion board of a computer or a function expansion unit connected to the computer, the function expansion board or function is stored based on the instruction of the program code. The case where the CPU or the like provided in the storage unit performs part or all of the actual processing, and the above-described embodiment is realized by the processing is also included in the scope of the present invention.

As described above, according to the present embodiment, a support diagnosis process is performed on a series of images stored and stored in advance, and a region of interest is included in a code sequence in a process of compression-encoding prior to transmission. did. This makes it possible to realize an efficient system in which the time for transmission and interpretation is reduced.

[0140]

As described above, according to the present invention, it is possible to reproduce a region of interest on the side of reproducing an image while distinguishing the region of interest from other regions.

[Brief description of the drawings]

FIG. 1 is a block diagram of a device on the image encoding side according to an embodiment.

FIG. 2 is a block diagram of an image decoding and reproducing apparatus according to the embodiment.

FIG. 3 is a block diagram of an encoding unit in FIG. 1;

FIG. 4 is a diagram for explaining a bit shift process of a region of interest according to the first embodiment.

FIG. 5 is a block diagram of a decoding unit in FIG. 2;

FIG. 6 is a diagram illustrating a data format for each layer according to the second embodiment.

FIG. 7 is a diagram illustrating an example of a relationship between a layer structure and a code block according to the second embodiment.

FIG. 8 is a diagram illustrating a region of interest and component information indicating the region of interest in the third embodiment.

FIG. 9 is a diagram showing a format of code data in a fourth embodiment.

FIG. 10 is a diagram showing a data format depending on the presence or absence of a region of interest in the fifth embodiment.

FIG. 11 is a diagram illustrating a structure of a file in an image sequence according to a sixth embodiment.

FIG. 12 is a diagram illustrating an example of application of the embodiment to a system.

FIG. 13 is a functional block configuration diagram of an encoding device according to a seventh embodiment.

FIG. 14 is a functional block configuration diagram of a decoding device according to a seventh embodiment.

FIG. 15 is a diagram illustrating operation contents of diagnostic information interpreting means in the seventh embodiment.

FIG. 16 is a block diagram of a region-of-interest determination unit according to a seventh embodiment.

FIG. 17 is a diagram illustrating a learning phase of a segmentation unit according to the seventh embodiment.

FIG. 18 is a diagram illustrating an example of a chest front image.

FIG. 19 is a diagram showing learning data of manual segmentation means.

FIG. 20 is a diagram illustrating a use phase of a segmentation unit according to the seventh embodiment.

FIG. 21 is a diagram illustrating an example of a segmentation image output according to the seventh embodiment.

FIG. 22 is a diagram showing a correspondence between an imaging part and a segment number.

FIG. 23 is a diagram showing an operation procedure of the area dividing means in the seventh embodiment.

FIG. 24 is a diagram showing an example of the operation of the area dividing means.

FIG. 25 is a diagram showing definition words for region division.

FIG. 26 is a diagram illustrating an arrangement of layers in the entire code string according to the seventh embodiment.

FIG. 27 is a diagram illustrating a basic configuration of an encoder of JPEG2000.

FIG. 28 is a diagram illustrating a relationship between a region of interest and a quantization index outside the region of interest.

FIG. 29 is a diagram showing that a region of interest is preferentially encoded in a layered encoded data configuration.

FIG. 30 shows an image input unit 511 according to the seventh embodiment.
FIG. 11 is a diagram showing a configuration in a case where a code string is dynamically generated between the image server SV and the image server SV.

FIG. 31 is a diagram illustrating code string transmission when an image is divided into tiles and encoded.

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) H04N 7/30 A61B 5/05 380 390 H04N 7/133 Z F term (Reference) 4C093 AA26 CA29 CA30 DA03 FA35 FA44 FA49 FF17 FF28 FG18 FG19 FH01 FH03 FH06 FH07 4C096 AB27 AB29 AD14 AD16 DC20 DC28 DD18 DD19 DE01 DE03 DE06 DE07 5C059 MA24 MA35 MA35 MA41 PP01 PP02 PP28 RB02 RB09 RC11 RC32 RC33 RC34 RC40 SS20 TA69 TB18 TC24 ACO4 TC34 TC78 DA01 DA02 DB19 EA00

Claims (39)

[Claims] 1. A region of interest in a given medical image is set, or an image having information on the region of interest is encoded using a region of interest setting means for detecting the region of interest in the medical image. An image coding apparatus, comprising: a conversion unit that converts an image into data of a sub-band component for each frequency component; and component data obtained by the conversion unit, component data in the region of interest, and data outside the region of interest. Encoding means for separating and encoding the component data; and output means for adding and outputting at least information for identifying a region of interest and outside the region of interest to the encoded data encoded by the encoding means. An image encoding device comprising: 2. The image encoding apparatus according to claim 1, wherein said transforming means is a wavelet transform. 3. The encoding means according to claim 1, wherein: in the data of the subband components obtained by the wavelet transform, the component data in the region of interest is shifted by a bit number that can be distinguished from the outside of the region of interest. Means for generating a bit plane composed of the same bit positions in the data shifted by the bit shifting means, and encoding each bit plane.
3. The image encoding apparatus according to claim 2, wherein the output unit adds the number of bits to be shifted as the information and outputs the information. 4. The apparatus according to claim 3, wherein said output means does not output coded data in a bit plane lower than a bit plane including at least the region of interest by a predetermined lower level after being shifted by said bit shift means. An image encoding device according to claim 1. 5. The encoding means has means for encoding component data in the region of interest and component data outside the region of interest into different layers in subband component data obtained by wavelet transform. 3. The apparatus according to claim 2, wherein the output unit stores and outputs information indicating zero length in a packet for storing code data of the region of interest when there is no region of interest. Image coding device. 6. The encoding unit includes: a generation unit configured to generate a binary image for distinguishing the region of interest from the outside of the region of interest; and a component different from the image to be encoded. 3. The image encoding apparatus according to claim 2, further comprising: encoding means, and wherein the output means adds and outputs the number of components as the information. 7. The image coding apparatus according to claim 2, wherein said coding means has means for adding information on the region of interest to a predetermined header. 8. The image processing apparatus according to claim 1, wherein the encoding unit has a unit that reduces the number of layers to be encoded by one when there is no region of interest and performs encoding differently from the number when there is a region of interest. 3. The image encoding device according to claim 2, wherein: 9. The image processing apparatus according to claim 1, wherein the image is a medical X-ray image, an X-ray slice by CT, or a slice image of MRI. Image coding device. 10. The apparatus according to claim 1, wherein the region-of-interest setting means performs a computer-aided diagnosis process on the input medical image and sets a region of interest based on a result of the diagnosis assisted process. Item 10. The image encoding device according to any one of Items 9 to 9. 11. The region-of-interest setting means includes diagnostic information input means for inputting diagnostic information corresponding to the input medical image, and sets a region of interest for the medical image based on the input diagnostic information. The image encoding device according to any one of claims 1 to 10, wherein: 12. The image encoding apparatus according to claim 10, wherein the diagnostic information is text information. 13. The image coding apparatus according to claim 10, wherein said diagnosis information is audio information information. 14. The apparatus according to claim 9, wherein the region of interest setting means does not determine the region of interest when the diagnostic information is non-existent. An image encoding device according to claim 1. 15. An image decoding device for decoding image data encoded by the image encoding device according to claim 3, wherein: decoding means for decoding the input encoded data for each bit plane; In accordance with the number of shift bits, inverse bit shift means for shifting the data of the sub-band component constituted by the bit plane group of the region of interest in the reverse direction, and each of the data obtained by the decoding means and the inverse bit shift means. From a bit plane, generating image data of a sub-band component, performing an inverse conversion to the conversion unit, and providing an image reproduction unit that reproduces the region of interest in a different form from the region other than the region of interest. Image decoding device. 16. An image decoding device for decoding image data encoded by the image encoding device according to claim 5, wherein information indicating zero length is included in a packet storing code data of a region of interest. Determining means for determining whether or not there is, and when there is data other than zero length, the determining means determines that there is a region of interest and decodes an image including the region of interest, and makes the region of interest different from that other than the region of interest. An image decoding apparatus, comprising: an image reproducing unit configured to reproduce the image in a form. 17. An image decoding device for decoding image data encoded by the image encoding device according to claim 6, wherein said detecting means detects the number of components and said detecting means detects the number of components. If there are two components, the first
Is decoded as a binary image for identifying the region of interest and the region of interest, and when decoding and reproducing the data of other components, the region of interest is associated with the region of interest according to the binary image. An image decoding device, comprising: an image reproducing unit that reproduces the image in a form different from the above. 18. An image decoding device for decoding image data encoded by the image encoding device according to claim 7, wherein: decoding means for decoding the encoded image data; An image reproducing unit that reproduces a region of interest in the image decoded by the decoding unit in a different form from the region of interest based on information about the region of interest added to a predetermined header. An image decoding apparatus characterized by the above-mentioned. 19. An image coding method for setting a region of interest in a given medical image, or using a region of interest setting means for detecting a region of interest to code an image having information about the region of interest. A conversion step of converting an image into sub-band component data for each frequency component; and separating component data in the region of interest and component data outside the region of interest in the component data obtained in the conversion step. And an output step of adding at least information for identifying a region of interest and an area outside the region of interest to the encoded data encoded in the encoding step and outputting the encoded data. Image encoding method. 20. The image encoding method according to claim 19, wherein the conversion process in the conversion step is a wavelet transform. 21. The encoding step, wherein the subband component data obtained by the wavelet transform includes:
For the component data in the region of interest, a bit shift step of shifting the data by the number of bits that can be distinguished from the outside of the region of interest, and generating a bit plane composed of the same bit positions in the data resulting from the shift in the bit shift step, A step of encoding for each bit plane, wherein the output step adds the number of bits to be shifted as the information and outputs the information.
Item 10. The image encoding method according to Item 9. 22. The method according to claim 21, wherein the output step does not output coded data in a bit plane lower than a bit plane including at least the region of interest after being shifted by the bit shift means. An image encoding device according to claim 1. 23. The encoding step, wherein the subband component data obtained by the wavelet transform includes:
Encoding the component data in the region of interest and the component data outside the region of interest in different layers, wherein the output step includes, when there is no region of interest, the packet for storing the code data of the region of interest. 21. An apparatus according to claim 20, wherein information indicating zero length is stored and output.
An image encoding method according to any one of the preceding claims. 24. The encoding step, comprising: a generating step of generating a binary image for distinguishing the region of interest from the outside of the region of interest; and converting the generated binary image into a component different from the image to be encoded. 21. The image encoding method according to claim 20, further comprising: encoding as component information, and wherein the output step adds and outputs the number of components as the information. 25. The image encoding method according to claim 20, wherein said encoding step includes means for adding information on said region of interest to a predetermined header. 26. The method according to claim 26, wherein the encoding step includes a step of reducing the number of layers to be encoded by one when there is no region of interest and performing encoding differently from the number of layers when there is a region of interest. The image encoding method according to claim 20, characterized in that: 27. The image processing apparatus according to claim 19, wherein the image is a medical X-ray image, an X-ray slice by CT, or a slice image of MRI. Image coding method. 28. The region-of-interest setting means includes diagnostic information input means for inputting diagnostic information corresponding to the input medical image, and sets a region of interest for the medical image based on the input diagnostic information. The image encoding method according to any one of claims 19 to 27, wherein: 29. The image encoding method according to claim 28, wherein the diagnostic information is text information. 30. The image encoding apparatus according to claim 28, wherein said diagnostic information is audio information information. 31. The apparatus according to claim 28, wherein the region-of-interest setting means does not determine the region of interest when the diagnostic information indicates no finding. An image encoding device according to claim 1. 32. A computer which has a program code corresponding to each of the steps according to claim 19, and functions as a device which encodes image data by being read and executed by a computer. Computer program for. 33. A storage medium storing the computer program according to claim 32. 34. An image decoding method for decoding image data encoded by the image encoding device according to claim 3, wherein: a decoding step of decoding input encoded data for each bit plane; In accordance with the number of shift bits, an inverse bit shift step of shifting the data of the sub-band component constituted by the bit plane group of the region of interest in the reverse direction, and each of the data obtained in the decoding step and the inverse bit shift step An image reproduction step of generating data of a subband component from a bit plane, performing a reverse conversion to the conversion unit, and reproducing the region of interest in a different form from the region other than the region of interest. Image decoding method. 35. An image decoding method for decoding image data encoded by the image encoding device according to claim 5, wherein the information indicating zero length is included in a packet storing the code data of the region of interest. A determination step of determining whether or not there is, and if there is data other than zero length in the determination step, it is determined that there is a region of interest, an image including the region of interest is decoded, and the region of interest is different from the region other than the region of interest. And an image reproduction step of reproducing the image in a form. 36. An image decoding method for decoding image data encoded by the image encoding device according to claim 6, wherein the detecting step detects the number of components, and detects the number of components. If there are two components, the first
Is decoded as a binary image for identifying the region of interest and the region of interest, and when decoding and reproducing the data of other components, the region of interest is associated with the region of interest according to the binary image. An image reproducing step of reproducing the image in a different form. 37. An image decoding method for decoding image data encoded by the image encoding device according to claim 7, wherein: a decoding step of decoding the encoded image data;
An image reproduction step of reproducing a region of interest in the image decoded in the decoding step in a different form from the region of interest based on the information on the region of interest added to the predetermined header. A featured image decoding method. 38. It has a program code corresponding to each step according to any one of claims 34 to 37, and functions as a device that decodes image data by being read and executed by a computer. Computer programs. 39. A storage medium storing the computer program according to claim 38.