EP3973504A1 - Procede, dispositif et support lisible par ordinateur pour classifier automatiquement une lesion coronarienne selon la classification cad-rads par un reseau de neurones profond - Google Patents
Procede, dispositif et support lisible par ordinateur pour classifier automatiquement une lesion coronarienne selon la classification cad-rads par un reseau de neurones profondInfo
- Publication number
- EP3973504A1 EP3973504A1 EP20725201.6A EP20725201A EP3973504A1 EP 3973504 A1 EP3973504 A1 EP 3973504A1 EP 20725201 A EP20725201 A EP 20725201A EP 3973504 A1 EP3973504 A1 EP 3973504A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- images
- neural network
- coronary
- parts
- applied directly
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- 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]
- A61B6/032—Transmission computed tomography [CT]
-
- 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
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
- A61B6/503—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of the heart
-
- 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
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
- A61B6/504—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of blood vessels, e.g. by angiography
-
- 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
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
- A61B6/507—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for determination of haemodynamic parameters, e.g. perfusion CT
-
- 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
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
- A61B6/5217—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data extracting a diagnostic or physiological parameter from medical diagnostic data
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10081—Computed x-ray tomography [CT]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30048—Heart; Cardiac
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30096—Tumor; Lesion
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Definitions
- the present invention relates to a computer-implemented method for automatically detecting the presence or not of a coronary lesion and classifying it by assigning a value according to its severity, using a deep neural network, as well.
- a device capable of automatically detecting the presence or not of such a coronary lesion and a non-transient computer-readable medium storing computer-readable program instructions for automatically detecting the presence or not of such a coronary lesion are capable of automatically detecting the presence or not of such a coronary lesion and a non-transient computer-readable medium storing computer-readable program instructions for automatically detecting the presence or not of such a coronary lesion.
- Coronary artery disease is the second leading cause of death in developed countries, after cancer. In particular, it affects more than fifteen million Americans. It can turn out to be brutal, it is by far the leading cause of sudden death in the world. There are around 50,000 cases of sudden death per year in France, most of them from myocardial infarction. It can affect young people, sometimes in their thirties. Its incidence increases with the aging of the population and the development of chronic diseases such as diabetes and high blood pressure.
- Cardiovascular disease includes a number of disorders
- Coronary heart disease also called coronary heart disease
- coronary artery disease or coronary heart disease
- coronary artery disease is an obstructive disease of the coronary arteries, which supply blood to the heart.
- stenosis narrowing to occlusion
- coronary injury results in coronary artery disease, or coronary artery disease, or coronary artery disease.
- Coronary insufficiency usually results in myocardial ischemia, which means insufficient blood supply (ischemia) to the heart muscle (myocardium), due in part to vascular obstruction.
- myocardial the main ones being the electrocardiogram, stress test, MRI, myocardial scintigraphy, coronary angiography, and more recently the coronary angiogram.
- the image quality depends on the heart rate, any artifacts in staircase walking between two beats, the quality of the injection of the contrast product, the level of noise in the image and the possible presence of calcifications. .
- the latest cardiac scanner technology achieves the best average image quality while reducing most of the cited artifacts.
- the coronary CT angiography (or CCTA for Coronary CT angiography) is a
- CT angiography has poorer specificity (approximately 50-70%) due to frequent false-positive cases. Its positive predictive value of the scanner is therefore lower. False-positive cases are observed, in particular in the case of coronary calcifications and / or in the case of movement artifacts during image acquisition. Thus, a reading expertise is necessary to minimize the number of false positives.
- a good expertise can be acquired over several years (at least 5 years) for radiologists or cardiologists working in a center specializing in cardiac imaging.
- AI Artificial Intelligence
- Machine learning or Machine Learning tools in English, and in particular neural networks (RN) make it possible to reproduce expertise, which is widely used in the field of image recognition. This is why multiple projects are developing in the field of medical imaging.
- the present invention consists in particular in adapting an expertise in the reading of CT angiography using AI techniques.
- Computed tomography coronary angiography is a sensitive method for detecting coronary lesions (plaque or stenosis), which in practice allows coronary lesion to be ruled out when the examination is normal.
- an automatic examination detection normal classified CAD-RADS 0
- the present invention relates to an automated determination of the value according to the CAD-RADS classification (Cury RC et al., "Coronary Artery Disease - Reporting and Data System (CAD-RADS): An Expert Consensus Document of SCCT, ACR and NASCI: Endorsed by the ACC. ”, JACC Cardiovasc Imaging. 2016 Sep; 9 (9): 1099-1113.)
- CAD-RADS Coronary Artery Disease - Reporting and Data System
- the present invention relates to a computer-implemented method for automatically detecting the presence or absence of a coronary lesion and classifying it by assigning a value from 0 to 5 according to its severity, according to the CAD-RADS classification (for Coronary Artery Disease - Reporting and Data System value), using a neural network, as well as a device capable of automatically detecting the presence or not of a coronary lesion and classifying it in assigning a value from 0 to 5 depending on its severity, according to the CAD-RADS classification (for Coronary Artery Disease - Reporting and Data System value or system of reports and data) using a neural network, and a non-transient computer readable medium storing computer readable program instructions for automatically detecting the presence or not of a coronary lesion and classifying it by assigning a value from 0 to 5 according to of its severity, according to
- angiography are based on visual estimates of strictures, at a threshold of 50% in diameter, corresponding to the CAD-RADS classification 3, 4 or 5 (occlusion). The relevance of this detection is very dependent on the observer and his level of reading experience.
- a problem which the present invention proposes to solve consists in particular in limiting the interobserver interpretation variability of coronary stenosis linked to the expertise by determining the value according to the CAD-RADS classification directly on images. anatomical.
- the first object of the solution to this problem is a computer-implemented method for determining the presence of a coronary lesion for a patient, comprising:
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value or system of reports and data
- Its second object is a device capable of determining the presence of a coronary lesion in a patient, comprising:
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value or system of reports and data
- a final object of the invention is a non-transient computer readable medium storing computer readable program instructions for determining the presence of a coronary lesion for a patient, comprising the execution by a processor of instructions.
- computer readable programs that perform the following operations:
- CT scan curvilinear or stretched multiplanar medical image
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value
- the Applicant has in particular been able to develop a method which has qualities which make it possible to automatically detect a potentially significant coronary artery stenosis from a hemodynamic point of view (preferably CAD-RADS 3 or 4), on the basis of high-level expertise. level.
- the method also has the advantage of being able to predict the absence of coronary injury by determining a CAD-RADS value of 0 with a high probability. This system therefore makes it possible, in a single coronary CT angiogram examination, to provide reliable results which advantageously help in the diagnosis and subsequently to adapt therapeutic management.
- FIG. 1 represents a curvilinear RMP image of the coronary artery presenting a
- Figure 2 shows a stretched RMP image of coronary artery stenosis.
- Figure 3 is a block diagram illustrating the different possible steps of a process according to the invention.
- Figure 4 shows some of the anatomical criteria for predicting whether a stenosis is hemodynamic or not.
- FIG. 5 represents the results of the CAD-RADS 0 detection by the neural network.
- Figure 6 illustrates an example of analysis of multiple incidence images of the same artery analyzed by neural network.
- the first object of the invention is a computer-implemented process for
- the first step of said method is a step of receiving at least one curvilinear or stretched multiplanar medical image by computed tomography (X-ray scanner) of said patient including a coronary artery to be studied.
- a multiplanar image is an image reconstructed from the centerline of a tubular anatomical structure such as a coronary artery. The major axis of the image plane is then aligned with the anatomical structure by following this central line. This allows the structure to be included entire anatomy (here a coronary artery) in a single image.
- An RM P image can thus follow the curvilinear path of the vessel, and the adjacent structures are then distorted.
- the axis of the vessel can also be stretched by projection in a fixed direction. Visualization can be done on an axis of rotation of 360 ° in both cases (curvilinear RM P or stretched RMP).
- the second step of the process is a step of determining a
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value or system of reports and data
- the images or parts of images come from a coronary CT angiography (or CCTA for Coronary Computed Tomography Angiography).
- the first neural network is trained to read RMP images
- multiplanar reconstructions curvilinear, single images or, for diagnostic accuracy, multiple RMP images of the same artery viewed from multiple views, spaced at least 20 ° apart, preferably with at least 180 ° coverage.
- a base of at least 5,000, preferably 10,000, artery images were analyzed and captioned, with or without coronary or coronary injury, by a recognized expert with more than 20 years of experience in reading these images.
- Curvilinear or stretched RMP images are obtained from the centerline of a coronary artery. This central line is extracted by common software on X-ray workstations, but it is sometimes necessary to correct the central line manually so that this line always remains in the center of the circulating light.
- Each coronary artery is generally analyzed with multiple PMRs by multiplying the incidences over 360 °. In particular, this makes it easier to detect asymmetric lesions, which may only appear under certain conditions.
- the method according to the invention therefore includes a step of determining a value according to the CAD-RADS classification (for Coronary Artery Disease - Reporting and Data System value) of a coronary lesion by use of a first trained deep neural network applied directly to the detected images or parts of images.
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value
- Figure 6 illustrates an example of analysis of multiple incidence images of the same artery analyzed by neural network. The average of the probabilities by CAD-RADS classification of each image is calculated as well as the frequency of each
- a specific algorithm makes it possible to classify a coronary artery according to CAD-RADS from multiple images.
- the algorithm takes the most frequent classification of images of different incidences (from 0 to 5), and compares it to the average of the probabilities of each classification. If the most frequent classification is also the one with the highest average probability, this is retained by the algorithm. In the event of a discrepancy, the more severe classification of the two is retained: it is indeed better in clinical screening practice to overestimate a coronary lesion than to underestimate it, to avoid false-negative examinations.
- the algorithm eliminates probability scores below a certain decision threshold from calculations in order to optimize the diagnostic performance of the neural network.
- CAD-RADS 4 Five RMP images are classified CAD-RADS 4, two RMP images are classified CAD-RADS 3, two RMP images are classified CAD-RADS 2-
- the average CAD-RADS 4 probability is 0.8, that CAD-RADS 3 is 0.5, that of CAD-RADS 2 is 0.3.
- the lesion is classified CAD-RADS 4 because this classification is more frequent and its average probability is higher.
- the lesion will therefore be classified CAD-RADS 4 with a probability of 0.8.
- CAD-RADS 2 Five RMP images are classified CAD-RADS 2, four RMP images are classified CAD-RADS 3.
- the average CAD-RADS 2 probability is 0.6, that CAD-RADS 3 is 0.7.
- the lesion is classified CAD-RADS 3 because this classification is the most severe of the 2.
- the lesion will be classified CAD-RADS 3 with a probability of 0.7.
- the CAD-RADS classification allows a rational and standardized classification of coronary atheromatous lesions. This classification in six degrees of severity (from 0 to 5), makes it possible to propose optimal therapeutic choices for the patient in the light of the results of the coronary CT scan (Cury RC et al., “Coronary Artery Disease - Reporting and Data System (CA D-RADS): An Expert Consensus Document of SCCT , ACR and NASCI: Endorsed by the ACC. ”, JACC Cardiovasc Imaging. 2016
- CAD-RADS 1 plaque ⁇ 25%
- CAD-RADS 2 plaque between 25 and 49%
- CAD-RADS 3 50-69% stenosis
- categories 1 and 2 and categories 3 and 4 can be grouped as follows:
- CAD-RADS 1 or 2 non-obstructive coronary artery disease
- CAD-RADS 3 or 4 obstructive coronary artery disease
- the method according to the invention further comprises a step of predicting an interval of coronary reserve flow value by manual, semi-automated and / or automated measurement of at least two morphological criteria. chosen from:
- anatomical criteria are extracted from the image, manually, semi-automatically or automatically:
- a the minimum diameter of the stenosis in mm
- e the length of the stenosis in mm
- manual image extraction is meant a manual measurement of the diameter and minimum area of the vessel (on an image of an artery section) at the narrowest place of the stenosis, a tracing or contouring manual of diameter and area (on an artery section image), at a healthy artery segment closest to the stenosis; manual length measurement; a visual estimate of myocardial mass and the percentage of myocardium supplied by an artery downstream from a stenosis on that same artery.
- semi-automatic image extraction is meant a measurement obtained by prior creation of central lines from the pointing of a vessel by the user. At each point of the vessel, the values of the minimum diameter and the surface area of the vessel are displayed by an algorithm on the radiological workstation. The various parameters of interest are readable at the level of the area of interest with the possibility of manual correction of central lines and contours.
- Dedicated specific algorithms can calculate the volume of vascularized myocardium downstream of a stenosis (depending on the image processing software used).
- Dedicated specific algorithms can calculate the volume of vascularized myocardium downstream of a stenosis (depending on the image processing software used). [73]
- the combination of at least two of these criteria and the evaluation by neural network provides a prediction of the functional character by the FFR above or below the threshold of 0.8.
- the most relevant anatomical criteria for predicting an FFR value are:
- the myocardial mass downstream of a stenosis is downstream of a stenosis.
- the anatomical criteria for qualifying a stenosis are:
- Coronary reserve further includes the use of a second trained deep neural network applied directly to the detected images or parts of images.
- the second neural network is trained to read curvilinear or stretched RMP images, single images or, for more precision, multiple RMP images of the same stenosis according to several views, separated by at least 20 °, covering at least 180 °.
- the neural network has been successfully trained on real images in which the real value of FFR has been measured.
- a specific algorithm predicts the FFR of a given coronary artery from multiple images, at a threshold of 0.8.
- the algorithm takes the most frequent classification of images of different impacts, and compares it to the average of the probabilities of each classification. If the most frequent classification is also the one with the highest average probability, this is retained by the algorithm.
- the FFR + classification (FFR less than or equal to 0.8) is retained in order to avoid false negative results as much as possible, since it is considered more serious not to diagnose a lesion. more severe than overestimating a lesion in terms of screening.
- the method according to the invention advantageously further comprises at least one of the following steps, which can be performed in any order:
- an automated image quality determination step providing a diagnostic confidence index using a third trained neural network applied directly to the detected images or portions of images;
- a step of determining a high risk plaque (PHR or HRP) of a cardiac event using a fifth trained neural network, applied directly to the images or parts of images detected.
- An automatic image quality rating is useful for quality control, and for comparing images from center to center.
- this automatic evaluation is advantageously used to provide an index of diagnostic confidence in the final interpretation.
- the third neural network for the automated image quality determination step has been successfully trained by supervised learning on RMP images of arteries whose image qualities have been assessed by a recognized expert. The images were classified according to a score of 0 to 4 on the subjective scale (detailed below).
- the neural network provides an overall image quality score from one to nine images of the same artery.
- the algorithm takes the average of the classifications of the images of an artery according to different incidences (classified 0 to 4).
- the fourth network for the step of determining a global calcification score was trained directly on RMP images of arteries for which the Agatston score was known by a prior CT examination without injection of contrast product. ; After training, the degree of calcification is semi-quantitatively predicted on an injected CT angiogram according to four categories, for each of the arteries extracted:
- Moderate calcifications Agatston score predicted between 1 and 99
- Agatston Score is recognized as an important and independent risk marker for predicting the probability of coronary events, in the same way as known risk factors. like high cholesterol, diabetes or hypertension.
- the actual calcium score is first calculated on a CT scan without contrast injection, because high density contrast interferes with calcium and therefore does not allow a score calculation.
- Machine learning has been used precisely to estimate the score automatically on contrasted exams, learning from the score obtained on a non-contrast CT scan of the same patient.
- the Agatston calcium score predicted by the system will be between 200 and 400, corresponding to an intermediate risk.
- the fifth array for the high-risk plaque determination step was successfully trained by supervised learning on RMP images or section images of arteries, perpendicular to the RMP images, in which the possible presence of 'a vulnerable plaque was or was not detected by a recognized expert. After training, the presence of a vulnerable plaque is confirmed according to a probability threshold (between 0 and 1) calculated as the optimal threshold to obtain the best performance from this neural network. This performance is measured
- High-risk plaques are characterized by the presence of the following elements: low density plaque (LDP), positive remodeling plaque (PR) by increasing the vessel wall towards the exterior of the vessel, presence of an area of negative density within the plaque (lipid core). The presence of at least two of the first three criteria confirms the presence of a high-risk plaque.
- LDP low density plaque
- PR positive remodeling plaque
- a specific algorithm can determine the presence of a vulnerable plaque. Due to its asymmetric nature, a plate may not be visible on one or more of the RMP image views due to a different viewing angle.
- V The presence of a vulnerable plaque is noted V, if the probability threshold reaches or exceeds 0.5, its absence is noted 0.
- the method of the invention comprises the following five steps which can be performed in any order:
- an automated image quality determination step providing a diagnostic confidence index from a third trained neural network applied directly to the images or parts of images detected;
- a step of determining plaque at high risk of cardiac event using a fifth trained neural network applied directly to the images or parts of images detected.
- the detection of images or parts of images corresponding to the lesion of the patient of the method according to the invention comprises the detection of parts of images. corresponding to the coronary lesion, a coronary tree, coronary ostia or coronary vessels.
- a subject of the invention is also a device capable of determining the presence of a coronary lesion for a patient, comprising means for receiving at least one curvilinear or stretched multiplanar medical image by computed tomography (X-ray scanner), d a coronary artery of said patient;
- X-ray scanner computed tomography
- said device further comprises means for determining a value according to the CAD-RADS classification (for Coronary Artery Disease - Reporting and Data System value) of a coronary lesion on said image or on a part of said image by the use of a first trained deep neural network applied directly to the images or parts of images detected.
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value
- the first neural network is as described above.
- the device further comprises means for predicting an interval of coronary reserve flow value by use of a second trained deep neural network applied directly to the images or parts of images detected.
- the second neural network is as described above.
- the device according to the invention further comprises at least one of the following means:
- - automated image quality determination means providing a diagnostic confidence index by using a third trained neural network applied directly to the images or parts of images detected;
- the device comprises the following five means:
- - automated image quality determination means providing a diagnostic confidence index by using a third trained neural network applied directly to the images or parts of images detected;
- the invention relates to a non-transient computer readable medium storing computer readable program instructions for determining the presence of a coronary lesion for a patient, comprising the execution by a processor of instructions to computer-readable programs that perform the following operations:
- CT scan curvilinear or stretched multiplanar medical image
- CAD-RADS classification for Coronary Artery Disease - Reporting and Data System value or system of reports and data
- the support is able to further generate the realization by said
- processor of an operation of predicting an interval of coronary reserve flux value by use of a second trained deep neural network applied directly to the detected images or parts of images.
- the support is suitable for furthermore causing said processor to perform at least one of the following operations:
- the support generates the
- a large database of more than 10,000 RMP images from coronary CT angiography was used for supervised learning. All the images were classified and labeled by an expert with more than 20 years of experience in reading these images (corresponding to approximately 50,000 cases analyzed). The images in this database were classified in terms of image quality, degree of calcification, and degree of stenosis according to the CAD-RADS classification. The possible presence of risk plaque (vulnerable plaque) has been specified.
- a binary classification at the FFR threshold of 0.8 was performed on 4500 images of patients with stenosis and known FFR value. A neural network could thus be trained to predict on a new image whether the value of FFR will be greater or less (> or ⁇ ) than 0.8.
- Various neural networks available in open access have been tested, for example: GOOGLENET TM, RESNET TM, and INCEPTION TM V3, VGG11 TM, VGG13 TM, VGG19 TM, in order to obtain the best classification rate.
- Various measurements were carried out on a test database, independent of the training base: calculations of test precision, sensitivity, specificity, positive predictive value, negative predictive value, F1 score (harmonic mean between the sensitivity and positive predictive value), area under the ROC curve.
- RMP multiplanar images are provided by the scanner consoles of all different scanner manufacturers as standard, from the centerlines. Usually curvilinear or stretched RMP images, and one to nine images of the same artery. Images can be exported from workstations in a DICOM or other radiological standard image format (eg: JPEG, PNG, ...), and secondarily uploaded to a dedicated website. They can also be loaded directly onto an Internet site from the workstation. The produced evaluation result is then returned to the reader's usual environment.
- DICOM or other radiological standard image format
- a large stretched or curvilinear RMP image database was created from examinations performed on a 64-slice scanner, a 256-slice scanner, and a 320-slice scanner at four different institutions.
- Each RMP image was classified by an expert with regard to image quality, degree of calcification, presence or absence of a vulnerable plaque, and degree of stenosis using the CAD-RADS classification. .
- Figure 5 illustrates the results of the CAD-RADSO detection by the network of
- the detection sensitivity of normal arteries is 275/275 + 35 or 89%.
- the specificity is 495/495 + 27 or 95%.
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Public Health (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Pathology (AREA)
- High Energy & Nuclear Physics (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Heart & Thoracic Surgery (AREA)
- Optics & Photonics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Quality & Reliability (AREA)
- Databases & Information Systems (AREA)
- Physiology (AREA)
- Vascular Medicine (AREA)
- Cardiology (AREA)
- Pulmonology (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| FR1905408A FR3096497B1 (fr) | 2019-05-23 | 2019-05-23 | Procédé, dispositif et support lisible par ordinateur pour classifier automatiquement une lésion coronarienne selon la classification CAD-RADS par un réseau de neurones profond |
| PCT/EP2020/063798 WO2020234233A1 (fr) | 2019-05-23 | 2020-05-18 | Procede, dispositif et support lisible par ordinateur pour classifier automatiquement une lesion coronarienne selon la classification cad-rads par un reseau de neurones profond |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP3973504A1 true EP3973504A1 (fr) | 2022-03-30 |
Family
ID=68210917
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP20725201.6A Pending EP3973504A1 (fr) | 2019-05-23 | 2020-05-18 | Procede, dispositif et support lisible par ordinateur pour classifier automatiquement une lesion coronarienne selon la classification cad-rads par un reseau de neurones profond |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US12033323B2 (fr) |
| EP (1) | EP3973504A1 (fr) |
| FR (1) | FR3096497B1 (fr) |
| WO (1) | WO2020234233A1 (fr) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP4589528A3 (fr) | 2019-08-05 | 2025-10-22 | Lightlab Imaging, Inc. | Affichage longitudinal de la charge calcique de l'artère coronaire |
| CN111709925B (zh) * | 2020-05-26 | 2023-11-03 | 深圳科亚医疗科技有限公司 | 用于血管斑块分析的装置、系统及介质 |
| US11775822B2 (en) * | 2020-05-28 | 2023-10-03 | Macronix International Co., Ltd. | Classification model training using diverse training source and inference engine using same |
| CN112700421B (zh) * | 2021-01-04 | 2022-03-25 | 推想医疗科技股份有限公司 | 冠脉图像分类方法及装置 |
| CN112967234B (zh) * | 2021-02-09 | 2022-12-09 | 复旦大学附属中山医院 | 冠状动脉功能生理学病变模式定量评价方法 |
| US12586180B2 (en) * | 2021-06-18 | 2026-03-24 | AI Optics Inc. | Assessment of image quality for a medical diagnostics device |
| WO2024178199A1 (fr) * | 2023-02-22 | 2024-08-29 | Q Bio, Inc. | Moteur de recommandation dynamique |
| CN116681890B (zh) * | 2023-05-31 | 2025-07-11 | 同济大学 | 一种基于目标检测的血管狭窄病变识别方法及其应用 |
| WO2025262634A1 (fr) | 2024-06-21 | 2025-12-26 | Universita' Degli Studi Di Torino | Procédé et système de traitement d'image angiographique |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101779224B (zh) * | 2007-08-03 | 2016-04-13 | 皇家飞利浦电子股份有限公司 | 用于绘制并显示3d管状结构的曲面重组视图的方法、装置和系统 |
| WO2019025270A1 (fr) * | 2017-08-01 | 2019-02-07 | Siemens Healthcare Gmbh | Évaluation non invasive et guidage de thérapie pour une coronaropathie dans des lésions diffuses et en tandem |
-
2019
- 2019-05-23 FR FR1905408A patent/FR3096497B1/fr active Active
-
2020
- 2020-05-18 WO PCT/EP2020/063798 patent/WO2020234233A1/fr not_active Ceased
- 2020-05-18 US US17/613,455 patent/US12033323B2/en active Active
- 2020-05-18 EP EP20725201.6A patent/EP3973504A1/fr active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| FR3096497B1 (fr) | 2021-04-30 |
| WO2020234233A1 (fr) | 2020-11-26 |
| FR3096497A1 (fr) | 2020-11-27 |
| US20220215541A1 (en) | 2022-07-07 |
| US12033323B2 (en) | 2024-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3973503B1 (fr) | Procede, dispositif et support lisible par ordinateur pour detecter automatiquement une stenose coronarienne hemodynamiquement significative | |
| EP3973504A1 (fr) | Procede, dispositif et support lisible par ordinateur pour classifier automatiquement une lesion coronarienne selon la classification cad-rads par un reseau de neurones profond | |
| US20230394654A1 (en) | Method and system for assessing functionally significant vessel obstruction based on machine learning | |
| US12484871B2 (en) | System for and method of identifying coronary artery disease | |
| US11389130B2 (en) | System and methods for fast computation of computed tomography based fractional flow reserve | |
| US12079988B2 (en) | Medical image reconstruction apparatus and method for screening for plurality of types of lung diseases | |
| US12380560B2 (en) | Systems, methods, and devices for image-based plaque analysis and risk determination | |
| EP3564963A1 (fr) | Système et procédés de calcul rapide de reserve de débit fractionnaire à base de tomodensitométrie | |
| Omori et al. | Determination of lipid-rich plaques by artificial intelligence-enabled quantitative computed tomography using near-infrared spectroscopy as reference | |
| Kawasaki et al. | Evaluation of significant coronary artery disease based on CT fractional flow reserve and plaque characteristics using random forest analysis in machine learning | |
| FR3082981A1 (fr) | Determination de la confiance d'une mesure dans un clip video d'imagerie medicale basee sur la qualite de l'image du clip video | |
| US20250329015A1 (en) | Systems, methods, and devices for plaque analysis, vessel and fluid flow analysis, and/or risk determination or prediction thereof | |
| Revathi et al. | A deep learning based approach for diagnosing coronary inflammation with multi-scale coronary response dynamic balloon tracking (MSCAR-DBT) based artery segmentation in coronary computed tomography angiography (CCTA) | |
| Teßmann et al. | Automatic detection and quantification of coronary calcium on 3D CT angiography data | |
| EP3566204B1 (fr) | Procede de determination d'un indice d'heterogeneite de la perfusion myocardique a partir d'images medicales | |
| Kapoor et al. | Comparison of Hemodynamic Significant Coronary Stenosis Using Artificial Intelligence and Conventional Cross Sectional Area in CT Coronary Angiography | |
| EP4706013A1 (fr) | Systèmes, procédés et dispositifs d'analyse de plaque basée sur une image et de détermination de risque | |
| Ayala et al. | Studying hypertension in ocular fundus images using Hausdorff dispersion ordering | |
| WO2024107691A1 (fr) | Systèmes, dispositifs et procédés d'analyse de plaque basée sur une image non invasive et de détermination de risque | |
| FR3161496A1 (fr) | Procédé de détermination de la densité mammaire | |
| WO2024137796A1 (fr) | Systèmes, dispositifs et méthodes d'analyse de plaque basée sur une image non invasive et de détermination de risque | |
| WO2005050569A2 (fr) | Procede pour generer des donnees d'imagerie medicale 3d | |
| Doğan | Pulmonary embolism and right ventricular function: cardiac CT research: right |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20211223 |
|
| AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) | ||
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: EXAMINATION IS IN PROGRESS |
|
| 17Q | First examination report despatched |
Effective date: 20241107 |