Method and apparatus for determining presence of a tumor

Abstract

A method (200) and apparatus (100) for determining the presence of a tumor are provided. In one aspect, the method (200) includes receiving a medical image associated with the patient, wherein the medical image includes a region of interest associated with the patient. Further, the method (200) includes identifying one or more blood vessels associated with the region of interest in the medical image. Additionally, the method (200) includes determining a set of characteristics associated with the blood vessels using a trained machine learning model. The method (200) also includes determining whether the blood vessels are feeding blood vessels associated with a tumor based on the set of characteristics associated with the blood vessels. Further, the method (200) includes detecting a tumor region in the region of interest based on the feeding blood vessels if the blood vessels are feeding blood vessels associated with a tumor.

Description

Technical Field

[0001] This invention relates to methods and apparatus for determining the presence of tumors in a patient's body. Background Technology

[0002] Cancer is one of the leading causes of death worldwide. Therefore, it takes time to effectively treat tumors. One method of treating tumors includes embolization, a procedure involving injecting material directly into an artery to block or reduce blood flow to a tumor in a particular area. Current methods for detecting blood vessels rely on the doctor's experience to identify vessels that appear to nourish a tumor (also known as feeder fessels). Therefore, the physical and biological characteristics of such vessels may not be considered during an embolization procedure. This can lead to normal tissue being identified as tumor tissue, damaging healthy tissue within the patient and missing areas of tumor.

[0003] Currently, there is no effective and accurate method for identifying blood vessels supplying cancer and associated tumor regions for cancer treatment. Summary of the Invention

[0004] Therefore, the object of the present invention is to provide a method and apparatus that can effectively determine the presence of tumors in a patient's body.

[0005] This invention achieves this objective by a method for determining the presence of a tumor in a patient. The method includes receiving medical images associated with the patient. The medical images include regions of interest associated with the patient. In one embodiment, the medical images may be computed tomography (CT) angiography images, X-ray angiography images, or magnetic resonance imaging (MRI) images. The medical images can be received from medical imaging equipment such as X-ray imaging devices, computed tomography devices, MRI devices, etc. The region of interest imaged in the medical images may be an area in the patient's body suspected of containing a tumor. In another embodiment, the region of interest in the medical images may include one or more organs associated with the patient, the vascular system associated with the patient, and / or bone or tissue information associated with the patient.

[0006] The method also includes identifying one or more blood vessels associated with the region of interest. Specifically, the blood vessels to be identified may include vessels that supply blood to tumors present in the region of interest. In one embodiment, one or more segmentation techniques known to those skilled in the art can be used to segment blood vessels in a medical image. For example, a region growing algorithm can be used to segment blood vessels. Region growing algorithms segment blood vessels based on pixel information associated with blood vessels in a medical image. In another embodiment, a topology-preserving thinning algorithm can be used to skeletonize the segmented blood vessels. The advantage is that skeletonizing the blood vessels preserves the topology of the vessels.

[0007] The method also includes determining a set of characteristics associated with blood vessels. This can be performed using a trained machine learning model. This set of characteristics may include, for example, the diameter associated with the blood vessel, the branching of the blood vessel, and / or the tortuosity of the blood vessel. Since the blood vessels supplying blood to the tumor are generated by the tumor itself, the morphological characteristics associated with such blood vessels differ from those of normal blood vessels. For example, in the case of a tumor vascular system, the diameter associated with the blood vessel is irregular or non-uniform. Similarly, the branching of such blood vessels is anomalous, and the vascular network is disordered; that is, there may be no hierarchy in blood vessels such as arteries, arterioles, capillaries, venules, and veins. Therefore, a machine learning model can be trained to detect such anomalous characteristics in blood vessels segmented from medical images. Furthermore, the method also includes, if the blood vessel is a blood-supplying vessel associated with the tumor, then based on the set of characteristics associated with the blood vessel. A blood-supplying vessel is a blood vessel that supplies blood to the tumor. The method also includes detecting tumor regions in a region of interest based on the blood-supplying vessels. In one embodiment, the tumor region can be identified based on the blood-supplying vascular network in the region of interest. An advantage is that the method can accurately identify the tumor vascular system, thereby making the embolization process precise and more effective. Another advantage of the invention is that it reduces damage to healthy tissues associated with the patient.

[0008] According to one embodiment, determining whether a blood vessel is a tumor-associated feeding vessel includes using a trained machine learning model to measure the diameter associated with the vessel. In one embodiment, the diameter may be the Euclidean distance between two points, wherein the first of the two points lies on a parallel wall of the vessel from the second of the two points. The method also includes determining whether the diameter of the vessel decreases as it branches into branch vessels. In a normal vascular system, the diameter of the parent vessel is always larger than the diameter of the branch vessels. Therefore, as a vessel progresses from its parent vessel, its diameter should decrease. However, this may not hold true for feeding vessels. The method also includes classifying the vessel as a tumor-associated feeding vessel if its diameter does not decrease as it branches into branch vessels. An advantage is that the machine learning model enables efficient identification of feeding vessels based on their diameter. Another advantage of the invention is that vessels associated with a patient's healthy tissue are not damaged due to cancer treatment.

[0009] According to another embodiment, determining whether a blood vessel is a tumor-associated feeding vessel includes determining the radius associated with that blood vessel. Furthermore, the radii associated with branch vessels originating from the blood vessel are also determined. Branch vessels largely conform to Murray's law, according to which the cube of the vessel radius is equal to the sum of the cubes of the branch vessel radii. Therefore, the method also includes calculating the cube of the radius associated with the blood vessel and the cube of the radius associated with the branch vessels. Additionally, it is determined whether the cube of the vessel radius is equal to the sum of the cubes of the branch vessel radii. If the vessel radius is not equal to the sum of the cubes of the branch vessel radii, then a trained machine learning model is used to classify the blood vessel as a tumor-associated feeding vessel. An advantage is that the machine learning model enables the determination of the presence of irregular branches associated with the blood vessel. This can be a characteristic of tumor-associated feeding vessels. Therefore, embolization can be performed precisely.

[0010] According to another embodiment, determining whether a blood vessel is a feeding vessel associated with a tumor includes determining the tortuosity of the blood vessel. The tortuosity of the blood vessel is determined based on the angular deviation of the blood vessel from the straight path associated with it. This can be performed using a trained machine learning model. Tortuosity allows for understanding whether a blood vessel is, in fact, disordered. In one embodiment, the tortuosity of the blood vessel can be determined using a distance metric or an angle metric. The method also includes comparing the tortuosity of the blood vessel with predetermined criteria associated with the blood vessel using a trained machine learning model. For example, if a blood vessel has two or more consecutive bends with angles greater than or equal to 180°, the blood vessel can be considered severely tortuous. Furthermore, the method includes classifying the blood vessel as a feeding vessel if the tortuosity of the blood vessel does not match the predetermined criteria. In one embodiment, the measurement of the tortuosity of the blood vessel can be considered together with the diameter of the blood vessel and / or the disordered characteristics of the vascular network to determine whether the blood vessel is a feeding vessel. An advantage of this invention is that it enables accurate identification of feeding vessels used for cancer treatment. Additionally, tumor regions can also be accurately identified based on the feeding vessel network.

[0011] According to one embodiment, determining the tumor region within the region of interest based on the feeding vessels includes identifying a region in the medical image based on classified feeding vessels, wherein the identified region in the medical image is the region surrounding the feeding vessels. For example, feeding vessels can be tracked to identify the area where the feeding vessels terminate. For example, the feeding vessels may terminate in the region where the feeding vessels end. Such regions can be labeled and identified as areas surrounding the feeding vessels. Furthermore, the method includes determining pixel intensities associated with the region of interest in the medical image. Pixel intensities can be pixel values ​​associated with pixels in the medical image. The method includes determining the tumor region based on pixel intensities and the region surrounding the feeding vessels. The tumor region has a higher pixel intensity than other regions in the medical image. For example, connectivity component analysis can be performed to accurately identify the tumor region in the medical image. Connectivity component analysis can identify tumor regions based on pixel intensities associated with the medical image. An advantage of this invention is that it enables the precise determination of tumor regions and associated feeding vessels, based on which cancer treatment can be effectively planned for patients.

[0012] According to another embodiment, the trained machine learning model is a Classification and Regression Tree (CART) model. This model is a supervised machine learning model that can classify medical images as normal or suspected to include tumors with feeding vessels. The CART model can perform classification based on the diameter, branching, and / or tortuosity of blood vessels associated with them in the medical image. In one embodiment, the model can use a Gini index to create decision points for performing classification. The Gini index stores the sum of squared probabilities for each class. The Gini index can be represented as:

[0013]

[0014] Where i is the number of categories, and Pi is the probability of each category.

[0015] The advantage is that the model is configured to accurately identify tumor regions and blood vessels in medical images, thereby enabling effective cancer treatment for patients.

[0016] The objective of this invention is achieved through a method for training a machine learning model to determine the presence of a tumor in a patient. The method includes receiving medical images associated with the patient, wherein the medical images include a region of interest associated with the patient. The medical images can be received from a source such as a medical database or a medical imaging device. Additionally, the method includes identifying one or more blood vessels associated with the region of interest. Furthermore, the method includes receiving a machine learning model and determining a set of characteristics associated with the blood vessels using the model. The method also includes having the machine learning model determine, based on the set of characteristics associated with the blood vessels, whether the blood vessel is a feeding vessel associated with the tumor.

[0017] The method further includes receiving tumor data associated with a medical dataset, wherein the tumor data indicates whether a blood vessel is a feeding vessel associated with the tumor. In one embodiment, the tumor data may include medical images that have been labeled to indicate the presence or absence of feeding vessels associated with a tumor in a patient's body. In a further embodiment, the labeled tumor data may be associated with multiple patients who have been historically monitored and treated for cancer. In an alternative embodiment, the tumor data may be data received from a physician / specialist, which may include analysis of medical images associated with the patient, indicating the presence or absence of feeding vessels associated with the patient.

[0018] The method further includes adjusting the machine learning model based on a comparison between the blood supply vessels determined by the model and the tumor data. This comparison can indicate the accuracy of the machine learning model in detecting blood supply vessels. Therefore, if a difference is identified between the determined blood supply vessels and the tumor data in the comparison, the machine learning model can be adjusted. The advantage is that the machine learning model is made more robust, thereby improving the accuracy of the model in determining blood supply vessels. Therefore, it can make the determination of the presence of tumors in a patient effective and accurate.

[0019] The object of the present invention is achieved through a device for determining the presence of a tumor in a patient. The device includes one or more processing units and a medical database connected to the one or more processing units, the medical database including multiple medical images and tumor data associated with the patient. The device also includes a memory coupled to the one or more processing units. The memory includes a tumor determination module configured to perform the method steps described above using at least one trained machine learning model.

[0020] In one aspect, the present invention relates to a computer program product comprising a computer program that can be loaded into a system's storage unit, including program code segments, so that when the computer program is executed in the system, the system performs a method according to one aspect of the present invention.

[0021] In one aspect, the present invention relates to a computer-readable medium on which a program code segment of a computer program is stored, the program code segment being loadable into and / or executable in a system such that when the program code segment is executed in the system, the system performs a method according to one aspect of the present invention.

[0022] Implementing the invention through a computer program product and / or a computer-readable medium has the advantage that software updates can easily adopt existing management systems to function as proposed in the invention.

[0023] A computer program product can be, for example, a computer program or include another element besides a computer program. This other element can be hardware, such as a memory device storing the computer program thereon, a hardware key for using the computer program, etc., and / or software, such as documentation or software keys for using the computer program. Attached Figure Description

[0024] The invention will now be further described with reference to the illustrative embodiments shown in the accompanying drawings, wherein:

[0025] Figure 1 A block diagram of an embodiment of a device for determining the presence of a tumor in a patient is shown.

[0026] Figure 2 A flowchart illustrating a method for determining the presence of a tumor in a patient according to an embodiment of the present invention is shown.

[0027] Figure 3 A flowchart illustrating a method for determining whether a blood vessel is a blood supply vessel associated with a tumor, according to an embodiment of the present invention.

[0028] Figure 4 A flowchart illustrating a method for determining whether a blood vessel is a blood supply vessel associated with a tumor, according to another embodiment of the present invention.

[0029] Figure 5 A flowchart illustrating a method for detecting a tumor region in a region of interest based on a blood supply vessel, according to an embodiment of the present invention, is shown.

[0030] Figure 6 A flowchart illustrating a method for training a machine learning model to determine the presence of a tumor in a patient's body, according to an embodiment of the present invention, is shown. Detailed Implementation

[0031] The embodiments for carrying out the present invention are described in detail below. Various embodiments are described with reference to the accompanying drawings, wherein the same reference numerals are always used to refer to the same elements. In the following description, numerous specific details are set forth for purposes of explanation in order to provide a thorough understanding of one or more embodiments. It will be apparent that these embodiments may also be practiced without these specific details.

[0032] The solution according to the invention is described below with respect to the claimed providing system and the claimed method. Features, advantages, or alternative embodiments described herein may be assigned to other claimed objectives, and vice versa. In other words, the claims for the providing system can be improved using features described or claimed in the context of the method. In this case, the functional features of the method are embodied by the target unit of the providing system.

[0033] Furthermore, the solutions according to the invention are described below for methods and systems for determining the presence of tumors in a patient and for methods and systems for training machine learning models to determine the presence of tumors in a patient. Features, advantages, or alternative embodiments described herein may be assigned to other claimed objectives, and vice versa. In other words, the claims for methods and systems for training machine learning models to determine the presence of tumors in a patient can be modified using features described or claimed in the context of methods and systems for determining the presence of tumors in a patient, and vice versa. Specifically, the trained machine learning model of the methods and systems for determining the presence of tumors in a patient can be adapted by methods and systems for training machine learning models for determining the presence of tumors in a patient. Furthermore, input data may include advantageous features and embodiments for training input data, and vice versa. Furthermore, output data may include advantageous features and embodiments for outputting training data, and vice versa.

[0034] Figure 1 This is a block diagram of device 100, wherein embodiments may be implemented as, for example, device 100 for determining the presence of a tumor in a patient, configured to perform the processes described herein. Figure 1 In the device 100, there are processing unit 101, memory 102, storage unit 103, input unit 104, bus 106, output unit 105 and network interface 107.

[0035] As used herein, processing unit 101 represents any type of computing circuitry, such as, but not limited to, a microprocessor, microcontroller, complex instruction set computing microprocessor, reduced instruction set computing microprocessor, very long instruction word microprocessor, explicit parallel instruction computing microprocessor, graphics processor, digital signal processor, or any other type of processing circuitry. Processing unit 101 may also include embedded controllers, such as general-purpose or programmable logic devices or arrays, application-specific integrated circuits, single-chip computers, etc.

[0036] Memory 102 can be volatile or non-volatile memory. Memory 102 can be coupled for communication with the processing unit 101. The processing unit 101 can execute instructions and / or code stored in memory 102. Various computer-readable storage media can be stored in and accessed from memory 102. Memory 102 can include any suitable element for storing data and machine-readable instructions, such as read-only memory, random access memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, hard disk drive, removable media drive for processing optical discs, digital video discs, magnetic disks, tape cartridges, memory cards, etc. In this embodiment, memory 102 includes a tumor determination module 110 stored in the form of machine-readable instructions on any of the above-described storage media, and can communicate with and be executed by processor 101. When executed by processor 101, tumor determination module 110 causes processor 101 to process medical images to determine the presence of tumors in a patient. Figure 2 , 3 Sections 4, 5, and 6 detail the method steps executed by processor 101 to achieve the above functions.

[0037] Storage unit 103 may be a non-transitory storage medium for storing medical database 112. Medical database 112 is a repository of medical images and tumor data related to one or more patients, maintained by a healthcare provider. Input unit 104 may include input devices capable of receiving input signals such as medical images, such as a keypad, touch-sensitive display, camera (such as a camera that receives gesture-based input), etc. Bus 106 serves as an interconnect between processor 101, memory 102, storage unit 103, input unit 104, output unit 105, and network interface 107.

[0038] Those skilled in the art should understand that Figure 1The hardware described herein can vary for specific implementations. For example, in addition to or instead of the hardware shown, other peripheral devices such as optical disc drives, local area network (LAN) / wide area network (WAN) / wireless (e.g., Wi-Fi) adapters, graphics adapters, disk controllers, and input / output (I / O) adapters may be used. The examples depicted are provided for illustrative purposes only and are not intended to imply any architectural limitations with respect to this disclosure.

[0039] Device 100 according to one embodiment of this disclosure includes an operating system employing a graphical user interface. The operating system allows multiple display windows to be presented simultaneously in the graphical user interface, wherein each display window provides an interface to a different application or a different instance of the same application. A cursor in the graphical user interface can be manipulated by a user using a pointing device. The cursor position can be changed and / or events such as clicking a mouse button can be generated to initiate a desired response.

[0040] With appropriate modifications, one of various commercial operating systems may be used, such as a version of Microsoft Windows™, a product of Microsoft Corporation, located in Redmond, Washington. The operating system described herein is modified or created in accordance with this disclosure.

[0041] The disclosed embodiments provide systems and methods for processing medical images. Specifically, the systems and methods can be used to determine the presence of tumors in a patient's body.

[0042] Figure 2 A flowchart of a method 200 for determining the presence of a tumor in a patient according to an embodiment of the present invention is shown. In step 201, a medical image associated with the patient is received from a medical imaging device. The medical image is a computed tomography angiography image. In an alternative embodiment, the medical image can be obtained from any imaging modality capable of imaging the vascular system associated with the patient. The medical image depicts a region of interest associated with the patient. The region of interest can be part of the patient's body suspected of having a tumor. The region of interest, for example, depicts an organ associated with the patient and the associated vascular system. Method 200 further includes step 202, wherein one or more blood vessels associated with the region of interest are identified. In one embodiment, image segmentation techniques such as region growing algorithms can be used to identify blood vessels. Region growing algorithms identify objects to be segmented based on pixel information present in the medical image. In an alternative embodiment, a topology-preserving thinning algorithm can be used to preserve the topology of the blood vessels. This algorithm is capable of emphasizing the topology of the blood vessels, particularly their length, orientation, and width.

[0043] In step 203, a set of features associated with blood vessels is determined. In one embodiment, this set of features associated with blood vessels includes the diameter of the blood vessel, the branching of the blood vessel, and / or the tortuosity of the blood vessel. A trained machine learning model is used to determine this set of features. This model is a Classification and Regression Tree (CART) model. The model classifies the medical image as normal or suspected of containing a tumor with feeding vessels based on the set of features associated with blood vessels. In step 204, based on the set of features associated with blood vessels, it is determined whether the blood vessel is a feeding vessel associated with a tumor. The CART model includes a root node, multiple child nodes, multiple branches, and leaf nodes, upon which the model makes a final decision. The root node splits into child nodes. The leaf nodes include output variables, upon which the model performs a prediction of the presence of feeding vessels in the medical image. The model uses a metric called the Gini index to create decision points for the classification task performed by the model. Therefore, during the classification task, a weighted sum of the Gini indices associated with the child nodes is calculated. This is performed for all child node splits, and the split with the lowest Gini index is selected as the best split. Calculate the Gini index using the following expression:

[0044]

[0045] Where "i" is the number of categories in the tree, and "Pi" is the probability of each category.

[0046] Furthermore, in step 205, the tumor region is detected within the region of interest based on the blood vessels associated with the tumor. This process can be performed if a blood vessel is identified as a feeding vessel. However, if the model determines in step 204 that no feeding vessel exists, the medical image is labeled as not containing tumor / feeding vessels in step 206. Figure 5 The method steps associated with detecting tumor regions in medical images are described in more detail.

[0047] Figure 3 A flowchart of a method 300 for determining whether a blood vessel is a blood supply vessel associated with a tumor, according to an embodiment of the present invention, is shown. In step 301, the diameter associated with the blood vessel is determined. Changes in the blood vessel diameter are considered an indicator of an abnormal vascular system. The diameter of the blood vessel is calculated using the Euclidean distance between points A and B on the blood vessel. The Euclidean distance is calculated using the following expression:

[0048] Diameter =

[0049] Among them, X A and Y A Let X be the Cartesian coordinates of point A, and X... B and Y BLet B be the Cartesian coordinate of point B in the Euclidean plane.

[0050] In a vascular system associated with normal tissue, the diameter of the parent vessel is larger than the diameter of its branch vessels. Therefore, the vessel diameter decreases along its length. In step 302, a CART model is used to determine whether the vessel diameter decreases as the vessel branches into branch vessels. If the vessel diameter does not decrease, the vessel is classified as a feeding vessel using the CART model. However, if the vessel diameter decreases along its length, in step 304, the vessel is classified as a normal (non-tumor vascular system).

[0051] In an alternative embodiment, in step 304, the radius associated with the blood vessel and its associated branch vessels is determined. In step 305, the cube of the radius associated with the blood vessel and the radius associated with each branch vessel is calculated. In step 306, it is determined whether the sum of the cubes of the radii associated with the branch vessels is equal to the sum of the cubes of the radii associated with the blood vessel. Generally, the branch vascular system of the circulatory system follows Murray's law, that is, the cube of the radius of the parent vessel is equal to the sum of the cubes of the radii of the branch vessels.

[0052]

[0053] Where "r" is the radius of the blood vessel, and r1 to r n It is the radius of the branch vessel.

[0054] Therefore, if the vascular system obeys Murray's law, the blood flow in the vascular system is laminar. However, if it does not obey Murray's law, then the blood flow in the vascular system is turbulent and can be an indication of the presence of feeding vessels in the region of interest. Therefore, if the sum of the cubes of the radii associated with branch vessels is not equal to the cube of the radius associated with the vessel, then in step 304, the vessel can be classified as a feeding vessel associated with a tumor. However, if the sum of the cubes of the radii associated with branch vessels is equal to the cube of the radius associated with the vessel, then in step 307, the vessel is classified as a normal (non-tumor vascular system).

[0055] Figure 4A flowchart of another method 400 for determining whether a blood vessel is a blood supply vessel associated with a tumor, according to an embodiment of the present invention, is shown. In step 401, the length associated with the blood vessel is determined. The length of the blood vessel may be the path length associated with the blood vessel. In step 402, the angle value at each corner point of the blood vessel is determined. The corner point may be each deviation of the blood vessel from the straight path followed by the blood vessel. This is defined as the tortuosity of the blood vessel. In one embodiment, in step 403, the tortuosity of the blood vessel is calculated using a distance metric method or an angle and measure (SOAM) method. According to the angle and measure method, the following expression can be used to calculate the tortuosity:

[0056]

[0057] Where α is the angular deviation associated with the blood vessel in degrees, and L1 is the path length associated with the blood vessel.

[0058] In step 404, the measured tortuosity of the blood vessel is compared with predetermined criteria associated with the vessel. For example, if there are more than two consecutive bends with an angle greater than or equal to 180°, the vessel is classified as tortuous at step 405. However, if the tortuosity of the vessel is equal to or less than the predetermined criteria, the vessel is classified as normal (non-tumor vascular system) at step 406. In one embodiment, the tortuosity of a vessel may not be the only characteristic on which the classification of a vessel as a feeding vessel is based. The tortuosity of the vessel, along with other characteristics such as the diameter associated with the vessel and its branches, may also be considered. This avoids false positive results.

[0059] Figure 5 A flowchart of a method 500 for detecting a tumor region in a region of interest based on a blood supply vessel according to an embodiment of the present invention is shown. In step 501, regions in a medical image are identified based on classified blood supply vessels. Specifically, regions in the region of interest are identified based on the location of the blood supply vessels exiting the image. Regions in the medical image where the blood supply vessels gradually taper can be identified as areas where the blood supply vessels exit. Additionally, in step 502, pixel intensities associated with the region of interest are identified. Pixel intensity is a pixel value associated with a pixel in the medical image. In step 503, the tumor region is identified based on the pixel intensity and the region surrounding the blood supply vessel. The region surrounding the blood supply vessel is used as input to determine the tumor region. The pixel intensity associated with the region surrounding the blood supply vessel is unique and different relative to the rest of the region of interest. Therefore, a connectivity component labeling mechanism is used to accurately determine the tumor region.

[0060] Figure 6A flowchart of a method 600 for training a machine learning model to determine the presence of a tumor in a patient, according to an embodiment of the present invention, is shown. In step 601, a medical image associated with the patient is received. The medical image includes a region of interest that may be suspected of containing a tumor. In step 602, one or more blood vessels associated with the region of interest are identified. In step 603, a processing unit 101 receives a machine learning model. In step 604, a set of characteristics associated with the blood vessels is determined by the machine learning model. Furthermore, in step 605, the machine learning model determines whether the blood vessel is a blood supply vessel associated with a tumor based on this set of characteristics associated with the blood vessel.

[0061] In step 606, tumor data associated with medical images is received, wherein the tumor data indicates whether a blood vessel is a feeding vessel associated with a tumor. In one embodiment, the tumor data includes medical images associated with a patient that have been pre-labeled to indicate the presence or absence of a feeding vessel. Alternatively, the tumor data may also include medical images from multiple historical examinations of patients with a feeding vessel / tumor. In another embodiment, the tumor data may include medical images associated with multiple body parts associated with a patient.

[0062] In step 607, a comparison is made to determine whether the output of the machine learning model matches the tumor data. If the model's output does not match the tumor data, it is determined that the machine learning model needs to be adjusted. Therefore, if it is determined that the model needs to be adjusted, the model is adjusted in step 608. The machine learning model is adjusted so that the accuracy of the model in predicting the presence of blood vessels in medical images is improved. In an alternative embodiment, if the model's output does not match the tumor data, a notification may be generated for the user of the method to determine whether the model should be adjusted according to the tumor data. The model can then be adjusted based on the user's input.

[0063] The advantages of this invention lie in its ability to effectively detect tumor regions in medical images. Furthermore, this invention enables the identification of the correct blood supply vessels for embolization. Therefore, healthy tissues associated with the patient are not damaged due to cancer treatment. Additionally, this invention reduces the need for manual identification of tumor regions in medical images. This invention also enables targeted cancer therapy, based on which tumor vascular networks can be precisely embolized. Moreover, this invention allows for the embolization of multiple tumors in a single medical procedure.

[0064] The embodiments provided above are for illustrative purposes only and should not be construed as limiting the invention disclosed herein. Although the invention has been described with reference to various embodiments, it should be understood that the terms used herein are descriptive and illustrative, not restrictive. Furthermore, while the invention has been described herein with reference to specific apparatuses, materials, and embodiments, it is not intended to be limited to the details disclosed herein; rather, the invention extends to all functionally equivalent structures, methods, and uses, such as those within the scope of the appended claims. Various modifications can be made to it by those skilled in the art who benefit from the teachings of this specification, and changes can be made in various aspects without departing from the scope and spirit of the invention.

Claims

1. A method (200) for determining the presence of a tumor in a patient by a processing unit (101), the method (200) comprising: The processing unit (101) receives a medical image associated with the patient, wherein the medical image includes a region of interest associated with the patient; The processing unit (101) identifies one or more blood vessels associated with the region of interest in the medical image; The processing unit (101) uses a trained machine learning model to determine a set of characteristics associated with the blood vessel; The trained machine learning model determines whether the blood vessel is a blood supply vessel associated with the tumor based on the set of characteristics associated with the blood vessel; and If the blood vessel is a blood supply vessel associated with the tumor, then the processing unit (101) detects the tumor region in the region of interest based on the blood supply vessel. Detecting the tumor region within the region of interest based on the blood supply vessels includes: Regions in the medical image are identified based on classified blood supply vessels, wherein the identified regions in the medical image are regions surrounding the blood supply vessels; Determine the pixel intensity associated with the region of interest in a medical image; and The tumor region is determined based on the pixel intensity and the area surrounding the blood supply vessel, wherein the tumor region has a higher pixel intensity than other regions in the medical image.

2. The method (200) according to claim 1, wherein, The set of characteristics associated with the blood vessel includes one or more of the following: the diameter associated with the blood vessel, the branching of the blood vessel, and the tortuosity of the blood vessel.

3. The method (200) according to claim 1, wherein, Determining whether the blood vessel is a blood supply vessel associated with the tumor includes: Use a trained machine learning model to measure the diameter associated with blood vessels; A trained machine learning model is used to determine whether the diameter of the blood vessel decreases as it branches into branching vessels; and Using the trained machine learning model, if the diameter of the blood vessel does not decrease as the vessel branches into branching vessels, the blood vessel is classified as a blood supply vessel associated with the tumor.

4. The method (200) according to claim 1, wherein, Determining whether the blood vessel is a blood supply vessel associated with the tumor includes: A trained machine learning model is used to determine the radius associated with the blood vessel; A trained machine learning model is used to determine the radius associated with the branch vessels originating from the said vessel. A trained machine learning model is used to calculate the radius associated with the blood vessel and the cube of the radius associated with each of the branch blood vessels. The trained machine learning model is used to determine whether the sum of the cubes of the radii associated with the branch vessels is equal to the sum of the cubes of the radii associated with the vessels; and Using the trained machine learning model, if the sum of the cubes of the radii associated with the branch vessel is not equal to the cube of the radius associated with the vessel, then the vessel is classified as a supply vessel.

5. The method (200) according to claim 1, wherein, Determining whether the blood vessel is a blood supply vessel associated with the tumor includes: The tortuosity of the blood vessel is determined using the trained machine learning model, wherein the tortuosity of the blood vessel is determined based on the angular deviation of the blood vessel from the straight path associated with the blood vessel; Using the trained machine learning model, the tortuosity of the blood vessel is compared with a predetermined standard associated with the blood vessel; and Using the trained machine learning model, if the tortuosity of the blood vessel does not match the predetermined criteria, the blood vessel is classified as a blood supply vessel.

6. The method (200) according to claim 1, wherein, Identifying the one or more blood vessels associated with the region of interest in the medical image includes using a segmentation algorithm to segment the one or more blood vessels.

7. The method (200) of claim 6 further includes using a topology-preserving refinement algorithm to preserve the topology of the blood vessels.

8. The method (200) according to any one of claims 1 to 7, wherein, The trained machine learning model is a classification and regression tree (CART) model.

9. A method (600) for training a machine learning model (700) to determine the presence of a tumor in a patient, the method (600) comprising: The processing unit (101) receives a medical image associated with the patient, wherein the medical image includes a region of interest associated with the patient; The processing unit (101) identifies one or more blood vessels associated with the region of interest; The processing unit (101) receives the machine learning model; The machine learning model is used to determine a set of characteristics associated with the blood vessels; The machine learning model determines whether the blood vessel is a blood supply vessel associated with the tumor based on the set of characteristics associated with the blood vessel; The processing unit (101) receives tumor data related to the medical image, wherein the tumor data indicates whether the blood vessel is a blood supply vessel associated with the tumor; and The machine learning model is adjusted based on the comparison between the blood supply vessels determined by the machine learning model and the tumor data. Detecting the tumor region within the region of interest based on the blood supply vessels includes: Regions in the medical image are identified based on classified blood supply vessels, wherein the identified regions in the medical image are regions surrounding the blood supply vessels; Determine the pixel intensity associated with the region of interest in a medical image; and The tumor region is determined based on the pixel intensity and the area surrounding the blood supply vessel, wherein the tumor region has a higher pixel intensity than other regions in the medical image.

10. The method (600) according to claim 9, wherein, The tumor data includes medical images labeled to indicate the presence or absence of blood vessels that are associated with tumors in the patient's body.

11. An apparatus (100) for determining the presence of a tumor in a patient, the apparatus comprising: One or more processing units (101); A medical database (112), coupled to the one or more processing units (101), the medical database (112) including multiple medical images and tumor data associated with the patient; and A memory (102) coupled to the one or more processing units (101), the memory including a tumor determination module (110) configured to perform the method steps according to any one of claims 1 to 8 using at least one trained machine learning model.

12. A computer program product comprising machine-readable instructions that, when executed by one or more processing units (101), cause the one or more processing units (101) to perform the method steps according to any one of claims 1 to 8.

13. A computer-readable medium on which a program code segment of a computer program is stored, the program code segment being loadable into and / or executable in a system (100) such that when the program code segment is executed in the system (100), the system (100) performs the method steps according to any one of claims 1 to 8.

Images