Dynamic image analysis apparatus, method, and program

Abstract

To solve the problem that all information cannot be extracted, even though a periodic spectral analysis, such as FFT, is performed on a dynamic image.SOLUTION: A dynamic image analysis apparatus according to one embodiment of the present disclosure includes an acquisition unit which acquires a dynamic image of a dynamically imaged photographic subject, and an analysis unit which unsteadily spectrally analyzes the dynamic image acquired by the acquisition unit.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a dynamic image analysis device, method, and program. [Background technology]

[0002] Dynamic imaging is performed by a radiation generator repeatedly emitting radiation pulses at a frequency (pulse period) of multiple times per unit time (e.g., 15 times per second) for a predetermined time (duration) while an emission command is being issued, and a radiation detection device reads out the amount of charge generated in accordance with the radiation dose received through the subject as a signal value (intensity). Dynamic imaging captures dynamic images consisting of multiple (series of) still images captured at different times, each captured at a pulse period. The frequency at which still images are captured is called the frame rate, and is equal to the radiation pulse period. Doctors diagnose diseases based on the captured dynamic images. By performing dynamic imaging of organs such as the lungs and heart, doctors can diagnose lung and heart diseases based on the movement of the lungs and heart. Furthermore, by performing dynamic imaging of bones, doctors can make diagnoses based on the movement of joints.

[0003] Analysis is being performed in which a fast Fourier transform (FFT) is performed on a dynamic image to extract a specific frequency spectrum, such as a periodic signal synchronized with the heartbeat. A specific frequency spectrum can be extracted by FFT. Extracting frequencies synchronized with the heartbeat as the frequency spectrum extracted by FFT on a dynamic image makes it possible to extract the contraction and expansion of blood vessels accompanying the heartbeat in the dynamic image. Detecting areas where the contraction and expansion of blood vessels accompanying the heartbeat decrease can provide information useful for diagnosing pulmonary embolism. Areas where the signal decreases can be detected, for example, by the difference in the amount of signal change from a reference frame (see, for example, Patent Documents 1 and 2). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2023-121104 [Patent Document 2] Japanese Patent Publication No. 2022-095871 Summary of the Invention [Problem to be solved by the invention]

[0005] Dynamic images show changes in blood vessel size and blood volume due to cardiac induced vessel dilation caused by cardiac contraction and dilation, which are detected as changes in X-ray intensity by an X-ray detection device. Because the pulmonary artery undergoes large changes in contraction and dilation due to cardiac pulsation, changes in blood vessel contraction and dilation can be detected by performing an FFT on the changes in X-ray intensity obtained from dynamic images and extracting frequencies synchronized with cardiac pulsation.

[0006] On the other hand, although the lung field (anatomical lung parenchyma) on the image, which is dominated by capillaries (peripheral blood vessels), undergoes contraction and expansion of blood vessels in synchronization with the heartbeat, the change in X-ray intensity is thought to be attenuated relative to the change in intensity of the pulmonary artery.

[0007] Furthermore, when the tissue properties of the lung parenchyma become non-uniform due to pulmonary disease or other reasons, at least one of resonance, attenuation, and reflection occurs in the vibration propagation due to the contraction and expansion of blood vessels. The phenomena of resonance, attenuation, and reflection in vibration propagation are considered to be observed as non-stationary signals.

[0008] In this way, we realized that because biological signals (constriction and expansion of blood vessels due to heartbeat) are thought to contain non-stationary signals, it is not possible to extract information related to non-stationary signals even if stationary spectral analysis such as FFT is performed on dynamic images. Therefore, with conventional methods, for example, it is not possible to extract blood flow signals related to capillaries (peripheral blood vessels), which are observed as non-stationary signals, and it is thought that useful information about peripheral pulmonary embolism cannot be obtained. [Means for solving the problem]

[0009] A dynamic image analysis device in one embodiment of the present disclosure includes an acquisition unit that acquires a dynamic image of a subject that has been photographed in dynamic photography, and an analysis unit that performs non-stationary spectral analysis of the dynamic image acquired by the acquisition unit.

[0010] A dynamic image analysis method according to an embodiment of the present disclosure acquires a dynamic image of a subject who has been photographed in dynamic photography, and performs non-stationary spectral analysis on the acquired dynamic image.

[0011] A dynamic image analysis program according to an embodiment of the present disclosure causes a computer to acquire a dynamic image of a subject who has been photographed in dynamic photography, and to perform non-stationary spectral analysis on the acquired dynamic image.

[0012] These comprehensive or specific aspects may be realized as a system, an apparatus, a method, an integrated circuit, a computer program, or a recording medium, or may be realized as any combination of a system, an apparatus, a method, an integrated circuit, a computer program, and a recording medium. [Effects of the Invention]

[0013] According to the present disclosure, by performing non-stationary spectral analysis on dynamic images, useful information for pulmonary diseases, cardiac diseases, and the like can be obtained. [Brief explanation of the drawings]

[0014] [Figure 1] Diagram showing the configuration of a dynamic image analyzer [Figure 2] Processing circuit functional block diagram [Figure 3] ROI location example [Figure 4] Examples of biosignals for each ROI [Figure 5] Scalogram Example [Figure 6] Scalogram Example [Figure 7] Frequency Spectrum Example [Figure 8] Frequency Spectrum Example [Figure 9] Flowchart of a program executed by the processing circuit [Figure 10] Image of block division [Figure 11] Flowchart of a program for diagnosing emphysema [Figure 12] An example of selecting high frequency bands from the calculated frequency spectrum for each block. [Figure 13] Example of calculated high-frequency band strength for each block [Figure 14] An example of selecting low frequency bands from the calculated frequency spectrum for each block. [Figure 15] Example of calculated low-frequency band strength for each block DETAILED DESCRIPTION OF THE INVENTION

[0015] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings as appropriate.

[0016] <Dynamic image analysis device> The configuration of a dynamic image analysis device 100 according to an embodiment of the present disclosure will be described.

[0017] [composition] FIG. 1 is a diagram showing the configuration of a dynamic image analyzer 100.

[0018] The dynamic image analysis device 100 includes a processing circuit 110, an input / output unit 120, a communication unit 130, and a memory 140. The input / output unit 120 includes an input unit 121 and an output unit 122. The input unit 121 and the output unit 122 may be integrated. If input and output are performed via the communication unit 130, the input / output unit 120 may be omitted.

[0019] The processing circuit 110 is configured with a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), etc., and may include a neural network. The processing circuit 110 performs non-stationary spectral analysis on the input medical image based on the input medical image. Details of the processing circuit 110 will be described later.

[0020] The input unit 121 includes at least one of a touch panel, a keyboard, a mouse, a microphone, etc., and receives input based on the operation of a user (doctor, radiologist, etc.).

[0021] The output unit 122 includes at least one of a display, a speaker, a printer, etc., and outputs the results of the non-stationary spectrum analysis performed by the processing circuit 110 to the outside.

[0022] The communication unit 130 communicates with external devices wirelessly or via wired buses, LANs (Local Area Networks), the Internet, VPNs (Virtual Private Networks), public lines, etc. The communication unit 130 communicates with hospital information systems (HIS), radiology information systems (RIS), picture archiving and communication systems (PACS), dynamic imaging devices, etc.

[0023] The memory 140 is composed of a ROM (Read Only Memory), a RAM (Random Access Memory), an EPROM (Erasable Programmable ROM), an EEPROM (Electrically EPROM), an HDD (Hard Disk Drive), etc., and stores dynamic images, various programs, etc.

[0024] FIG. 2 is a functional block diagram of the processing circuit 110.

[0025] The processing circuit 110 includes an acquisition unit 111 , an analysis unit 112 , and a classification unit 113 .

[0026] The acquisition unit 111 acquires a dynamic image. The dynamic image may be acquired from an external system such as a RIS via the communication unit 130 based on an external input via the input unit 121 or the communication unit 130, or may be acquired from the memory 140. The dynamic image is, for example, a dynamic image of the lungs.

[0027] The analysis unit 112 performs non-stationary spectrum analysis based on the dynamic image acquired by the acquisition unit 111. Details will be described later.

[0028] The classification unit 113 classifies the dynamic image based on the analysis result of the analysis unit 112. Details will be described later.

[0029] [Dynamic image] Dynamic images are composed of multiple (series of) still images taken at different times, each taking a pulse period. Dynamic imaging is performed by emitting radiation pulses repeatedly over a predetermined time (duration) at a frequency (pulse period) of multiple times per unit time (for example, 15 times per second) while an emission command is issued, and the radiation detection device reads out the amount of charge generated according to the radiation dose received through the subject as a signal value (intensity).

[0030] The intensity I of the X-rays after passing through a material is expressed as follows, where I0 is the intensity of the X-ray incident on the material, μ is the linear attenuation coefficient, ρ is the material density, and x is the thickness of the material: I=I0exp(-μx) (1) Here, the linear attenuation coefficient μ[1 / cm] is the mass attenuation coefficient μ m [cm 2 / g], material density ρ[g / cm 3 ] as μ=μ m ρ (2) Therefore, I=I0exp(-μ m ρx) (3) is.

[0031] By having the subject hold their breath during dynamic imaging, dynamic images can be obtained with the subject and lung thickness (volume) kept constant. m is constant, and the X-ray radiation intensity I0 of each pulse used for dynamic imaging is also constant. Therefore, according to equation (3), if the lung thickness x is constant, the intensity I of the X-rays captured is related to the lung density ρ. It is presumed that lung density is affected by changes in biosignals, for example, the state of the blood vessels in the lungs expanding and contracting with heartbeat. In other words, it is presumed that the temporal change in X-ray intensity in dynamic images is due to the influence of changes in biosignals, for example, the contraction and expansion of blood vessels with heartbeat.

[0032] That is, the dynamic images acquired by the acquisition unit 111 are dynamic images captured while the subject is holding his / her breath.

[0033] [Biological signals] A biosignal is acquired from a dynamic image. The acquired biosignal is the X-ray intensity for each pixel (element constituting the detector) of the imaging device in each frame. The biosignal may be acquired for each region of interest (ROI) of the dynamic image, such as the pulmonary artery, peripheral blood vessels, upper lobe, middle lobe, and lower lobe. The biosignal is the temporal change in X-ray intensity (time-intensity characteristics).

[0034] 3 is a diagram showing examples of ROI positions. For example, ID1 is the right pulmonary artery center, ID2 is the right upper lobe of the lung (upper lung field, S1), ID3 is the right upper lobe of the lung (middle lung field, S3), ID4 is the right middle lobe of the lung (lower lung field, S4), ID5 is the left pulmonary artery center, ID6 is the left upper lobe of the lung (upper lung field, S1+2), ID7 is the left upper lobe of the lung (middle lung field, S3), ID8 is the left lower lobe of the lung (lower lung field, S8), and ID9 is the right lower lobe of the lung (lower lung field, S8). The positions and number of ROIs are arbitrary and are not limited to the nine shown in FIG. 3.

[0035] Figure 4 shows an example of a biological signal for each ROI.

[0036] The analysis unit 112 analyzes the acquired biological signals. Specifically, the analysis unit 112 logarithmically transforms the X-ray intensities of the acquired dynamic images, blocks the logarithmically transformed X-ray intensities, and analyzes the blocked dynamic images for each block. The blocking is a process of, for example, averaging the X-ray intensities for each of multiple pixels of the detector. For example, the detector pixel is 0.4 mm square, and the blocked block size is 3 mm square. The analysis unit 112 may block ROIs and perform analysis only on the ROIs.

[0037] The analysis can be performed using stationary or non-stationary spectral analysis. The ROI can be a predetermined region. Each block of a dynamic image can be an ROI, or multiple blocks of a dynamic image can be a single ROI. The ROI can be an organ (e.g., the entire lung, upper lobe, middle lobe, lower lobe, or the entire heart).

[0038] The signal is mathematically A i sinω i The sum (composition) of t, i.e.,

number

[0039] Steady-state spectrum analysis is a method for expressing a signal as the sum of frequencies of a constant amplitude, and is a useful signal processing method for stationary signals or when biological signals are assumed to be stationary signals. FFT is widely used for steady-state spectrum analysis. By extracting specific frequency components of a stationary signal from a dynamic image using FFT, the biological signal of the extracted frequency components can be understood. For example, by extracting frequency components corresponding to cardiac pulsation, the contraction and expansion of blood vessels accompanying cardiac pulsation in a dynamic image can be understood.

[0040] Since Fourier transforms including FFT do not provide time information, non-stationary spectral analysis, which performs analysis including time information, is suitable for analyzing non-stationary signals whose frequency or amplitude changes over time.

[0041] Because the contraction and expansion of blood vessels are induced by the heartbeat, the signal generally changes periodically. However, due to arrhythmia, for example, the heartbeat does not repeat the same fluctuations every time. Therefore, it is best to assume that the signal is non-stationary.

[0042] In addition, the pulmonary blood vessels branch into the pulmonary artery, and blood perfuses the capillaries, undergoes gas exchange with the alveoli, and then circulates to the pulmonary veins. The contraction and dilation of the capillaries are attenuated signals induced by cardiac pulsation. Furthermore, because the blood flow in the pulmonary veins is steady and slower than that of the pulmonary arteries, there is no contraction or dilation of the pulmonary veins. In other words, the contraction and dilation of the pulmonary blood vessels (cardiac-induced vessel dilation) are shear waves originating from the pulmonary arteries on dynamic images. Since the vibration propagation includes attenuated signals in the lung parenchyma, which contains different tissues, such as the pulmonary arteries and capillaries, it should be assumed to be a non-stationary signal.

[0043] In other words, it is best to assume that the heartbeat itself is a non-stationary signal, and since vibration propagation is also a non-stationary signal, it can be said that the biological signal extracted from a dynamic image is a non-stationary signal.

[0044] Dynamic images detect changes in lung density in the direction of irradiation, which includes pulmonary arteries, capillaries, and veins, and is understood to mean that lung tissue is not uniform. When lung tissue is not uniform, resonance, attenuation, and reflection occur in the blood flow in the lungs, and the degree of resonance, attenuation, and reflection of the blood flow changes over time. In other words, it is best to assume that the blood flow in a given ROI is a non-stationary signal, just like the heartbeat itself, and that resonance, attenuation, and reflection also change over time. Therefore, it is appropriate to assume that the biological signals in each ROI are also non-stationary signals.

[0045] Therefore, steady-state spectrum analysis of biological information is a simple analysis that analyzes the steady-state signal portion contained in the non-stationary signal, and in order to analyze the details of the biological information, it is useful for the analysis unit 112 to perform non-stationary spectrum analysis of the biological information.

[0046] Nonstationary spectral analysis is a signal processing technique useful for nonstationary signals, and is an analytical method that uses time-frequency analysis, multiresolution analysis, or a combination of these. Examples of nonstationary spectral analysis include wavelet transform, windowed FFT, and Wigner distribution.

[0047] A steady-state signal becomes a non-stationary signal when it resonates, attenuates, or reflects. When lung tissue is non-uniform due to pulmonary disease or other conditions, resonance, attenuation, and reflection occur in the pulmonary blood flow. Therefore, performing non-stationary spectral analysis on dynamic images can provide useful information about pulmonary diseases.

[0048] Generally, low-frequency signals change slowly over time, while high-frequency signals change rapidly over time. In analysis, there is a trade-off between time resolution and frequency resolution (uncertainty principle). However, wavelet transform is suitable for analyzing the above signals because it can perform a transformation that prioritizes frequency resolution over time resolution in the low-frequency region and time resolution over frequency resolution in the high-frequency region. In the following explanation, we will use an example of wavelet transform as a non-stationary spectrum analysis.

[0049] [Wavelet Transform] The wavelet transform is a method of expressing a mother wavelet function as the sum of signals compressed, expanded (scaled), and shifted in time. The results of the wavelet transform are displayed as time-frequency characteristics, i.e., as a scalogram with the horizontal axis representing time and the vertical axis representing frequency, and the intensity of each frequency component (wavelet coefficient) displayed by color.

[0050] Fig. 5 shows an example of a scalogram (time-frequency characteristics) at ID1 in Fig. 3. Fig. 6 shows an example of a scalogram at ID3 in Fig. 3. Figs. 5 and 6 are displayed with densities corresponding to the intensities over time and frequency, but they may also be output with colors corresponding to the intensities over time and frequency.

[0051] The scalogram can be reduced in dimension to generate (calculate) frequency-intensity characteristics (frequency spectrum). The maximum, average, or sum of wavelet coefficients (intensity) for each frequency component is calculated for the time period from the first time to the second time in the scalogram to generate a frequency spectrum. For example, if a dynamic image has a total of 120 frames, the maximum, average, or sum of intensity may be calculated based on frames 21 to 100. Frequency spectra can also be generated for multiple time periods. The generated frequency spectrum is displayed, for example, with frequency on the horizontal axis and intensity on the vertical axis. At least one of the maximum, average, and sum of intensity is displayed. Multiple intensities, such as the maximum and average, may also be displayed.

[0052] The dimensionality reduction method for the scalogram may be to calculate features from a covariance matrix using principal component analysis (PCA).

[0053] The intensity information may be the intensity of each frequency, or the ratio of the intensity of each frequency to the intensity of all frequencies (energy frequency distribution).

[0054] FIG. 7 shows an example of a frequency spectrum at ID1 in FIG. 3, i.e., a frequency spectrum generated from the scalogram in FIG. 5. FIG. 8 shows an example of a frequency spectrum at ID3 in FIG. 3, i.e., a frequency spectrum generated from the scalogram in FIG. 6. The solid line indicates the average, and the dotted line indicates the maximum value. Although FIGS. 7 and 8 show the intensity of one ROI, the intensities of multiple ROIs may be displayed simultaneously. If regions A to C are set as ROIs, for example, region A may be displayed in red, region B in yellow, and region C in blue, and the maximum value of each region may be displayed in dotted lines, and the average in each region may be displayed in solid lines.

[0055] [Classification] The classification unit 113 can provide information useful for diagnosing diseases by classifying the blocked dynamic image based on the results of non-stationary spectrum analysis. The classification unit 113 can provide information useful for diagnosing lung diseases and cardiac diseases by classifying, for example, based on at least one of a scalogram, which is the result of wavelet transformation, and a frequency spectrum. The classification unit 113 can classify based on the intensity of any frequency band in the frequency spectrum. The frequency band for calculating the intensity is arbitrary and selectable.

[0056] Information useful for diagnosis includes, for example, qualitative diagnoses of lung and heart diseases, such as whether the disease is malignant or benign, the type of lung disease, the level (severity) of lung disease, and the progression of the disease (degree of progression).

[0057] The classification unit 113 can classify scalograms by unsupervised learning. The classification unit 113 can classify which dynamic images are similar to which dynamic images from multiple scalograms. Furthermore, the classification unit 113 can perform classification based on frequency spectra by, for example, principal component analysis (dimensionality reduction).

[0058] By classifying the images, the doctor can determine which past dynamic images the current dynamic image is similar to, and can therefore use the subject's information from the past dynamic images to make a diagnosis of the current subject.

[0059] The classification unit 113 can classify the scalograms by supervised learning. For example, the classification can be performed based on machine learning using training data, such as regression analysis or deep learning. For example, the training data can be dynamic images of a healthy patient, dynamic images of a patient with mild pulmonary embolism, and dynamic images of a patient with severe emphysema. The classification unit 113 may classify the scalogram of the subject's dynamic image based on which of the training data scalograms it resembles.

[0060] By classification, it is possible to provide information about the disease, such as information about the possibility of a disease, information about the level (severity) of a certain disease, etc., for the current dynamic imaging.

[0061] The scalograms of at least two different regions of interest (ROIs) are compared to determine their correlation (coherence). For example, if the scalogram of the pulmonary artery and the scalogram of the lung field are compared and the same frequency components are present at the same time (or at a time delayed by a specified time), the coherence is determined (classified) as high.

[0062] If the coherence is high, it can be inferred that blood in the pulmonary artery is flowing smoothly to the lung field, while if the coherence is low, it can be inferred that blood in the pulmonary artery is not flowing smoothly to the lung field for some reason. It can be inferred that the blood is not flowing smoothly because of the presence of a lung disease. In other words, by comparing scalograms of different regions and classifying them based on their relevance (coherence), it is possible to provide useful information about lung diseases.

[0063] Based on the frequency spectrum generated from the scalogram, non-stationary spectral analysis features for dynamic imaging can be identified. Non-stationary spectral analysis features can also be identified based on the intensity of any frequency region in the frequency spectrum. When the intensity of a specific frequency is high, it can be inferred that resonance or reflection occurs in the blood. Resonance or reflection in the blood may indicate that lung tissue is not uniform. In other words, classification based on the frequency spectrum, for example, when the intensity of a high frequency in the frequency spectrum is high, can provide useful information for lung disease.

[0064] Frequency spectra can also be trained and classified. Similar to training scalograms, frequency spectra can be classified based on unsupervised or supervised learning.

[0065] For example, if the signal intensity of a specific frequency in the frequency spectrum is high, it is possible to provide information indicating the possibility of emphysema or interstitial pneumonia. Also, if the frequency close to the heartbeat frequency is high in a specific region in the frequency spectrum, it is possible to provide information indicating the possibility of pulmonary embolism.

[0066] Dynamic images of the same subject taken at different times can provide information about the progression of the disease.

[0067] The data to be classified is optimized depending on the disease to be diagnosed. Optimization can be considered to be equivalent to preprocessing of machine learning. For example, necessary data is selected depending on the disease. For example, when diagnosing disease A, a scalogram with a different specific ROI is selected, and when diagnosing disease B, a frequency spectrum is selected. The processing circuit 110 may calculate only the characteristic data to be selected, or may calculate all the characteristic data and select from all the characteristic data.

[0068] The classification unit 113 performs classification based on the coherence of the scalogram of the specific ROI based on the optimized data, and performs classification based on the frequency spectrum.

[0069] Since dynamic chest images capture both the lungs and the heart, it is possible to analyze not only the pulmonary blood vessels but also the cardiac blood vessels. For example, non-stationary spectral analysis of the expansion and contraction of the atria and ventricles, and the expansion and contraction of the aorta, can provide useful information about heart disease.

[0070] Furthermore, dynamic imaging of the brain allows for analysis of cerebral blood vessels, which can provide useful information regarding brain diseases.

[0071] In the case of dynamic imaging of the subject's lungs (heart), the diseases to be diagnosed include circulatory diseases such as pulmonary disease, cardiac disease, and brain disease. Pulmonary diseases include the presence and severity (level) of emphysema, subtypes of interstitial pneumonia, pulmonary embolism, peripheral pulmonary embolism, and pleural effusion. Cardiac diseases include heart failure and atrial fibrillation.

[0072] For example, it can provide useful information for distinguishing between heart failure and pneumonia, diagnosing whether heart failure is HFrEF (heart failure with reduced ejection fraction: systolic dysfunction) or HFpEF (heart failure with preserved ejection fraction: diastolic dysfunction), screening for spirometry, distinguishing between atelectasis and pleural effusion, diagnosing the degree of pleural adhesion and thickness, and estimating left atrial pressure.

[0073] FIG. 9 shows a flowchart of the program executed by the processing circuit 110.

[0074] The acquisition unit 111 of the processing circuit 110 acquires a dynamic image captured while the subject holds their breath (step S901). The dynamic image may be acquired from an external device such as a dynamic imaging device via the communication unit 130, or may be acquired from the memory 140. The dynamic image is acquired as the intensity of each pixel constituting the detector of the X-ray imaging device. If the pixel of the detector is, for example, 0.4 mm square, the acquired dynamic image is the intensity of each 0.4 mm square pixel.

[0075] The analysis unit 112 of the processing circuit 110 performs logarithmic transformation on the intensity of the acquired dynamic image to calculate a logarithmic transformation value (step S902). Since the X-ray intensity is a value with lung density ρ as an exponent, as expressed by the above formula (3), by performing logarithmic transformation, a linear value with respect to lung density ρ can be obtained.

[0076] The analysis unit 112 of the processing circuit 110 divides the dynamic image acquired in step S901 into blocks (step S903). The analysis unit 112 converts the dynamic image into blocks of, for example, 3 mm square. For example, by calculating the average of every 8 × 8 pixels from the data acquired for each pixel (0.4 mm square) in step S901, the block size can be converted into a block size of 3.2 mm square. By dividing the dynamic image into blocks, the influence of noise can be reduced. Increasing the block size (reducing the spatial resolution) during blocking increases the S / N (intensity resolution), while decreasing the block size (increasing the spatial resolution) decreases the S / N (intensity resolution). In other words, the size of the block is determined according to at least one of the spatial resolution and intensity resolution to be acquired. Figure 10 is a diagram showing an image of the blocking process.

[0077] The analysis unit 112 of the processing circuit 110 performs non-stationary spectral analysis on each block of the blocked dynamic image to calculate the time-frequency characteristics (scalogram) for each block (step S904). The non-stationary spectral analysis may be, for example, a wavelet transform, a windowed FFT, or a Wigner distribution.

[0078] The analysis unit 112 of the processing circuit 110 calculates the maximum, average, or sum of the wavelet coefficients (intensity) for each frequency component for the time period from the first time to the second time of the calculated scalogram for each block, thereby generating (calculating) a frequency spectrum for each block (step S905). For example, if the dynamic image has a total of 120 frames, the processing circuit 110 may calculate the maximum, average, or sum of the intensities based on the frames from the 21st to the 100th frames. The processing circuit 110 displays the generated frequency spectrum for each block, with the horizontal axis representing frequency and the vertical axis representing intensity. For example, in response to the selection of an ROI displayed in FIG. 3, the processing circuit 110 displays the frequency spectrum of the selected ROI. The processing circuit 110 displays at least one of the maximum, average, and sum from the first time to the second time as the intensity. The processing circuit 110 may display the intensity of each frequency, or the proportion of the intensity of each frequency to the intensity of all frequencies (energy frequency distribution). The intensities of multiple ROIs may be displayed.

[0079] The analysis unit 112 of the processing circuit 110 acquires an arbitrary spectrum (intensity of an arbitrary frequency band) of the frequency spectrum for each calculated block (step S906). The analysis unit 112 displays the acquired intensities for each block as an image of the chest. The frequency band to be acquired can be selected. A doctor can diagnose a lung disease based on the intensity for each block.

[0080] FIG. 11 shows an example flowchart of a program executed by processing circuitry 110 for diagnosing emphysema.

[0081] The processing from step S901 to step S905 is the same as that in FIG. 9, and therefore description thereof will be omitted.

[0082] The analysis unit 112 of the processing circuit 110 obtains the intensity of a high-frequency band (for example, a frequency band higher than heartbeat) for each block from the frequency spectrum for each block calculated in step S905 (step S1106). The intensity of the frequency band is calculated as the average or sum of the selected frequency bands. FIG. 12 is a diagram showing an example of selecting a high-frequency band from the frequency spectrum for each block calculated in step S905 in step S1106. FIG. 13 is a diagram showing an example of the calculated intensity of the high-frequency band for each block. It can be assumed that signals with higher frequency components than the heartbeat are signals generated when signals induced by the heartbeat collide with air and are reflected. Therefore, by acquiring the intensity of the high-frequency band for each block, it is possible to obtain information useful for diagnosing which blocks contain a lot of air, i.e., which parts have emphysema.

[0083] The analysis unit 112 of the processing circuit 110 executes lung field masking (step S1107). Lung field masking is a process for extracting only the lung field area. By removing areas that are not relevant to the diagnosis, it is possible to provide doctors with only information relevant to determining whether or not a lung disease has occurred.

[0084] <Modification> On the other hand, when the lungs become fibrotic, they become stiff, and it is presumed that resonance occurs in the low-frequency band. Therefore, the analysis unit 112 of the processing circuit 110 can obtain information useful for diagnosing which part of the lungs has interstitial pneumonia (pulmonary fibrosis) by acquiring a frequency spectrum in the low-frequency band (for example, a frequency band lower than the heartbeat) for each calculated block. Figure 14 is a diagram showing an example in which, in step S906, a low-frequency band is selected from the frequency spectrum for each block calculated in step S905. Figure 15 is a diagram showing an example in which the intensity of the calculated low-frequency band is shown for each block.

[0085] In step S906 of FIG. 9, the frequency band of the frequency spectrum calculated for each block can be selected according to the type of disease.

[0086] Although the embodiments have been described above with reference to the drawings, the present disclosure is not limited to such examples. It is clear that a person skilled in the art can conceive of various modifications or alterations within the scope of the claims. It is understood that such modifications or alterations also fall within the technical scope of the present disclosure. Furthermore, the components in the embodiments may be combined in any manner without departing from the spirit of the present disclosure.

[0087] (1) A dynamic image analysis device according to one embodiment of the present disclosure includes an acquisition unit that acquires a dynamic image of a subject who has been photographed in dynamic photography, and an analysis unit that performs non-stationary spectral analysis of the dynamic image acquired by the acquisition unit.

[0088] (2) In one embodiment of the dynamic image analysis device of the present disclosure, in the dynamic image analysis device of (1), the non-stationary spectral analysis is an analysis based on a non-stationary signal of a biological signal acquired from the dynamic image.

[0089] (3) In one embodiment of the present disclosure, the dynamic image analysis device is the dynamic image analysis device of (1), in which the dynamic images are captured to capture effects associated with cardiac pulsation.

[0090] (4) In one embodiment of the present disclosure, the dynamic image analyzer is the dynamic image analyzer of (2), wherein the unsteady signal is a signal whose frequency and / or amplitude change.

[0091] (5) In one embodiment of the dynamic image analysis device of the present disclosure, in the dynamic image analysis device of (1), the non-stationary spectral analysis is an analysis method for non-stationary signals, such as time-frequency analysis, multi-resolution analysis, or a combination of these.

[0092] (6) In one embodiment of the dynamic image analysis device of the present disclosure, in the dynamic image analysis device of (4) or (5), the non-stationary spectral analysis is a wavelet transform, a windowed fast Fourier transform, or a transform using a Wigner distribution.

[0093] (7) In one embodiment of the dynamic image analysis device of the present disclosure, in the dynamic image analysis device of (6), the non-stationary spectral analysis is a wavelet transform, and the analysis unit outputs a scalogram that has undergone wavelet transformation.

[0094] (8) In one embodiment of the dynamic image analysis device of the present disclosure, in the dynamic image analysis device of (6), the non-stationary spectral analysis is a wavelet transform, and the analysis unit outputs a frequency spectrum generated from a scalogram that has undergone wavelet transformation.

[0095] (9) In one embodiment of the present disclosure, the dynamic image analysis device is the dynamic image analysis device of (1), wherein the dynamic image is a dynamic image captured while the subject is holding their breath.

[0096] (10) In one embodiment of the dynamic image analysis device of the present disclosure, in the dynamic image analysis device of (1), the analysis unit blocks the dynamic image acquired by the acquisition unit, and the non-steady-state spectral analysis is non-steady-state spectral analysis for each block of the blocked dynamic image.

[0097] (11) In one embodiment of the dynamic image analysis device of the present disclosure, in the dynamic image analysis device of (10), the block size of the blocking is determined according to at least one of the spatial resolution and the intensity resolution to be obtained.

[0098] (12) In one embodiment of the present disclosure, the dynamic image analysis device is the dynamic image analysis device of (8), wherein the analysis unit calculates the intensity of a predetermined band of the frequency spectrum.

[0099] (13) In one embodiment of the present disclosure, the dynamic image analyzer is the dynamic image analyzer of (12), wherein the predetermined band is selectable.

[0100] (14) A dynamic image analysis method according to an embodiment of the present disclosure acquires a dynamic image of a subject who has been photographed in dynamic photography, and performs non-stationary spectral analysis on the acquired dynamic image.

[0101] (15) In one embodiment of the present disclosure, a dynamic image analysis method is the dynamic image analysis method described in (14), wherein the non-stationary spectral analysis is an analysis based on a non-stationary signal of a biological signal obtained from the dynamic image.

[0102] (16) In one embodiment of the present disclosure, the dynamic image analysis method is the dynamic image analysis method described in (14), in which the dynamic image captures the influence of signals associated with cardiac pulsation.

[0103] (17) In one embodiment of the present disclosure, the dynamic image analysis method is the dynamic image analysis method described in (15), wherein the unsteady signal is a signal in which at least one of frequency and amplitude changes.

[0104] (18) In one embodiment of the present disclosure, the dynamic image analysis method is the dynamic image analysis method described in (14), wherein the non-stationary spectral analysis is a time-frequency analysis, a multi-resolution analysis, or a combination of these analysis methods for non-stationary signals.

[0105] (19) In one embodiment of the present disclosure, the dynamic image analysis method is the dynamic image analysis method described in (17) or (18), wherein the non-stationary spectral analysis is a wavelet transform, a windowed fast Fourier transform, or a transform using a Wigner distribution.

[0106] (20) In one embodiment of the dynamic image analysis method of the present disclosure, in the dynamic image analysis method described in (19), the non-stationary spectral analysis is a wavelet transform, and a scalogram obtained by performing the wavelet transform is output.

[0107] (21) In one embodiment of the dynamic image analysis method of the present disclosure, in the dynamic image analysis method described in (19), the non-stationary spectral analysis is a wavelet transform, and a frequency spectrum generated from a scalogram subjected to the wavelet transform is output.

[0108] (22) In one embodiment of the present disclosure, the dynamic image analysis method is the dynamic image analysis method described in (14), wherein the dynamic image is a dynamic image captured while the subject is holding their breath.

[0109] (23) In one embodiment of the dynamic image analysis method of the present disclosure, in the dynamic image analysis method described in (14), the non-stationary spectral analysis is performed by dividing the acquired dynamic image into blocks and performing non-stationary spectral analysis on each block of the blocked dynamic image.

[0110] (24) In one embodiment of the dynamic image analysis method of the present disclosure, in the dynamic image analysis method described in (23), the block size of the blocking is determined according to at least one of the spatial resolution and intensity resolution to be obtained.

[0111] (25) In one embodiment of the present disclosure, the dynamic image analysis method according to (21) further comprises calculating the intensity of a predetermined band of the frequency spectrum.

[0112] (26) In one embodiment of the present disclosure, in the dynamic image analysis method according to (25), the predetermined band is selectable.

[0113] (27) A dynamic image analysis program according to an embodiment of the present disclosure causes a computer to acquire a dynamic image of a subject who has been photographed in dynamic photography, and to perform non-stationary spectral analysis on the acquired dynamic image.

[0114] (28) In one embodiment of the present disclosure, the dynamic image analysis program is the program of (27), wherein the non-stationary spectrum analysis is an analysis based on a non-stationary signal of a biological signal acquired from the dynamic image.

[0115] (29) In one embodiment of the present disclosure, the dynamic image analysis program is the program of (27), in which the dynamic image captures the influence of signals accompanying cardiac pulsation.

[0116] (30) In one embodiment of the present disclosure, the dynamic image analysis program is the program of (28), wherein the unsteady signal is a signal in which at least one of frequency and amplitude changes.

[0117] (31) In one embodiment of the dynamic image analysis program of the present disclosure, in the program of (27), the non-stationary spectral analysis is an analysis method for non-stationary signals, such as time-frequency analysis, multi-resolution analysis, or a combination of these.

[0118] (32) In one embodiment of the present disclosure, the dynamic image analysis program is the program of (30) or (31), wherein the non-stationary spectral analysis is a wavelet transform, a windowed fast Fourier transform, or a transform using a Wigner distribution.

[0119] (33) In one embodiment of the present disclosure, the dynamic image analysis program is the program of (32), in which the non-stationary spectral analysis is a wavelet transform, and a scalogram obtained by the wavelet transform is output.

[0120] (34) In one embodiment of the dynamic image analysis program of the present disclosure, in the program of (32), the non-stationary spectral analysis is a wavelet transform, and a frequency spectrum generated from a scalogram that has undergone wavelet transform is output.

[0121] (35) In one embodiment of the present disclosure, the dynamic image analysis program is the program of (27), wherein the dynamic image is a dynamic image captured while the subject is holding their breath.

[0122] (36) In one embodiment of the dynamic image analysis program of the present disclosure, in the program of (27), the non-stationary spectral analysis is performed by dividing the acquired dynamic image into blocks and performing non-stationary spectral analysis on each block of the blocked dynamic image.

[0123] (37) In one embodiment of the present disclosure, the dynamic image analysis program of (36) is configured such that the block size of the blocking is determined according to at least one of the spatial resolution and the intensity resolution to be obtained.

[0124] (38) In one embodiment of the present disclosure, the dynamic image analysis program of (34) calculates the intensity of a predetermined band of the frequency spectrum.

[0125] (39) In one embodiment of the present disclosure, the dynamic image analysis program is the dynamic image analysis method described in (38), wherein the predetermined band is selectable. [Industrial Applicability]

[0126] The present disclosure is useful for dynamic image analysis devices, methods, and programs. [Explanation of symbols]

[0127] 100 Dynamic image analysis device 110 Processing circuit 111 Acquisition Department 112 Analysis Department 113 Classification Department 120 Input / output section 121 Input section 122 Output section 130 Communications Department 140 memory

Claims

1. an acquisition unit for acquiring a dynamic image of a subject that has been photographed in dynamic motion; an analysis unit that performs non-stationary spectrum analysis on the dynamic image acquired by the acquisition unit; A dynamic image analysis device comprising:

2. The non-stationary spectrum analysis is an analysis based on a non-stationary signal of a biological signal acquired from the dynamic image. The dynamic image analyzer according to claim 1 .

3. The dynamic image captures the effects of heartbeat. The dynamic image analyzer according to claim 1 .

4. The non-stationary signal is a signal in which at least one of frequency and amplitude changes. The dynamic image analyzer according to claim 2 .

5. The non-stationary spectrum analysis is an analysis method for a non-stationary signal, such as time-frequency analysis, multi-resolution analysis, or a combination thereof. The dynamic image analyzer according to claim 1 .

6. The non-stationary spectral analysis is a wavelet transform, a windowed fast Fourier transform, or a transform using a Wigner distribution. The dynamic image analyzer according to claim 4 or 5.

7. the non-stationary spectral analysis is a wavelet transform; The analysis unit outputs a scalogram that has been subjected to wavelet transformation. The dynamic image analyzer according to claim 6.

8. the non-stationary spectral analysis is a wavelet transform; The analysis unit outputs a frequency spectrum generated from the scalogram subjected to wavelet transformation. The dynamic image analyzer according to claim 6.

9. The dynamic image is a dynamic image captured while the subject is holding their breath. The dynamic image analyzer according to claim 1 .

10. The analysis unit divides the dynamic image acquired by the acquisition unit into blocks, The non-stationary spectral analysis is a non-stationary spectral analysis for each block of the blocked dynamic image. The dynamic image analyzer according to claim 1 .

11. The block size of the blocking is determined according to at least one of a spatial resolution and an intensity resolution to be obtained. The dynamic image analyzer according to claim 10.

12. the analysis unit calculates the intensity of a predetermined band of the frequency spectrum. The dynamic image analyzer according to claim 8.

13. The predetermined band is selectable. The dynamic image analyzer according to claim 12.

14. Acquire a dynamic image of the subject who has been photographed in dynamic motion; performing non-stationary spectral analysis on the acquired dynamic image; Dynamic image analysis methods.

15. The non-stationary spectrum analysis is an analysis based on a non-stationary signal of a biological signal acquired from the dynamic image. The dynamic image analysis method according to claim 14.

16. The dynamic image captures the influence of a signal accompanying heartbeat. The dynamic image analysis method according to claim 14.

17. The non-stationary signal is a signal in which at least one of frequency and amplitude changes. The dynamic image analysis method according to claim 15.

18. The non-stationary spectrum analysis is an analysis method for a non-stationary signal, such as time-frequency analysis, multi-resolution analysis, or a combination thereof. The dynamic image analysis method according to claim 14.

19. The non-stationary spectral analysis is a wavelet transform, a windowed fast Fourier transform, or a transform using a Wigner distribution. The dynamic image analysis method according to claim 17 or 18.

20. the non-stationary spectral analysis is a wavelet transform; Outputs the wavelet transformed scalogram. The dynamic image analysis method according to claim 19.

21. the non-stationary spectral analysis is a wavelet transform; Output the frequency spectrum generated from the wavelet transformed scalogram. The dynamic image analysis method according to claim 19.

22. The dynamic image is a dynamic image captured while the subject is holding their breath. The dynamic image analysis method according to claim 14.

23. The non-stationary spectrum analysis is performed by dividing the acquired dynamic image into blocks and performing non-stationary spectrum analysis on each of the blocks of the blocked dynamic image. The dynamic image analysis method according to claim 14.

24. The block size of the blocking is determined according to at least one of a spatial resolution and an intensity resolution to be obtained. The dynamic image analysis method according to claim 23.

25. Calculating the intensity of a predetermined band of the frequency spectrum; The dynamic image analysis method according to claim 21.

26. The predetermined band is selectable. The dynamic image analysis method according to claim 25.

27. On the computer, A dynamic image of the subject is acquired by dynamic photography; subjecting the acquired dynamic image to non-stationary spectral analysis; Dynamic image analysis program.

28. The non-stationary spectrum analysis is an analysis based on a non-stationary signal of a biological signal acquired from the dynamic image. The dynamic image analysis program according to claim 27.

29. The dynamic image captures the influence of a signal accompanying heartbeat. The dynamic image analysis program according to claim 27.

30. The non-stationary signal is a signal in which at least one of frequency and amplitude changes. The dynamic image analysis program according to claim 28.

31. The non-stationary spectrum analysis is an analysis method for a non-stationary signal, such as time-frequency analysis, multi-resolution analysis, or a combination thereof. The dynamic image analysis program according to claim 27.

32. The non-stationary spectral analysis is a wavelet transform, a windowed fast Fourier transform, or a transform using a Wigner distribution. The dynamic image analysis program according to claim 30 or 31.

33. the non-stationary spectral analysis is a wavelet transform; Output the wavelet transformed scalogram. The dynamic image analysis program according to claim 32.

34. the non-stationary spectral analysis is a wavelet transform; Output the frequency spectrum generated from the wavelet transformed scalogram. The dynamic image analysis program according to claim 32.

35. The dynamic image is a dynamic image captured while the subject is holding their breath. The dynamic image analysis program according to claim 27.

36. The non-stationary spectrum analysis is performed by dividing the acquired dynamic image into blocks and performing non-stationary spectrum analysis on each of the blocks of the blocked dynamic image. The dynamic image analysis program according to claim 27.

37. a block size of the blocking is determined according to at least one of a spatial resolution and an intensity resolution to be obtained; The dynamic image analysis program according to claim 36.

38. Calculating the intensity of a predetermined band of the frequency spectrum; The dynamic image analysis program according to claim 34.

39. The predetermined band is selectable. The dynamic image analysis program according to claim 38.

Images