WO2017150497A1 - 肺野病変の診断支援装置、該装置の制御方法及びプログラム - Google Patents
肺野病変の診断支援装置、該装置の制御方法及びプログラム Download PDFInfo
- Publication number
- WO2017150497A1 WO2017150497A1 PCT/JP2017/007668 JP2017007668W WO2017150497A1 WO 2017150497 A1 WO2017150497 A1 WO 2017150497A1 JP 2017007668 W JP2017007668 W JP 2017007668W WO 2017150497 A1 WO2017150497 A1 WO 2017150497A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- lung
- image
- lesion
- region
- interstitial pneumonia
- Prior art date
Links
- 210000004072 lung Anatomy 0.000 title claims abstract description 206
- 230000003902 lesion Effects 0.000 title claims abstract description 138
- 238000003745 diagnosis Methods 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 title claims description 65
- 208000029523 Interstitial Lung disease Diseases 0.000 claims abstract description 104
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 201000004071 non-specific interstitial pneumonia Diseases 0.000 claims description 70
- 230000001684 chronic effect Effects 0.000 claims description 29
- 230000003176 fibrotic effect Effects 0.000 claims description 27
- 208000005069 pulmonary fibrosis Diseases 0.000 claims description 19
- 201000010099 disease Diseases 0.000 claims description 14
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 14
- 238000003384 imaging method Methods 0.000 claims description 12
- 206010035664 Pneumonia Diseases 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 6
- 230000004069 differentiation Effects 0.000 abstract description 10
- 230000002093 peripheral effect Effects 0.000 abstract description 10
- 210000000481 breast Anatomy 0.000 abstract 2
- 208000036971 interstitial lung disease 2 Diseases 0.000 description 122
- 201000009794 Idiopathic Pulmonary Fibrosis Diseases 0.000 description 99
- 238000004458 analytical method Methods 0.000 description 64
- 206010014561 Emphysema Diseases 0.000 description 14
- 238000012545 processing Methods 0.000 description 14
- 210000000038 chest Anatomy 0.000 description 13
- 201000003838 Idiopathic interstitial pneumonia Diseases 0.000 description 12
- 208000019693 Lung disease Diseases 0.000 description 11
- 230000008569 process Effects 0.000 description 11
- 150000003431 steroids Chemical class 0.000 description 10
- 206010016654 Fibrosis Diseases 0.000 description 9
- 230000004761 fibrosis Effects 0.000 description 9
- 208000027932 Collagen disease Diseases 0.000 description 8
- 210000001370 mediastinum Anatomy 0.000 description 7
- 238000004393 prognosis Methods 0.000 description 7
- 201000009803 desquamative interstitial pneumonia Diseases 0.000 description 6
- 239000003814 drug Substances 0.000 description 6
- 238000001914 filtration Methods 0.000 description 6
- 230000007170 pathology Effects 0.000 description 6
- 238000013399 early diagnosis Methods 0.000 description 5
- 239000005337 ground glass Substances 0.000 description 5
- 230000001575 pathological effect Effects 0.000 description 5
- 230000004202 respiratory function Effects 0.000 description 5
- 238000001574 biopsy Methods 0.000 description 4
- 238000007596 consolidation process Methods 0.000 description 4
- 201000009805 cryptogenic organizing pneumonia Diseases 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000000491 multivariate analysis Methods 0.000 description 4
- 210000004204 blood vessel Anatomy 0.000 description 3
- 230000001413 cellular effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000010191 image analysis Methods 0.000 description 3
- 238000011158 quantitative evaluation Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 229940124597 therapeutic agent Drugs 0.000 description 3
- 210000000115 thoracic cavity Anatomy 0.000 description 3
- 208000025247 virus-associated trichodysplasia spinulosa Diseases 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 208000031071 Hamman-Rich Syndrome Diseases 0.000 description 2
- 206010058467 Lung neoplasm malignant Diseases 0.000 description 2
- 238000000585 Mann–Whitney U test Methods 0.000 description 2
- 201000004073 acute interstitial pneumonia Diseases 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000017531 blood circulation Effects 0.000 description 2
- 210000000621 bronchi Anatomy 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000004195 computer-aided diagnosis Methods 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000003748 differential diagnosis Methods 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 230000002526 effect on cardiovascular system Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000008595 infiltration Effects 0.000 description 2
- 238000001764 infiltration Methods 0.000 description 2
- 201000005202 lung cancer Diseases 0.000 description 2
- 208000020816 lung neoplasm Diseases 0.000 description 2
- 210000005259 peripheral blood Anatomy 0.000 description 2
- 239000011886 peripheral blood Substances 0.000 description 2
- 210000004224 pleura Anatomy 0.000 description 2
- 230000000241 respiratory effect Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000001356 surgical procedure Methods 0.000 description 2
- 206010006448 Bronchiolitis Diseases 0.000 description 1
- 206010061818 Disease progression Diseases 0.000 description 1
- 206010067472 Organising pneumonia Diseases 0.000 description 1
- 208000004756 Respiratory Insufficiency Diseases 0.000 description 1
- 206010039710 Scleroderma Diseases 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000001154 acute effect Effects 0.000 description 1
- 230000009798 acute exacerbation Effects 0.000 description 1
- 206010069351 acute lung injury Diseases 0.000 description 1
- 230000002300 anti-fibrosis Effects 0.000 description 1
- 230000003510 anti-fibrotic effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000000090 biomarker Substances 0.000 description 1
- 208000023367 bronchiolitis obliterans with obstructive pulmonary disease Diseases 0.000 description 1
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000005750 disease progression Effects 0.000 description 1
- 210000002950 fibroblast Anatomy 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000012333 histopathological diagnosis Methods 0.000 description 1
- 238000003703 image analysis method Methods 0.000 description 1
- 238000003706 image smoothing Methods 0.000 description 1
- 229960003444 immunosuppressant agent Drugs 0.000 description 1
- 239000003018 immunosuppressive agent Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- XZXHXSATPCNXJR-ZIADKAODSA-N nintedanib Chemical compound O=C1NC2=CC(C(=O)OC)=CC=C2\C1=C(C=1C=CC=CC=1)\NC(C=C1)=CC=C1N(C)C(=O)CN1CCN(C)CC1 XZXHXSATPCNXJR-ZIADKAODSA-N 0.000 description 1
- 229960004378 nintedanib Drugs 0.000 description 1
- 238000002640 oxygen therapy Methods 0.000 description 1
- 229960003073 pirfenidone Drugs 0.000 description 1
- ISWRGOKTTBVCFA-UHFFFAOYSA-N pirfenidone Chemical compound C1=C(C)C=CC(=O)N1C1=CC=CC=C1 ISWRGOKTTBVCFA-UHFFFAOYSA-N 0.000 description 1
- 201000003144 pneumothorax Diseases 0.000 description 1
- 238000010837 poor prognosis Methods 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000002685 pulmonary effect Effects 0.000 description 1
- 230000009257 reactivity Effects 0.000 description 1
- 210000001034 respiratory center Anatomy 0.000 description 1
- 201000004193 respiratory failure Diseases 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000000391 smoking effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000001225 therapeutic effect Effects 0.000 description 1
- 210000003437 trachea Anatomy 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
- A61B6/03—Computed tomography [CT]
Definitions
- the present invention relates to a diagnosis support apparatus for lung field lesions such as idiopathic interstitial pneumonia, a control method for the apparatus, and a program.
- Idiopathic pulmonary fibrosis a typical disease of chronic interstitial pneumonia, is a disease in which the lungs become fibrotic and hard, resulting in respiratory failure (Non-Patent Documents 1 and 2). Histologically, it is characterized by a normal interstitial pneumonia (UIP) pattern. There is currently no cure for the disease, a slowly progressive and fatal disease.
- therapeutic agents for IPF such as Pirfenidone and nintedanib have been developed, and it has been shown that the progression of the disease can be suppressed by these agents (Non-Patent Documents 3 to 5).
- it is expected that early diagnosis and treatment of IPF will lead to an improvement in survival rate, so accurate assessment of the severity and progression of the disease will be required for early diagnosis and evaluation of the effects of internal treatment. It is becoming.
- Non-Patent Documents 3 to 5 Forced vital capacity (FVC) is a representative index.
- FVC Forced vital capacity
- CT CT has the advantage that it can be performed even in cases where these conditions are poor.
- Non-Patent Document 6 it is known that CT findings gradually progress as the disease progresses.
- the severity of CT findings correlates well with respiratory function test data or life prognosis (Non-Patent Documents 7 to 10). Based on these findings, quantitative evaluation of CT findings is expected to be a biomarker that measures the severity of IPF or the degree of disease progression.
- Non-Patent Document 11 there is a method of separating a lesion from a normal lung with a simple CT value threshold (Non-Patent Document 11).
- AFM Adaptive multiple feature method
- Non-Patent Document 12 Adaptive multiple feature method
- smart vector machine using 25 feature values
- GHNC GaussianGaHistogram Normalized Correlation segmentation
- Non-patent Document 23 There is also no known computer-aided diagnosis technology that can differentiate between IPF and fNSIP.
- An object of the present invention is to provide a novel diagnostic support means that is excellent in diagnostic performance of lung field lesions such as interstitial pneumonia and that can also distinguish between IPF and NSIP, which has been very difficult in the past.
- the present invention provides an image acquisition unit that acquires a chest tomographic image obtained by imaging a subject; and extracts a lung border region from the chest tomographic image at an arbitrarily specified depth from the pleural surface.
- a lung margin region extraction unit; a feature amount acquisition unit that acquires one or more feature amounts from the lung margin region; and a lesion in the lung margin region is identified based on the acquired one or more feature amounts
- a diagnosis support device for lung field lesions comprising: a lesion identification unit; and an output unit that outputs a lesion identification result.
- the present invention also provides an image acquisition step of acquiring a chest tomographic image obtained by imaging a subject; and extracting a lung peripheral region from the chest tomographic image at an arbitrarily specified depth from the pleural surface; A lung marginal region extracting step; acquiring one or more feature amounts from the lung marginal region; and a feature amount acquiring step; identifying a lesion in the lung marginal region based on the acquired one or more feature amounts.
- a control method for a lung field lesion diagnosis support apparatus comprising: a lesion identification step; and an output step of outputting a lesion identification result.
- the present invention provides a program for causing a computer to execute each step of the control method of the lung field lesion diagnosis support apparatus of the present invention.
- the present invention provides a computer-readable recording medium on which the program of the present invention is recorded.
- the present invention provides a diagnosis support apparatus for lung field lesions that greatly improves the diagnostic performance of interstitial pneumonia and enables discrimination between IPF and NSIP.
- Various computer-aided diagnosis techniques for interstitial pneumonia have been reported as described above, but no technique that focuses on the peripheral region of the lung is known.
- the present inventors have found that it is possible to distinguish between IPF and NSIP based on the pixel value intensity of the differential image of the CT image, but there is a region where the pixel value intensity of the differential image is a certain value or more.
- the method itself using the ratio of the share as an index is also new.
- CT image that can be used as a teacher image in GHNC, and shows an original image and differential image of CT, and a histogram obtained from these images. It is a figure explaining the process of creating a Gaussian histogram from CT image (axial position discontinuity) of a subject's lung and its differential image in GHNC. It is a figure explaining the process of classifying the image of a lung of a subject by contrast with a teacher pattern (CT original image of a teacher image, and a Gaussian histogram of a differential image) in GHNC. It is a block diagram which shows schematic structure of an example of the diagnosis assistance apparatus of this invention.
- the lung field lesion targeted in the present invention is typically interstitial pneumonia.
- Interstitial pneumonia is roughly classified into idiopathic and secondary. Hereinafter, the classification of interstitial pneumonia will be described.
- the interstitial pneumonia targeted in the present invention includes these interstitial pneumonia.
- Idiopathic interstitial pneumonia is large, (1) Idiopathic pulmonary fibrosis (IPF) with histological features of normal interstitial pneumonia (UIP) pattern, (2) idiopathic nonspecific interstitial pneumonia (idiopathic nonspecific interstitial pneumonia; idiopathic NSIP or INSIP), (3) acute interstitial pneumonia (AIP), (4) Idiopathic organizing pneumonia (COP), (5) desquamative interstitial pneumonia (DIP), (6) Respiratory bronchiolitis-associated interstitial lung disease (RBILD) associated with respiratory bronchiolitis, an interstitial pneumonia associated with smoking Pleuroparenchymal fibroelastosis; PPFE is mentioned as a very special disease type (Non-patent Document 1).
- NSIP idiopathic NSIP
- cellular NSIP cellular NSIP
- fibrotic NSIP fibrosing NSIP
- IIP chronic fibrosing IP
- IPF includes chronic fibrosing IP and other unexplained interstitial pneumonia (including interstitial pneumonia in the acute course), and chronic fibrosing IP includes IPF and NSIP. included.
- a typical example of secondary interstitial pneumonia is interstitial pneumonia associated with collagen disease (combined with rheumatism, scleroderma, SLE, etc.) and is also called collagen disease lung.
- the collagen disease lung has many NSIP pathological patterns, but some have UIP patterns, and both are often mixed.
- Collagen disease lungs are usually expected to be treated to some extent with immunosuppressants and steroids, but cases of collagen disease lungs with UIP patterns are known to have poor prognosis, and there is no difference between IPF and prognosis There is also a report (Song JW, et al., EstChest. 2009; 136 (1): 23-30.).
- interstitial pneumonia characterized by a UIP pattern can be distinguished from other interstitial pneumonia and other diseases, so secondary interstitial pneumonia such as collagen disease lung is also present. It is included in the object of the invention.
- Interstitial pneumonia which has not yet been found autoantibodies and is currently diagnosed as idiopathic, can be classified as secondary interstitial pneumonia such as collagen disease lung, is also idiopathic or secondary As interstitial pneumonia, it is included in the subject of the present invention.
- airspace ⁇ enlargement with fibrosis is known as a disease that is difficult to differentiate from interstitial pneumonia (Kawabata Y, Hoshi E, Murai K, et al. Histopathology. 2008; 53 (6): 707-714.). This is because the emphysema wall is a little thick and not interstitial pneumonia, but there is a considerable amount of confusion with interstitial pneumonia on the image. Many lung cancer cases are complicated by emphysema and AEF, but some cases are accompanied by interstitial pneumonia.
- diagnosis of interstitial pneumonia includes the above-described diagnosis of interstitial pneumonia. Specifically, diagnosis of interstitial pneumonia (differentiation between interstitial pneumonia and normal, or other lung diseases such as emphysema and AEF), diagnosis of IIP (IIP and normal, or emphysema, AEF, etc.) (Differentiation from other lung diseases), diagnosis of chronic fibrosing IP (differentiation from chronic fibrosing IP and normal, or other lung diseases such as emphysema and AEF), diagnosis of IPF (IPF and normal, or emphysema and AEF) (Differentiation from other lung diseases such as), diagnosis of IPF and NSIP (differentiation between IPF and NSIP and normal, or other lung diseases such as emphysema and AEF, or IPF and NSIP in chronic fibrosing IP cases And diagnosis of pulmonary fibrosis (differentiation between pulmonary fibrosis (differentiation between
- IPF patients patients who should be treated with steroids
- IPF patients patients who should not be treated
- NSIP patients patients who should not be treated
- IPF patients Early diagnosis and early treatment of IPF patients who should receive antifibrotic drugs will be possible.
- early diagnosis of interstitial pneumonia complications showing UIP pattern in lung cancer etc. and AEF in secondary interstitial pneumonia such as collagen disease lung Diagnosis makes it possible to take appropriate measures early and contribute to improving prognosis.
- pulmonary fibrosis when simply saying “pulmonary fibrosis”, it means interstitial pneumonia showing a UIP pattern, and not only idiopathic pulmonary fibrosis (IPF) but also other stroma showing a UIP pattern. Pneumonia (including secondary interstitial pneumonia) is included.
- the peripheral region of the lung is the analysis target.
- the depth of the pleural surface when extracting the marginal region is not particularly limited, but is usually within a few mm from the pleural surface, and may be selected within a range of, for example, 5 mm. Specifically, an area of 5 mm, 4 mm, 3 mm, 2 mm, or 1 mm from the pleural surface may be extracted as a lung marginal area.
- the pleural surface depth may be selected from a range of 1.5 mm to 3.5 mm, such as, but not limited to, a range of 2 mm to 3 mm.
- the “pleura” is the visceral pleura.
- the extraction of the lung marginal region may be performed by, for example, extracting the lung region from the lung field image and then extracting from the lung contour (lung surface) to a certain depth.
- a lung extraction method in image diagnosis in addition to a method based on threshold processing and region expansion processing, a method of applying lung field contour correction using morphology calculation and rib extraction results, lung field atlas was used.
- a technique for recognizing an abnormal part of a lung field using a technique and a texture feature and extracting the lung field is known (for example, Wang J, Li F, Li Q. Med Phys 2009; 36: 4592-4599). Any method may be used to extract the lung region.
- the blood vessel region and the tracheal / bronchial region extraction step can be omitted.
- the chest tomographic image can be taken by CT or MRI.
- MRI for example, a ratio with a reference signal such as a signal of cerebrospinal fluid may be measured to determine various threshold values described later.
- a CT image can be preferably used.
- the direction of the cross section of the tomographic image is not particularly limited, but axial cuts and sagittal cuts are preferred, and axial cuts are particularly preferred.
- the CT image is preferably a thin slice CT image taken by thin layer scanning.
- a thin slice CT image composed of a plurality of scan images can be obtained by imaging using a multi-slice CT apparatus, for example.
- the slice thickness of the thin-layer slice CT image is usually several millimeters or less, for example, about 0.5 to 1 mm.
- a lung marginal region may be extracted and analyzed. More accurate analysis is possible by using all of the thin slice CT images obtained by slicing the lung in millimeters, but only a part (for example, about several to a dozen) may be used.
- CT values intensity, average value, variance, histogram (distribution), histogram skewness, kurtosis, gray level entropy, etc.
- differential image pixel values intensity, histogram (distribution)) Etc.
- run ⁇ length that is, a length of continuous pixels (such as short run emphasis, long run emphasis, gray level uniformity, run length nonuniformitiy, run percent), co-occurrence matrix, and the like.
- a lesion is identified by combining one or more of these.
- the method for identifying a lesion based on the feature amount is not particularly limited.
- various methods for supporting diagnosis of interstitial pneumonia by image analysis for the entire lung are known.
- the analysis target region is limited to the lung margin region, but those known methods can be applied to the computer analysis of the image itself.
- various texture analysis methods for pattern classification of a lesion are known.
- lung regions are classified into patterns such as normal, honeycomb lung, ground glass lesion, emphysematous lesion / emphysema based on various feature amounts obtained from image data.
- image data There are many methods that use a typical image of each lesion as a teacher image.
- an image of a lung marginal area can be analyzed using such a known technique, and the lung marginal area can be classified and evaluated as one or more lesions including a normal part and a honeycomb lung.
- specific examples of the texture analysis method will be given.
- the image analysis method applicable in the present invention is not limited to these.
- GHNC Gaussian Histogram Normalized Correlation
- Non-patent Document 17 a Gaussian function is convolved with the local histogram of the CT image and its differential image to form a Gaussian histogram, and the normalized correlation with the teacher image is taken to classify and quantify the lesion.
- Non-patent Document 18 Iwao et al. Developed a method for evaluating this three-dimensionally (Non-patent Document 18). Iwasawa has published a paper that applies these techniques to clinical cases and correlates with respiratory function and prognosis (Non-Patent Documents 19 to 21).
- Non-Patent Document 22 when analyzing the whole lung with GHNC, it was practically difficult to differentiate between IPF and NSIP from the analysis results. These problems can be solved by analyzing the marginal area rather than the entire lung.
- GHNC itself has already been reported and publicly known as described above.
- the analysis procedure is shown below.
- a number of typical images are registered in advance as samples (teacher images), and lesion histograms are created using the original CT images and differential images of the samples (FIG. 1).
- a differential image is an image in which adjacent portions with large changes in pixel values are displayed in white, and can be easily created from image data using a known algorithm or software.
- a histogram for each pixel of the image to be analyzed is created (FIG. 2).
- an image value of 50 pixels around a certain pixel is used.
- a gentle histogram is obtained by multiplying each pixel value by a Gaussian function. This is called a Gaussian histogram.
- a normalized correlation between the Gaussian histogram and a histogram of a sample prepared in advance (also referred to as a Gaussian histogram) is taken, and this pixel is classified into the best matching group (FIG. 3).
- the lungs are classified into each lesion in pixel units, so it is easy to display them as areas or volumes.
- a Gaussian histogram for each pixel is created for the marginal areas of the CT original image and differential image, and compared with the sample Gaussian histogram. do it.
- the sample that is a teacher image in GHNC may be selected by a radiologist who is proficient in interpretation of lung field lesions. Since the CT value of the lungs is affected by the distribution of blood flow, it tends to be low even at normal, low at the apex of the lung, high at the bottom of the lung, low at the ventral side, and high at the dorsal side. Desirably, samples are set for a total of nine regions on the ventral side, the middle side, and the back side for each of the three locations in the lung bottom.
- ground glass lesions, consolidation, and reticulated lesion samples are from NSIP cases (cases where steroids have been effective and lesions have improved), and honeycomb lung samples. May be selected from IPF cases.
- the lung border region is analyzed, but the sample may be selected from the lung border region or from a region other than the border.
- the teacher image may basically be selected from images having the same shooting and reconstruction conditions as the analysis target image. Noise in the high spatial frequency band increases in images taken with low doses and images using reconstruction functions that enhance edges. If the teacher image and the analysis target image have almost the same shooting and reconstruction conditions, and there is no significant difference in the amount of noise in the high spatial frequency band, the analysis target image is not filtered. Pattern classification can be performed. For example, if the image selected as the teacher image is an image reconstructed under the mediastinum condition and the analysis target image is also an image reconstructed under the mediastinum condition, the filtering process is not necessary. If the teacher image is an image reconstructed under the lung field condition, the pattern under the lung field condition can be classified without filtering.
- a teacher image is an image that is taken at a normal dose and reconstructed under mediastinal conditions, while an analysis target image is taken at a low dose or reconstructed under lung field conditions and has a high spatial frequency.
- pattern classification can be performed using the teacher image under the mediastinum condition as it is by processing the analysis target image with a filter that can reduce high spatial frequency noise.
- filters are known, and for example, a Gaussian filter can be preferably used.
- the Gaussian filter is a filter used for image smoothing, and has the effect of reducing high spatial frequency noise (Yoshinobu Sato, “Smoothing and Enhancement”, Medical Image Engineering Handbook, supervised by the Japanese Society for Medical Image Engineering, p395-418, (2012) ).
- Non-patent Document 14 Another preferred example of texture analysis is CALIPER (Computer Aided Lung Informatics for Pathology Evaluation and Rating) reported by the Mayo Clinic group (Non-patent Document 14).
- k patterns such as normal, ground glass lesions, reticular lesions, honeycomb lungs, emphysematous lesions / emphysema are given as teacher images, and a CT value histogram (CT value distribution) is used as a feature value.
- CT value histogram CT value distribution
- the whole lung is classified into k cluster using Earth Mover's Distance between the analysis object and the teacher image.
- output the centroid and weight of each cluster results from multicenter research using this method have also been reported (Non-Patent Documents 15 and 16).
- the above CALIPER is also a technique that uses teacher images. Selection of teacher images can be performed in the same manner as the selection of GHNC samples. Using the histogram of CT values of the lung border region extracted from the chest CT image, the correlation with the teacher image may be evaluated as described above, and the lung border may be classified into k types of patterns.
- a method of identifying a lesion such as a honeycomb lung using a neural network there is a method of identifying a lesion such as a honeycomb lung using a neural network.
- Specific examples include Adaptive multiple feature method (AMFM) (17) reported by Uppaluri et al., And 25 feature s reported by Rosas et al.
- AMFM Adaptive multiple feature method
- There are a method for quantitative evaluation (Non-patent Document 13) and a method reported by Yoon et al. (Yoon RG, et al., Eur Radiol. 2013 Mar; 23 (3): 692-701.). These are also techniques using teacher images.
- the lesions in the marginal area are classified by GHNC and the proportion of the honeycomb lung in the marginal area is calculated.
- any analysis method that can identify and classify the honeycomb lung is used. It is possible and the scope of the present invention is not limited to GHNC.
- the ratio of the honeycomb lung to the marginal area is calculated.
- Diagnosis of interstitial pneumonia for example, diagnosis of chronic fibrotic interstitial pneumonia, diagnosis of pulmonary fibrosis, interstitial showing UIP pattern Diagnosis of pneumonia is performed.
- chronic fibrotic interstitial pulmonary inflammation cases the proportion of the honeycomb lung is larger than normal in the marginal region of the lung, and among them, the proportion of the honeycomb lung is higher in the IPF case than in the NSIP case.
- two thresholds can be used in a determination based on the proportion of honeycomb lungs.
- the first threshold value is a threshold value for distinguishing between chronic fibrotic interstitial pneumonia and pulmonary diseases (such as emphysema and AEF) other than normal and chronic fibrotic interstitial pneumonia.
- the first threshold value may also be a threshold value for differentiating pulmonary fibrosis and NSIP from pulmonary diseases (such as emphysema and AEF) other than normal and chronic fibrotic interstitial pneumonia.
- the second threshold value is higher than the first threshold value, and is a threshold value for distinguishing between NSIP and IPF, or NSIP and pulmonary fibrosis. Specific examples of these threshold values are shown in the following examples (see Tables 5 to 7).
- CT value of -700HU or higher (Matsuoka S, Yamashiro T, Matsushita S, et al. J Comput Assist Tomogr. 2015, 39 (2): 153-159), CT value of -600HU or higher ( Kliment CR, Araki T, Doyle TJ, et al. BMC Pulm Med. 2015, 15: 134, and Lederer DJ, Enright PL, Kawut SM, et al. Am J Respir Crit Care Med. 2009, 180 (5): 407 (5): 407 -414) is known.
- the inventors of the present application can newly differentiate between IPF and NSIP based on the pixel value intensity of a differential image of a CT image (hereinafter sometimes simply referred to as “differential image”). I found.
- the proportion of the lung marginal region occupied by a region where the pixel value is equal to or greater than a certain value (sometimes referred to as a “high pixel value region”) is calculated.
- the constant value is a value selected from the range of 100 to 120. For example, the ratio of the area where the pixel value is 100 or more, the area of 110 or more, or the area of 120 or more may be calculated.
- the differential image can be created by the same method as the differential image used in the above-described GHNC.
- diagnosis based on the pixel value intensity of the differential image it is generally preferable to use a CT image of the mediastinum condition, but even in images of lung field conditions and low-dose images that have more noise in the high spatial frequency band than the mediastinum condition, By performing filter processing to reduce noise in a high spatial frequency band as necessary, diagnosis by pixel value intensity can be performed in the same manner as a mediastinum condition image.
- diagnosis by pixel value intensity can be performed in the same manner as a mediastinum condition image.
- a pixel value intensity that defines a high pixel value region, or a threshold value relating to a ratio of the high pixel value region may be set separately.
- the diagnosis based on the pixel value intensity of the differential image is particularly suitable for distinguishing between IPF and NSIP in a case of chronic fibrotic interstitial lung inflammation. Therefore, the diagnosis based on the pixel value intensity of the differential image is preferably performed in combination with a diagnosis based on a feature amount (for example, a CT value of ⁇ 700 HU or more) that can distinguish normal from chronic fibrotic interstitial pneumonia.
- a feature amount for example, a CT value of ⁇ 700 HU or more
- doctors who are proficient in the interpretation of images of interstitial pneumonia can sufficiently help differentiate (idiopathic) pulmonary fibrosis from NSIP using only the differential image pixel values. It is not essential to use in combination with other indicators such as 700HU or more.
- NSF is effective in NSIP, but it is possible to select the administration target of a therapeutic agent such as steroid which should be avoided in IPF.
- IPF can be identified and diagnosed at an early stage by distinguishing it from NSIP, so that IPF can be treated early, and prognosis can be improved in patients with IPF.
- secondary interstitial pneumonia that exhibits a UIP pattern can also be distinguished from other interstitial pneumonia and other lung diseases. Prognostic improvement is expected by finding and selecting appropriate countermeasures early.
- the diagnosis support technology for lung lesions according to the present invention includes not only diagnosis of lung diseases represented by interstitial lung disease, but also diagnosis of severity, monitoring of disease state, antifibrosis drug or candidate substance thereof. It can also be used to determine therapeutic effects.
- FIG. 4 is a block diagram showing a schematic configuration of an example of the apparatus of the present invention.
- 5 and 6 are flowcharts for explaining processing by the apparatus of the present invention.
- FIG. 5 is a mode in which teacher images are not used (for example, lesion identification is performed using a CT value of ⁇ 700 HU or higher or a pixel value intensity of a differential image as an index.
- FIG. 6 shows a processing example according to an aspect using a teacher image (for example, an aspect in which lesion identification is performed by texture analysis such as GHNC).
- the filtering process of the acquired image is performed after acquiring the chest tomographic image (S101, S201) and acquiring the feature amount of the lung marginal region (S103, S203). You may carry out before.
- the image acquisition unit 110 acquires a chest tomographic image obtained by imaging a subject (for example, an interstitial pneumonia patient) with the imaging device 20.
- the imaging device 20 can be, for example, a high-resolution multi-slice CT device.
- CT devices There are various commercially available CT devices, but all devices are normally calibrated so that the CT value is 0HU for water and -1000HU for air, and 0HU, -1000HU, etc. by measuring water phantoms and air phantoms, etc. Any CT device may be used as long as the CT value is properly calibrated.
- the conditions for imaging and reconstruction are not particularly limited, and may be, for example, a lung field condition or a mediastinum condition, or a normal dose or a low dose.
- the apparatus 10 may include a filter processing unit 170. If necessary, the acquired image is processed with an appropriate filter. For example, in the aspect using the teacher image, when the amount of noise in the high spatial frequency band is greatly different between the image acquired by the image acquisition unit 110 and the teacher image used by the teacher pattern acquisition unit 160, as described above.
- the acquired image may be processed with a filter such as a Gaussian filter.
- the filtering process of an acquired image may be performed as needed.
- the lung border region extraction unit 120 extracts a lung border region from the chest tomographic image. Perform extraction at an arbitrarily specified depth from the pleural surface.
- the pleural surface depth is as described above, and can be a value usually within a few mm, for example within 5 mm, particularly within a range of 1.5 mm to 3.5 mm.
- a lung outline may be determined by extracting a lung region, and a marginal region may be extracted from the lung contour at a predetermined depth.
- a specific depth may be set in the device 10 in advance, or a user of the device may input or select a desired depth from the input device 30 and set it.
- the feature amount acquisition unit 130 acquires a feature amount from the image data of the lung border region. If image processing is required to acquire feature quantities, that processing is also performed. For example, when acquiring the pixel value intensity of a differential image as a feature amount, a process of generating a differential image from a CT image is also performed. In addition, when a CT value distribution (histogram) is acquired as a feature amount, processing for generating a histogram is also performed.
- the order of lung margin region extraction and feature amount acquisition is not particularly limited. After obtaining the feature amount from the entire lung, only the feature amount of the lung margin region may be extracted and acquired. From this point of view, it is preferable to obtain the feature amount only for the lung marginal region.
- the feature amount to be acquired, and thus the identification method by the lesion identification unit 140 may be configured so that the user of the device 10 can select from the input device 30.
- the lesion identification unit After the feature values of the lung margin region are obtained by the processing in the lung margin region extraction unit and the feature amount acquisition unit, the lesion identification unit performs lesion identification based on the feature amount. As described above, various texture analysis methods, lesion identification methods based on CT values of ⁇ 700 HU or higher or ⁇ 600 HU or higher, pixel value intensity of differential images, and the like can be applied.
- the teacher pattern acquisition unit 160 is necessary in an example of an apparatus that performs analysis using a teacher image such as texture analysis.
- a teacher image such as texture analysis.
- the teacher pattern acquisition unit 160 can be omitted. is there.
- the teacher pattern acquisition unit 160 may be provided so that texture analysis can also be performed.
- the teacher pattern acquisition unit 160 acquires one or a plurality of feature amounts as a teacher pattern of each lesion from a teacher image of a plurality of lung field lesions including honeycomb lungs.
- the lesion identification unit classifies the lesion region in the lung margin region by comparing the feature amount obtained from the lung margin region image of the subject with the teacher pattern.
- the teacher image may be selected by a radiologist who is proficient in interpretation of lung field lesions.
- the setting of the teacher image and the acquisition of the teacher pattern may be performed before the lesion identification / classification step by the lesion identification unit 140. Therefore, the teacher image may be set before the chest tomographic image of the subject is acquired. It may be performed in any step from the acquisition to the lesion identification / classification process by the lesion identification unit 140.
- the teacher image once set and the teacher pattern acquired from the teacher image can be used repeatedly when performing lesion classification using the same apparatus. Appropriate filtering is performed by the filter processing unit 170 so that the teacher image can be changed even if the image is captured or reconstructed under conditions different from those of the set teacher image and noise with high spatial frequency is emphasized.
- the analysis can be performed using the same teacher image and the same teacher pattern. However, a teacher image may be added or changed as desired.
- a plurality of sets of teacher images having different shooting / reconstruction conditions may be set in advance, and the teacher images may be selected according to the acquired image conditions.
- the identification result by the lesion identification unit 140 is output to the display device 40 such as a monitor by the output unit 150 and displayed. Furthermore, the identification result can be output to a printing apparatus such as a printer, a recording medium, or the like. Furthermore, the output unit 150 may be configured to output the identification result via a network to an external storage device such as a database existing outside the device 10.
- the lesion identification unit 140 performs texture analysis and identifies a lesion based on the ratio of the honeycomb lung pattern to the lung marginal region.
- the lesion identification unit 140 identifies and classifies one or more lesions including a normal part and a honeycomb lung in the lung margin region based on the feature amount acquired by the feature amount acquisition unit, The output unit outputs this classification result.
- the apparatus 10 may further include a teacher pattern acquisition unit 160 and a filter processing unit 170.
- the lesion identification unit 140 may calculate the proportion of the honeycomb lung in the lung marginal region, and the output unit outputs the calculated proportion of the honeycomb lung together with the classification result.
- the lesion identification unit 140 can compare the ratio of the honeycomb lungs with a predetermined first threshold value. When the ratio of the honeycomb lung is below the first threshold, the judgment is normal, and when it exceeds, it is chronic fibrotic interstitial pneumonia or pulmonary fibrosis and idiopathic nonspecific interstitial pneumonia It is determined that Furthermore, in addition to the comparison with the first threshold, the lesion identification unit 140 may also perform a comparison with a second threshold that is higher than the first threshold. Specific examples of the first threshold value and the second threshold value are as shown in the following examples. When the proportion of the honeycomb lung in the peripheral region of the lung exceeds the first threshold and is equal to or less than the second threshold, it is determined as idiopathic nonspecific interstitial pneumonia. When the proportion occupied by the honeycomb lungs also exceeds the second threshold value, (idiopathic) pulmonary fibrosis is determined. The output unit outputs these determination results.
- the lesion identification unit 140 uses a CT value intensity as a feature amount, and identifies a lesion using an CT value of ⁇ 700 HU or more or ⁇ 600 HU or more as an index.
- the lesion identification unit 140 calculates the ratio of the high CT value region having a CT value of ⁇ 700 HU or more or ⁇ 600 HU or more to the lung marginal region, and the output unit 150 identifies the calculated ratio as the identification result. Output as.
- the lesion identification unit 140 may calculate a ratio occupied by the high CT value region, and further compare the calculated ratio with a predetermined threshold value.
- a predetermined threshold value are as shown in the following examples.
- the lesion identification unit 140 is normal when the high CT value region is equal to or lower than the threshold, and is chronic fibrotic interstitial pneumonia when the threshold is exceeded, or pulmonary fibrosis and idiopathic It is determined that it is any of non-specific interstitial pneumonia, and the output unit 150 outputs this determination result.
- the lesion identification unit 140 uses the pixel value intensity of the differential image as a feature amount, and identifies the lesion using the pixel value intensity of 100 to 120 or more as an index. In this aspect, for example, the lesion identification unit 140 calculates the ratio of the region (high pixel value region) in which the pixel value of the differential image is equal to or greater than the value selected from the range of 100 to 120 in the lung margin region, The output unit 150 outputs the calculated ratio as an identification result.
- the lesion identification unit 140 may calculate a ratio occupied by the high pixel value region, and further compare the calculated ratio with a predetermined threshold value.
- a predetermined threshold value are as shown in the following examples.
- the lesion identification unit 140 is (idiopathic) pulmonary fibrosis when the high pixel value region exceeds the threshold value, and idiopathic nonspecific interstitial pneumonia when the high pixel value region is equal to or less than the threshold value.
- the output unit 150 outputs the determination result.
- the lesion identification unit 140 performs a combination of determination based on the ratio occupied by the high CT value region and determination based on the ratio occupied by the high pixel value region. For example, it is first determined whether or not it is chronic fibrotic interstitial pneumonia based on the ratio of the high CT value region, or whether it is pulmonary fibrosis or idiopathic nonspecific interstitial pneumonia. If it is determined that the patient is chronic fibrotic interstitial pneumonia or pulmonary fibrosis and idiopathic nonspecific interstitial pneumonia, the determination based on the ratio of the high pixel value region is performed. Is called.
- the output unit 150 outputs a determination result of normal, (idiopathic) pulmonary fibrosis, or idiopathic nonspecific interstitial pneumonia.
- the feature amount acquisition unit may acquire both the CT value intensity and the pixel value intensity of the differential image at the same time, or the determination based on the ratio occupied by the high CT value region is performed, and the determination result is normal When it is not the determination, the pixel value of the differential image may be acquired again.
- the lesion identification part 140 may perform simultaneously the determination based on the ratio for which a high CT value area accounts, and the determination based on the ratio for which a high pixel value area accounts.
- the determination result based on the ratio occupied by the high CT value area is normal, the determination result based on the high pixel value area is ignored and the determination result is output as normal, and the ratio occupied by the high CT value area is
- the determination result based on is a disease determination, the determination result based on the ratio occupied by the high pixel value region may be output.
- the present invention further provides a method for controlling a diagnosis support apparatus for lung field lesions.
- the control method acquires an image of a chest tomographic image obtained by imaging a subject; and an image acquisition step; extracts a lung marginal region from the thoracic tomographic image at an arbitrarily specified depth from the pleural surface; A marginal region extraction step; a feature amount acquisition step of acquiring one or more feature amounts from the lung marginal region; and a lesion in the lung marginal region is identified based on the acquired one or more feature amounts A lesion identification step; and an output step of outputting a lesion identification result.
- the control method acquires one or a plurality of feature amounts as a teacher pattern of each lesion from a teacher image of a plurality of lung field lesions including a honeycomb lung, A teacher pattern acquisition step may be further included.
- the present invention further provides a program for causing a computer to execute each step of the above-described control method of the lung field lesion diagnosis support apparatus, and a computer-readable recording medium on which the program is recorded.
- the “recording medium” may be any “portable physical medium” (non-transitory recording medium) such as a flexible disk, a magneto-optical disk, ROM, EPROM, EEPROM, CD-ROM, MO, and DVD.
- it may be a “communication medium” that holds a program in a short period of time, such as a communication line or a carrier wave when transmitting a program via a network, represented by a LAN, WAN, or the Internet.
- IIP idiopathic interstitial pneumonia
- IPF idiopathic pulmonary fibrosis
- NSIP Nonspecific interstitial pneumonia
- fNSIP Fibrotic NSIP
- cNSIP Cellular NSIP
- COP idiopathic organized pneumonia
- DIP exfoliative interstitial pneumonia
- UIP Normal interstitial pneumonia
- GHNC 1-1 Diagnosis of pulmonary lesions by GHNC 1-1. Selection of GHNC sample images Cases with idiopathic interstitial pneumonia confirmed by histopathological diagnosis of surgical lung biopsy by 2005 at Kanagawa Cardiovascular Respiratory Disease Center, and no interstitial pneumonia A GHNC sample image (a teacher image for acquiring a teacher pattern) was selected from a CT image of a chest volunteer in a normal volunteer case. In addition, Toshiba's 64-row multi-slice CT was used for CT image capturing.
- G, C, and R were extracted from cases where steroids were effective and lesions were improved by steroid treatment (specifically, fNSIP, COP, DIP, cNSIP (cases with acute lung injury)).
- H was extracted from IPF cases.
- FIG. 7 shows the range of lesions on the peripheral edge of the lung.
- the range of “margin” is the range except the mediastinal side on the lung surface and the inside of the diaphragm surface is also excluded. In other words, when the left and right lungs were roughly combined and considered as one cylinder, the portion corresponding to the side surface was set as the analysis target.
- 2 mm and 5 mm from the lung surface were examined. The area 2 mm from the lung surface was 8% of the whole lung, and the proportion of 5 mm was about 20%.
- FIG. 8a shows the ratio of the H pattern in the range of the edge 2 mm.
- IPF and fNSIP shows the ratio of the H pattern in the range of the edge 2 mm.
- Fig. 9 shows the ratio of the H pattern of the entire lung and the ratio of the H pattern with a margin of 2 mm.
- NSIP open circles
- IPF black circle
- the heterogeneity of the overall findings, subpleural fibrosis, and fibroblast nest are features of the UIP pattern that is an image / pathological pattern of IPF. When cell infiltration is large, it is suggested that the reactivity of steroids and the like may be good instead of the UIP pattern.
- CT value of -700 or higher or -600HU or higher, which was conventionally used as an index of interstitial pneumonia, correlated with both fibrosis and cell infiltration.
- the average value of the pixel value of the differential image at the edge and the ratio of the area of the differential image with a pixel value of 120 or more in the edge were also examined.
- the UIP pattern (IPF) had a higher average value of the differential image on the lung surface and a higher percentage of the differential image with a pixel value of 120 or more.
- Multivariate analysis was significant up to 2mm, 3mm, 4mm and 5mm.
- the correct diagnosis rate was high because the average pixel value was 3mm margin (correct diagnosis rate 74.5%), and the ratio of differential image pixel value 120 or more was margin 2mm (correct The diagnosis rate was 78.7%.
- FIGS. 10 to 12 are graphs summarizing the ratio of the area of 700HU or more and the ratio of the area of the differential image having a pixel value of 120 or more.
- Tables 5, 6, and 8 to 11 show the results of studying threshold setting by ROC analysis. In these tables, the most significant threshold values and the correct diagnosis rate when the threshold values are adopted are shown.
- Table 7 shows an example of diagnosis based on the H pattern ratio
- Table 12 shows a combination of CT value intensity ( ⁇ 700 HU or more) and differential image pixel value intensity (100 or more, 110 or more, 120 or more). An example of diagnosis by is shown.
- the ratio of H (honeycomb lung) pattern could be divided into 3 groups.
- the CT value -700HU or more the graphs of fNSIP group and IPF group are almost overlapped, and it is difficult to distinguish the two groups, but normal group and fNSIP + IPF group (chronic fibrotic interstitial pneumonia group) A distinction was possible.
- the ROC analysis results show that the fNSIP group and the IPF Groups could be distinguished significantly.
- the ROC analysis showed that the diagnosis at the ratio of the pixel value of 100 or more and the diagnosis of the ratio of the pixel value of 110 or more is also significant.
- normal, NSIP, and IPF can be distinguished, and with this index alone, diagnosis of chronic fibrotic interstitial pneumonia and identification diagnosis of NSIP and IPF can be performed. It was shown to be possible. In addition, it was shown that normal, NSIP, and IPF can be identified by using the CT value and the pixel value of the differential image in combination. That is, first of all, a diagnosis of chronic fibrotic interstitial pneumonia is performed using a ratio of CT value -700 HU or more as an index. It was shown that differential diagnosis between NSIP and IPF is possible by distinguishing between NSIP and IPF using a ratio of ⁇ 120 or more as an index. However, doctors who are proficient in the interpretation of images of interstitial pneumonia can sufficiently help differentiate NSIP from IPF using only the pixel values of differential images.
- Table 5 shows the results of examining the threshold of the ratio of the H pattern that can distinguish normal from chronic fibrotic interstitial pneumonia (fNSIP + IPF) by ROC analysis. It was possible to set a significant threshold value for any edge from 2 mm to 5 mm. For example, when analyzing an area with a margin of 2 mm, normal if the H pattern ratio is 20% or less, idiopathic interstitial pneumonia if it exceeds 20% (chronic fibrotic interstitial pneumonia) Can be diagnosed.
- Table 6 shows the result of examining the threshold of the ratio of the H pattern that can distinguish IPF and fNSIP by ROC analysis.
- a significant threshold could not be set at the edge of 5 mm, but a significant threshold could be set at the edge of 2 mm to 4 mm.
- the cases used in this analysis were those that were subjected to surgical biopsy because IPF or fNSIP could not be determined by images. Given this, the correct diagnosis rate of 60-70% is a very good number. be able to.
- Table 8 shows the threshold values of the ratios of CT values of -700 HU or higher (Table 8-1) and -600 HU or higher (Table 8-2) that can distinguish normal from chronic fibrotic interstitial pneumonia (fNSIP + IPF), respectively. It is the result investigated by ROC analysis. It was possible to set a significant threshold value for any edge from 2 mm to 5 mm.
- idiopathic interstitial pneumonia if it exceeds 60% Pneumonia, or if the ratio of CT value -600HU or higher is 42% or less, normal, and if it exceeds 42%, it is diagnosed as idiopathic interstitial pneumonia (chronic fibrotic interstitial pneumonia) be able to.
- Tables 9 to 11 show the results obtained by examining the threshold values in the case of discriminating IPF and NSIP based on the ratio of the regions where the pixel values of the differential image are 100 or more, 110 or more, and 120 or more by ROC analysis. In any case, it was possible to set a significant threshold for all of the edge 2 mm to the edge 5 mm. The correct diagnosis rate was equal to or more than the diagnosis by the ratio of H pattern (Table 6). In cases diagnosed with idiopathic interstitial pneumonia (chronic fibrotic interstitial pneumonia) based on a diagnosis based on a ratio of CT value ⁇ 700 HU or higher, by comparing with the threshold values shown in Table 9 to Table 11, It can be distinguished whether the case is IPF or NSIP. However, as described above, doctors who are proficient in the interpretation of images of interstitial pneumonia can sufficiently use the thresholds shown in Tables 9 to 11 to fully distinguish between NSIP and IPF.
- Diagnosis support apparatus 110 Image acquisition part 120 Edge area extraction part 130 Feature-value acquisition part 140 Lesion identification part 150 Output part 160 Teacher pattern acquisition part 170 Filter processing part 20 Imaging device 30 Input device 40 Display apparatus
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Engineering & Computer Science (AREA)
- Radiology & Medical Imaging (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Optics & Photonics (AREA)
- Pathology (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- High Energy & Nuclear Physics (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
Description
(1) 通常型間質性肺炎(UIP)パターンを組織学的な特徴とする特発性肺線維症(idiopathic pulmonary fibrosis; IPF)、
(2) 特発性非特異性間質性肺炎(idiopathic nonspecific interstitial pneumonia; 特発性NSIP、又はINSIP)、
(3) 急性間質性肺炎(acute interstitial pneumonia; AIP)、
(4) 特発性器質化肺炎(cryptogenic organizing pneumonia; COP)、
(5) 剥離性間質性肺炎(desquamative interstitial pneumonia ; DIP)、
(6) 喫煙に関連した間質性肺炎である、呼吸細気管支炎に伴うILD(Respiratory bronchiolitis-associated interstitial lung disease; RBILD)
の6種に分けられ、さらに非常に特殊な病型としてPleuroparenchymal fibroelastosis; PPFEが挙げられる(非特許文献1)。単にNSIPと言った場合、特発性NSIPを指す。特発性NSIPには、細胞浸潤が主体の細胞性NSIP(cellular NSIP; cNSIP)と、線維化主体の線維化性NSIP(fibrosing NSIP; fNSIP)がある。
IIP; 特発性間質性肺炎
IPF: 特発性肺線維症
NSIP: 非特異性間質性肺炎
fNSIP: 線維化性NSIP
cNSIP: 細胞性NSIP
COP: 特発性器質化肺炎
DIP: 剥離性間質性肺炎
UIP: 通常型間質性肺炎
1-1. GHNCサンプル画像の選択
神奈川県循環器呼吸器病センターにおいて2005年までに外科的肺生検の病理組織診断で特発性間質性肺炎が確定した症例、および間質性肺炎のない正常ボランティアの症例の胸部軸位断CT画像から、GHNCのサンプル画像(教師パターンを取得するための教師画像)を選択した。なお、CT画像の撮影には東芝社の64列マルチスライスCTを使用した。
上記のサンプル画像を用いたGHNCにより、サンプルを抽出した症例とは異なる症例を用いて、正常、IPF、及びfNSIPを鑑別できるかをレトロスペクティブに検討した。正常3例、IPF 24例、fNSIP 39例の軸位断マルチスライスCT画像をGHNC解析に付した。ここで用いたIPF症例及びfNSIP症例は、2006年から2011年までに神奈川県循環器呼吸器病センターで行われた外科的肺生検により病理が確定し、少なくとも3年間の経過観察で2次性の間質性肺炎が否定され、臨床的に特発性と診断された63症例であった。
次にIPFとNSIPの鑑別がGHNCで可能かを検討した。まず、Mann-WhitneyのU検定で各指標に差があるか検討した。
(2-1) 外科的生検サンプルの病理とCT解析結果との対比
上記1-2で用いた特発性間質性肺炎63症例(IPF 24例、fNSIP 39例)にて、病変のVATS(ビデオ補助下胸部手術)部位(101部位)の病理と、CTの解析結果との対比を行った。VATS前のCTとVATS後のCT画像からVATSされた部位を同定し、できる限り病理切片と同様の断面で(矢状断もしくは冠状断像)、局所の(平均332ピクセル、130mm2)の画素について検討した。できる限り病変のみが含まれるように関心領域を設定した。
病理切片でNSIPパターン(fNSIP)、UIPパターン(IPF)と診断された領域に対応する画像の特徴量がその診断と一致するかを検討した。結果を表4に示す。
正常群(17例)、fNSIP群(25例)、IPF群(23例)を対象に、胸膜表面深度1mm~5mmの領域に占めるHパターンの割合、CT値-700HU以上の領域の割合、及び微分画像の画素値が120以上の領域の割合をまとめたグラフを図10~図12に示す。またROC解析により閾値の設定を検討した結果を表5、表6、表8~11に示す。これらの表には、最も有意であった閾値、及びその閾値を採用した場合の正診率等を示している。また、表7には、Hパターンの割合に基づく診断の一例を、表12には、CT値強度(-700HU以上)及び微分画像の画素値強度(100以上、110以上、120以上)の組み合わせによる診断の一例を示した。
110 画像取得部
120 辺縁領域抽出部
130 特徴量取得部
140 病変識別部
150 出力部
160 教師パターン取得部
170 フィルター処理部
20 撮像装置
30 入力装置
40 表示装置
Claims (25)
- 被検体を撮影して得られた胸部断層画像を取得する画像取得部と、
胸部断層画像より、胸膜表面から任意に指定された深度で肺辺縁領域を抽出する、肺辺縁領域抽出部と、
肺辺縁領域から1又は複数の特徴量を取得する、特徴量取得部と、
取得された1又は複数の特徴量に基づき、肺辺縁領域内の病変を識別する、病変識別部と、
病変の識別結果を出力する出力部と
を備える、肺野病変の診断支援装置。 - 肺野病変が間質性肺炎である、請求項1記載の装置。
- 間質性肺炎が慢性線維化性間質性肺炎である、請求項2記載の装置。
- 間質性肺炎が肺線維症である、請求項2記載の装置。
- 任意に指定された深度が、胸膜表面から5mm以内の深度である、請求項1~4のいずれか1項に記載の装置。
- 任意に指定された深度が、胸膜表面から1.5mm~3.5mmの範囲から選択される、請求項5記載の装置。
- 肺辺縁領域抽出部は、胸部断層画像より肺領域を抽出し、次いで肺辺縁領域の抽出を行なう、請求項1~6のいずれか1項に記載の装置。
- 断層画像が軸位断画像である、請求項1~7のいずれか1項に記載の装置。
- 断層画像が薄スライス画像である、請求項1~8記載の装置。
- 断層画像がCT画像である、請求項1~9記載の装置。
- 病変識別部は、特徴量取得部が取得した特徴量に基づき、肺辺縁領域内の正常部及び蜂巣肺を含む1以上の病変部をそれぞれ識別して分類し、出力部は、病変識別部による分類結果を出力する、請求項1~10のいずれか1項に記載の装置。
- 病変識別部はさらに、肺辺縁領域において蜂巣肺が占める割合を算出し、出力部は、分類結果及び算出された蜂巣肺が占める割合を出力する、請求項11記載の装置。
- 病変識別部はさらに、前記蜂巣肺が占める割合と所定の第1の閾値との比較を行ない、蜂巣肺が占める割合が第1の閾値以下である場合には正常であり、第1の閾値を超える場合には慢性線維化性間質性肺炎である、又は肺線維症及び特発性非特異性間質性肺炎のいずれかである、と判定し、出力部は該判定の結果を出力する、請求項12記載の装置。
- 病変識別部は、前記蜂巣肺が占める割合と所定の第2の閾値との比較をさらに行ない、蜂巣肺が占める割合が第1の閾値を超え第2の閾値以下である場合には特発性非特異性間質性肺炎であり、第2の閾値を超える場合には肺線維症である、と判定し、出力部は該判定の結果を出力する、請求項13記載の装置。
- 蜂巣肺を含む複数の肺野病変の教師画像より、各病変の教師パターンとして1又は複数の特徴量をそれぞれ取得する、教師パターン取得部をさらに含み、
病変識別部は、特徴量取得部が取得した特徴量と教師パターンとの対比により、肺辺縁領域の病変部を分類する、請求項11~14のいずれか1項に記載の装置。 - 特徴量が、CT画像のCT値分布を含む、請求項11~15のいずれか1項に記載の装置。
- 特徴量が、CT画像の微分画像の画素値分布をさらに含む、請求項16記載の装置。
- 特徴量取得部は、肺辺縁領域からCT値強度を特徴量として取得し、病変識別部は、肺辺縁領域においてCT値-700HU以上又は-600HU以上の高CT値領域が占める割合を算出し、出力部は、算出された高CT値領域が占める割合を識別結果として出力する、請求項10記載の装置。
- 病変識別部はさらに、前記高CT値領域が占める割合と所定の閾値との比較を行ない、高CT値領域が該閾値以下である場合には正常であり、該閾値を超える場合には慢性線維化性間質性肺炎である、又は肺線維症及び特発性非特異性間質性肺炎のいずれかである、と判定し、出力部は該判定の結果を出力する、請求項18記載の装置。
- 特徴量取得部は、肺辺縁領域からCT画像の微分画像の画素値強度を取得し、病変識別部は、肺辺縁領域において該画素値が所定の値以上である高画素値領域が占める割合を算出し、出力部は、算出された高画素値領域が占める割合を識別結果として出力し、前記所定の値は、100~120の範囲から選択される値である、請求項10、18及び19のいずれか1項に記載の装置。
- 病変識別部はさらに、前記高画素値領域が占める割合と所定の閾値との比較を行ない、高画素値領域が該閾値を超える場合には肺線維症であり、該閾値以下である場合には特発性非特異性間質性肺炎である、と判定し、出力部は該判定の結果を出力する、請求項20記載の装置。
- 病変識別部は、高CT値領域が占める割合に基づく判定と、高画素値領域が占める割合に基づく判定とを組み合わせて行ない、出力部は、正常、肺線維症、及び特発性非特異性間質性肺炎のいずれであるかの判定結果を出力する、請求項21記載の装置。
- 被検体を撮影して得られた胸部断層画像を取得する、画像取得工程と、
胸部断層画像より、胸膜表面から任意に指定された深度で肺辺縁領域を抽出する、肺辺縁領域抽出工程と、
肺辺縁領域から1又は複数の特徴量を取得する、特徴量取得工程と、
取得された1又は複数の特徴量に基づき、肺辺縁領域内の病変を識別する、病変識別工程と、
病変の識別結果を出力する出力工程と
を含む、肺野病変の診断支援装置の制御方法。 - 請求項23記載の肺野病変の診断支援装置の制御方法の各工程をコンピュータに実行させるためのプログラム。
- 請求項24記載のプログラムを記録したコンピュータ読み取り可能な記録媒体。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2018503316A JP7075882B2 (ja) | 2016-03-02 | 2017-02-28 | 肺野病変の診断支援装置、該装置の制御方法及びプログラム |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2016-039752 | 2016-03-02 | ||
JP2016039752 | 2016-03-02 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2017150497A1 true WO2017150497A1 (ja) | 2017-09-08 |
Family
ID=59743811
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2017/007668 WO2017150497A1 (ja) | 2016-03-02 | 2017-02-28 | 肺野病変の診断支援装置、該装置の制御方法及びプログラム |
Country Status (2)
Country | Link |
---|---|
JP (1) | JP7075882B2 (ja) |
WO (1) | WO2017150497A1 (ja) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018175226A (ja) * | 2017-04-10 | 2018-11-15 | 富士フイルム株式会社 | 医用画像分類装置、方法およびプログラム |
JP2019070729A (ja) * | 2017-10-10 | 2019-05-09 | オリンパス株式会社 | 観察システム |
CN111598853A (zh) * | 2020-04-30 | 2020-08-28 | 科大讯飞股份有限公司 | 针对肺炎的ct影像评分方法、装置及设备 |
CN111612755A (zh) * | 2020-05-15 | 2020-09-01 | 科大讯飞股份有限公司 | 肺部病灶分析方法、装置、电子设备和存储介质 |
JP2020181250A (ja) * | 2019-04-23 | 2020-11-05 | 富士通株式会社 | びまん性症例画像判定プログラム、システム、及び方法 |
JP2020188872A (ja) * | 2019-05-20 | 2020-11-26 | 国立研究開発法人理化学研究所 | 判別装置、学習装置、方法、プログラム、学習済みモデルおよび記憶媒体 |
CN112053319A (zh) * | 2020-07-22 | 2020-12-08 | 清华大学 | 图像处理的方法和装置 |
JP2021019677A (ja) * | 2019-07-24 | 2021-02-18 | 富士通株式会社 | 教師画像生成プログラム、教師画像生成方法、および教師画像生成システム |
WO2021177799A1 (ko) * | 2020-03-05 | 2021-09-10 | (주)제이엘케이 | 이미지 기반 코로나바이러스감염증 정량 분류 방법 및 시스템 |
JP2021525929A (ja) * | 2018-05-27 | 2021-09-27 | エルシド バイオイメージング インコーポレイテッド | 定量的イメージングを利用するための方法およびシステム |
US20220172367A1 (en) * | 2020-11-27 | 2022-06-02 | Vida Diagnostics, Inc. | Visualization of sub-pleural regions |
US11607179B2 (en) | 2015-08-14 | 2023-03-21 | Elucid Bioimaging Inc. | Non-invasive risk stratification for atherosclerosis |
US11676359B2 (en) | 2015-08-14 | 2023-06-13 | Elucid Bioimaging Inc. | Non-invasive quantitative imaging biomarkers of atherosclerotic plaque biology |
US11696735B2 (en) | 2015-08-14 | 2023-07-11 | Elucid Bioimaging Inc. | Quantitative imaging for instantaneous wave-free ratio |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111476774B (zh) * | 2020-04-07 | 2023-04-18 | 广州柏视医疗科技有限公司 | 基于新型冠状病毒肺炎ct检测的智能征象识别装置 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009195279A (ja) * | 2008-02-19 | 2009-09-03 | Kobe Univ | 中皮腫とアスベスト肺に関する医用画像診断支援装置と画像診断支援プログラム |
JP2010035635A (ja) * | 2008-07-31 | 2010-02-18 | Takeshi Joko | Ardsの重症度のスコアを算出する方法、装置およびプログラム |
JP2011048481A (ja) * | 2009-08-25 | 2011-03-10 | Fuji Xerox Co Ltd | 画像処理装置及びプログラム |
JP2015136480A (ja) * | 2014-01-22 | 2015-07-30 | 富士フイルム株式会社 | 3次元医用画像表示制御装置およびその作動方法並びに3次元医用画像表示制御プログラム |
-
2017
- 2017-02-28 JP JP2018503316A patent/JP7075882B2/ja active Active
- 2017-02-28 WO PCT/JP2017/007668 patent/WO2017150497A1/ja active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009195279A (ja) * | 2008-02-19 | 2009-09-03 | Kobe Univ | 中皮腫とアスベスト肺に関する医用画像診断支援装置と画像診断支援プログラム |
JP2010035635A (ja) * | 2008-07-31 | 2010-02-18 | Takeshi Joko | Ardsの重症度のスコアを算出する方法、装置およびプログラム |
JP2011048481A (ja) * | 2009-08-25 | 2011-03-10 | Fuji Xerox Co Ltd | 画像処理装置及びプログラム |
JP2015136480A (ja) * | 2014-01-22 | 2015-07-30 | 富士フイルム株式会社 | 3次元医用画像表示制御装置およびその作動方法並びに3次元医用画像表示制御プログラム |
Non-Patent Citations (1)
Title |
---|
TAE IWASAWA: "Lung Volume: Pitfalls in Quantitative CT Evaluation", DIAGNOSTIC IMAGING THE JULY ISSUE, vol. 31, no. 8, 25 June 2011 (2011-06-25), pages 850 - 863 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11696735B2 (en) | 2015-08-14 | 2023-07-11 | Elucid Bioimaging Inc. | Quantitative imaging for instantaneous wave-free ratio |
US11676359B2 (en) | 2015-08-14 | 2023-06-13 | Elucid Bioimaging Inc. | Non-invasive quantitative imaging biomarkers of atherosclerotic plaque biology |
US11607179B2 (en) | 2015-08-14 | 2023-03-21 | Elucid Bioimaging Inc. | Non-invasive risk stratification for atherosclerosis |
JP2018175226A (ja) * | 2017-04-10 | 2018-11-15 | 富士フイルム株式会社 | 医用画像分類装置、方法およびプログラム |
JP2019070729A (ja) * | 2017-10-10 | 2019-05-09 | オリンパス株式会社 | 観察システム |
JP7113916B2 (ja) | 2018-05-27 | 2022-08-05 | エルシド バイオイメージング インコーポレイテッド | 定量的イメージングを利用するための方法およびシステム |
JP2021525929A (ja) * | 2018-05-27 | 2021-09-27 | エルシド バイオイメージング インコーポレイテッド | 定量的イメージングを利用するための方法およびシステム |
JP7326845B2 (ja) | 2019-04-23 | 2023-08-16 | 富士通株式会社 | びまん性症例画像判定プログラム、システム、及び方法 |
JP2020181250A (ja) * | 2019-04-23 | 2020-11-05 | 富士通株式会社 | びまん性症例画像判定プログラム、システム、及び方法 |
JP7334900B2 (ja) | 2019-05-20 | 2023-08-29 | 国立研究開発法人理化学研究所 | 判別装置、学習装置、方法、プログラム、学習済みモデルおよび記憶媒体 |
JP2020188872A (ja) * | 2019-05-20 | 2020-11-26 | 国立研究開発法人理化学研究所 | 判別装置、学習装置、方法、プログラム、学習済みモデルおよび記憶媒体 |
US11361443B2 (en) | 2019-05-20 | 2022-06-14 | Riken | Identifying device, learning device, method, and storage medium |
JP7275961B2 (ja) | 2019-07-24 | 2023-05-18 | 富士通株式会社 | 教師画像生成プログラム、教師画像生成方法、および教師画像生成システム |
JP2021019677A (ja) * | 2019-07-24 | 2021-02-18 | 富士通株式会社 | 教師画像生成プログラム、教師画像生成方法、および教師画像生成システム |
WO2021177799A1 (ko) * | 2020-03-05 | 2021-09-10 | (주)제이엘케이 | 이미지 기반 코로나바이러스감염증 정량 분류 방법 및 시스템 |
CN111598853A (zh) * | 2020-04-30 | 2020-08-28 | 科大讯飞股份有限公司 | 针对肺炎的ct影像评分方法、装置及设备 |
CN111598853B (zh) * | 2020-04-30 | 2024-02-13 | 讯飞医疗科技股份有限公司 | 针对肺炎的ct影像评分方法、装置及设备 |
CN111612755A (zh) * | 2020-05-15 | 2020-09-01 | 科大讯飞股份有限公司 | 肺部病灶分析方法、装置、电子设备和存储介质 |
CN112053319B (zh) * | 2020-07-22 | 2022-12-02 | 清华大学 | 图像处理的方法和装置 |
CN112053319A (zh) * | 2020-07-22 | 2020-12-08 | 清华大学 | 图像处理的方法和装置 |
US20220172367A1 (en) * | 2020-11-27 | 2022-06-02 | Vida Diagnostics, Inc. | Visualization of sub-pleural regions |
Also Published As
Publication number | Publication date |
---|---|
JP7075882B2 (ja) | 2022-05-26 |
JPWO2017150497A1 (ja) | 2019-02-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP7075882B2 (ja) | 肺野病変の診断支援装置、該装置の制御方法及びプログラム | |
CN101231678B (zh) | 医用图像处理装置及医用图像处理方法 | |
Volpicelli et al. | Semi-quantification of pneumothorax volume by lung ultrasound | |
Boedeker et al. | Emphysema: effect of reconstruction algorithm on CT imaging measures | |
JP5159242B2 (ja) | 診断支援装置、診断支援装置の制御方法、およびそのプログラム | |
Van Rikxoort et al. | A method for the automatic quantification of the completeness of pulmonary fissures: evaluation in a database of subjects with severe emphysema | |
JP7264486B2 (ja) | 画像解析方法、画像解析装置、画像解析システム、画像解析プログラム、記録媒体 | |
EP3142556A1 (en) | Visualization and quantification of lung disease utilizing image registration | |
JP2002521896A (ja) | 軟組織胸部画像と標準胸部画像を利用したエネルギ差分法によるコンピュータ化肺小結節検出 | |
JP5098393B2 (ja) | 関心領域決定装置 | |
US11915822B2 (en) | Medical image reading assistant apparatus and method for adjusting threshold of diagnostic assistant information based on follow-up examination | |
JP2007068993A (ja) | 解剖学的構造内の疾患に関する疾患関連の変化の患者特異的解析を実行するための方法及びシステム | |
KR102280902B1 (ko) | 흉부 의료 영상 판독 지원 장치 및 방법 | |
Choromańska et al. | Role of computed tomography in quantitative assessment of emphysema | |
Boehm et al. | Automated classification of normal and pathologic pulmonary tissue by topological texture features extracted from multi-detector CT in 3D | |
JP5456132B2 (ja) | 診断支援装置、診断支援装置の制御方法、およびそのプログラム | |
Yoshiyasu et al. | Radiomics technology for identifying early-stage lung adenocarcinomas suitable for sublobar resection | |
Yanagawa et al. | Commercially available computer-aided detection system for pulmonary nodules on thin-section images using 64 detectors-row CT: preliminary study of 48 cases | |
KR20100010973A (ko) | 폐질환 자동 분류 방법 | |
Xing et al. | Automatic detection of A‐line in lung ultrasound images using deep learning and image processing | |
US7440601B1 (en) | Automated identification of ileocecal valve | |
Talwar et al. | Pulmonary nodules: assessing the imaging biomarkers of malignancy in a “coffee-break” | |
WO2018086893A1 (en) | Apparatus for the detection of opacities in x-ray images | |
Sørensen et al. | Learning COPD sensitive filters in pulmonary CT | |
Yin et al. | Computer-aided identification of interstitial lung disease based on computed tomography |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 2018503316 Country of ref document: JP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 17759963 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 31.01.2019) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 17759963 Country of ref document: EP Kind code of ref document: A1 |