Systems and methods for image segmentation

Abstract

Machine learning systems can be used to determine whether a segmentation of a medical image is valid, in the sense that it is a segmentation that could be made by a human user and does not contain any impossible combinations of pixel values, and the method is augmented by user input to avoid impossible combinations.

Description

[Technical Field]

[0001] The present invention relates to a method, computer program product, and computer system for machine-assisted segmentation of medical images, such as CT images. [Background technology]

[0002] Before performing a medical procedure, such as treating a patient with radiation therapy, it is standard practice to take medical images, e.g., CT images, of the relevant parts of the patient and create a treatment plan based on these images. For treatment planning purposes, various parts of the patient, such as the tumor and / or various organs, need to be delineated in the medical images. This delineation, known as segmentation, is recognized as a difficult task to perform both manually and automatically. Proper segmentation of the tumor, and sometimes of specific organs, can be difficult to determine with confidence. Even experienced oncologists can obtain completely different results when segmenting the same image.

[0003] Methods for automatic segmentation are known in the art. While such methods can produce good results, they are often limited to predicting one segmentation per patient. In some cases, especially when the input data is constructed by medical professionals with very different preferences, the model is implicitly forced to somehow average these preferences. This can sometimes produce unrealistic results.

[0004] A widely used method for automatically generating one-per-patient segmentation involves training a deep neural network of the U-net type. Formally, the setup is as follows: a dataset of n manually drawn medical images (x i ,y i )~p X,Y where i=1,2,…,n and x i is the i'th medical image, and y iis the relevant depiction done manually by a medical professional. The segmentation problem is then solved by searching for a classifier f ∈ F that minimizes the empirical loss with respect to a specific loss function L among a predefined set of U-net-based neural networks.

number

[0005] Instead of generating a single proposed segmentation per input image, we can consider a probability distribution over the space of segmentations conditional on the input image. This topic has been thoroughly explored in the classical machine learning paradigm, i.e., shallow models. One prominent example is conditional random fields, which are described in detail, for example, in Sutton, Charles, and Andrew McCallum. “An introduction to conditional random fields.” Foundations and Trends® in Machine Learning 4.4 (2012): 267-373. Conditional random fields, or CRFs, are used to obtain a probability distribution over the space of possible segmentations, which can then be used to sample proposed segmentations conditional on the input image. Knowing the probability distribution over the entire space allows users to manually modify parts of the segmentation and then perform maximum likelihood estimation conditional on the modifications. These properties make CRFs interesting candidate models for image segmentation. However, because they are shallow models, they are unable to learn complex relationships within the data. This makes them not viable for state-of-the-art medical image segmentation tasks.

[0006] Several deep learning methods have been proposed to enhance classical probabilistic segmentation methods with the versatility of deep models. For example, Simon AA Kohl et al.: A Probabilistic U-Net for Segmentation of Ambiguous Images, 32nd Conference on Neural Information Processing Systems (NeurIPS 2018) evaluated a probabilistic method based on a combination of a U-net and a variational autoencoder. In such methods, users sample proposed segmentations from a distribution trained on the input image, and in some cases compute maximum likelihood estimates. However, these prior art methods cannot be used for interactive segmentation. Summary of the Invention

[0007] It is an object of the present invention to provide improved image segmentation suitable for use in medical planning procedures such as radiation therapy planning.

[0008] The present invention provides a computer-implemented method for improving the determination of a segmentation of a medical image comprising a set of pixels or voxels, the method comprising: providing the medical image to a neural network-based machine learning system that has been trained with a training dataset that includes a manually created segmentation to determine whether the segmentation is similar to a segmentation in the training dataset, and obtaining an initial segmentation of the medical image from the machine learning system; manually adjusting the initial segmentation; updating the initial segmentation with a machine learning system to generate an updated segmentation; The present invention relates to a method, comprising:

[0009] Thus, the method of the present invention allows for the selection of a segmentation through a combination of automatic and manual input. The automatic selection is based on a probability determined as the probability that the entire segmentation is feasible, taking into account the user's input. This ensures that the resulting segmentation will be deemed acceptable by a qualified medical professional, while reducing the risk of errors caused by misinterpretation. The method is applicable to various types of images, including CT and MR images. The user is typically a medical professional, e.g., an oncologist.

[0010] Thus, according to the present invention, automatic segmentation is improved by using a machine learning system in combination with the possibility of manual adjustment based on the experience of a user, who may be a medical professional with experience in the relevant medical field, for example an oncologist, to avoid cases where segmentation is clearly impossible, for example because the outline has a shape that is unlikely to occur naturally.

[0011] One embodiment of the method approximates a non-normalized conditional probability distribution with a neural network.

number

[0012] In a preferred embodiment, the sets are vectors: Set X may be a vector of real values ​​and set Y may be a binary segmentation vector.

[0013] The adjustment and updating steps can be performed just once, or can be repeated using the updated segmentation as the initial segmentation until it is found to be good enough, in the sense that it is what a human operator could make, i.e., it does not contain interpretations that an untrained machine might make but that a human would consider invalid, such as sharp depressions in the structure, which in this context means that it has the same characteristics as the images in the training dataset.

[0014] In a preferred embodiment, the optimization is performed using an argmax function based on the pixels of the image, the segmentation values, and a function f.

[0015] The method of the present invention involves computing, with a deep neural network, a possibly non-normalized joint probability density function p x,y or the conditional density function p(y|x), which is designed to allow for interactive segmentation as well as to allow the model to capture complex relationships in the data.

[0016] The machine learning system used in accordance with the present invention is based on neural networks. More specifically, an almost-ubiquitously differential neural network is used to approximate a possibly non-normalized p(y|x) or p(x,y) that can be evaluated at all points. This network can be generated in various ways, for example by some variations of GANs or autoencoders, both of which are well known to those skilled in the art. Different types of networks have different advantages and disadvantages. In standard technical terms, the term "almost-ubiquitously differential" network means that the function may be non-differentiable at some points, but the total volume of these points is zero.

[0017] In some embodiments, the machine learning system is a generative adversarial network (GAN), and optimization is performed using a discriminator that is trained to recognize segmentations that are similar to segmentations in a training dataset of manually created or similar segmentations.

[0018] In some other embodiments, the machine learning system is an autoencoder, and the optimization is performed using the encoder, where the optimization is performed using a score f(x,y)=p Z It is based on (e(x,y)).

[0019] GANs are known in the literature to be able to generate images with a very high level of detail (in this case segmentation), but sometimes have problems with mode collapse where the full distribution properties are not captured satisfactorily. On the other hand, variations of autoencoders, such as variational autoencoders, are known to generate images with less detail but with better distribution properties. Depending on the dataset and the problem, the choice of method may differ.

[0020] Segmentation of medical images is used in certain medical procedures, in particular for planning purposes. The present invention therefore also relates to a medical method in which a segmentation generated according to such a method is used as a basis for planning a medical procedure. As mentioned above, segmentation is useful in planning various types of medical procedures, including radiation therapy and some surgical procedures.

[0021] Correspondingly, the present invention also provides a method for training a machine learning system capable of generating a score for generating a segmentation of at least one structure in a medical image, the method comprising: inputting training datasets into a machine learning system, the training datasets including image datasets, each image dataset including medical images along with one or more human-generated and / or one or more machine-generated segmentations of at least one structure; and training the machine learning system to recognize segmentations that are similar to the segmentations in the training dataset; The present invention relates to a method comprising:

[0022] The present invention also provides, when executed on a computer, the following: a. Segmentation value y indicating whether each pixel belongs to the structure i obtaining an initial segmentation of structures in the image in the form of a set Y of b. Segmentation value y i receiving manual input to adjust the initial segmentation by changing at least one of the following: c. updating the initial segmentation by optimizing the set of segmentation values ​​Y based on the given pixel value x and the adjusted set of segmentation values ​​Y'; d. Optionally, repeating steps b and c using the updated segmentation as the initial segmentation until the updated segmentation is found to be sufficiently good; for a given pixel value x i The present invention relates to a computer program product for improving the generation of a segmentation of a medical image comprising a set X of pixels having

[0023] The present invention also relates to a computer system for generating a segmentation of a medical image, the computer system comprising a processor, a program memory configured to hold at least one computer program to be executed on the processor, and a data memory, the program memory holding the computer program product as described above. The present invention also relates to a method of training a machine learning system and to a machine learning system trained according to said method.

[0024] The invention will now be explained in more detail, by way of example, with reference to the accompanying drawings, in which: FIG. [Brief explanation of the drawings]

[0025] [Figure 1] 1 is a general flowchart of a method according to one embodiment of the present invention. [Figure 2] 10 is an example of a possible segmentation with manual adjustments input by a human. [Figure 3] 10 is an example of a possible segmentation with manual adjustments input by a human. [Figure 4] 10 shows the probability p that each of several possible segmentations is a plausible segmentation. [Figure 5] 1 illustrates various portions of an embodiment of a method for using a GAN. [Figure 6] 1 illustrates various portions of an embodiment of a method for using a GAN. [Figure 7] FIG. 1 illustrates one embodiment of a method for using an autoencoder. [Figure 8] FIG. 1 illustrates a computer system on which the method of the present invention may be implemented. DETAILED DESCRIPTION OF THE INVENTION

[0026] Figure 1 is an overall flowchart of a method according to one embodiment of the present invention. The method uses a machine learning system that can generate a probability indicating how likely a physician is to generate the segmentation based on an input set including a CT image and a possible segmentation of the CT image. Before the system can be used to generate such a value, it must be trained based on several CT images, each having one or several segmentations associated with it. Various machine learning systems and how to train them are known to those skilled in the art. Some examples are described in some detail below.

[0027] In step S51, a segmentation of the structure in the CT image is obtained in any suitable manner. The segmentation can be completely arbitrary. In step S52, a physician reviews the segmentation and manually adjusts some of the structure. Some possible reasons for and types of adjustments are described below. In step S53, a discriminator is used to update the segmentation to identify one that is considered plausible in terms of what the physician is likely to arrive at. This is performed as an optimization, which is described below. Step S54 is a user decision step to determine if the updated segmentation is good enough. If yes, the segmentation is output and can be used, for example, for treatment planning. If no, the procedure returns to step S52 or S53 for further adjustments. Depending on the procedure in which the segmentation is used, the user is typically a medical professional.

[0028] In the following, the medical image to be segmented is denoted by x, and the proposed segmentation, y and f(x,y), is denoted by the approximation of the unnormalized likelihood, C·p(y|x), where C is a positive unknown constant. The update in step S53 can be performed as an optimization as follows: mathematically, f is a high-order function with 2N parameters, where the first N parameters, given by x, are related to the image and are real numbers, and the remaining N parameters, given by N, are binary and related to the segmentation. If there are no constraints on the number of voxels considered in the image, N is typically 100. 3 The order is:

[0029] In what follows, the part of the segmentation y that has been entered by the user will be denoted by y1, and the part of the segmentation that corresponds to the automatically generated segmentation will be denoted by y2. For notational convenience, and without loss of generality, in what follows we will only consider the case where there is one object to depict against a background. The optimization problem to be solved can be expressed as finding y2 that produces the maximum value of f for a given combination of x and y1, i.e.

number

[0030] This optimization can be approached in many ways, using methods known in the art. A simple solution would be simulated annealing, which is suitable for discrete problems without simple structures. This method is not feasible if all parameters of a single regular segmentation are optimized. However, the size of the problem can be reduced by excluding all voxels that are clearly part of the segmented structure or that are clearly not part of the segmented structure. In other words, the size can be reduced by excluding voxels with little or no uncertainty, resulting in a problem containing only voxels that are uncertain whether they belong to a structure or not. This can be defined by user input or by a method for identifying uncertain voxels. For example, the network can be a regular U-net, which assigns an independent probability to each voxel and filters out voxels whose probabilities associated with all structures are lower than a certain threshold probability.

[0031] In other words, the set X represents the pixels or voxels of the medical image, which may be a vector of real values. The sets Y and Y' are preferably vectors, such as binary segmentation vectors, where 1 indicates the foreground and zero indicates the background.

[0032] An alternative method that can handle larger problems is by successive approximations of the basic problem, i.e., y∈[0,1] N is the binary vector space {0,1} N The reason for this approximation is that we allow the continuum to take values ​​not only within the unit cube, but also within the unit cube itself. This continuum can then be interpreted as either the "fraction of voxels that belong to a structure" or the "probability of a voxel being within that structure." Such an approximation is sometimes made naturally to capture that voxels on the boundary are only likely to be part of a particular structure to a certain degree. The requirement for this approximation is that we can use the space {0,1} of binary vectors of dimension N to N as well as the unit cube y∈[0,1] of dimension N NThe function f must be differentiable almost everywhere with respect to y. M If is defined by y2 = sigmoid(z), and the sigmoid function operates element-wise, then gradient-based optimization methods such as gradient descent can solve the optimization problem

number

number

number

[0033] Assuming the machine learning system is a neural network, the backpropagation algorithm can be used as an efficient way to implement gradient descent in a neural network. Such a neural network can be achieved in a variety of ways. Examples of how to use a discriminator in conjunction with a GAN, or an encoder in conjunction with certain types of autoencoders, are provided below.

[0034] Figure 2 shows the initial segmentation y of the organ as a solid line 21. In this example, the user must select the region y outside the segmentation, delineated by the dashed line 22. 1 also belong to the structure. Bearing in mind that the images are 3D and may contain many slices, it would be a daunting task to re-depict the entire structure accordingly. As an example, a CT scan of the spine may contain around 200 slices.

[0035] According to an embodiment of the present invention, the user can mark one or more pixels in the segmentation as "clearly part of the structure," "not part of the structure," or "on the edge of the structure." In this example, the user modifies an initial segmentation marked with a solid line by marking several points along what are considered to be the edges of the structure. Alternatively, the region y 1 One or more points in the outline can be marked as belonging to the structure. A discriminator then selects the most probable segmentation from among the possible segmentations that match the user's markings. This means that the discriminator selects segmentations that look plausible, for example, in that they contain contiguous structures incorporating adjacent pixels with similar characteristics. The adjustment should be made at an appropriate location within the structure so that the discriminator can adjust the relevant part of the outline. For example, if only one adjustment point is entered, it should typically be in the center of the part of the outline to be modified.

[0036] Figure 3 shows another situation, illustrating the importance of considering the segmentation probability as a whole. In a) we consider the first possible segmentation y of the structure. init is shown as a solid line. In this example, the user evaluating the segmentation should select the smaller region y corr Everything within belongs to the structure, while the area y between the dashed line and the solid line on the right not It is clear that this region y not The voxels within are likely to have uniform characteristics. Therefore, this entire region should be included or excluded from the segmentation of the structure. In this example, the physician believes it should be excluded. It can be a lot of work to delineate parts of the structure that need to be manually modified in 3D, as in Figure 2.

[0037] As in the previous example, the user can mark one or more pixels in the segmentation as being "part of the structure", "not part of the structure", or "on the edge of the structure". b) shows three such points p1, p2, p3 set by the physician to identify the boundary of the structure. For ease of calculation, the points are selected to be approximately evenly distributed over the portion of the outline to be adjusted. As can be seen in b), the structure that results from adding these three points to the segmentation is smaller than the initial segmentation y. init There are three sharp indentations in x, which have a shape that is highly unlikely for any structure and would have a probability of zero. Therefore, according to this embodiment of the invention, an approximation of the non-normalized probability distribution C·p(y|x) is used to identify the most likely segmentation that satisfies the requirements added by the user. The resulting segmentation will likely be as shown in c), i.e., all areas outside the dashed lines are considered not to be part of the structure.

[0038] Figure 4 is a one-dimensional schematic diagram illustrating the unnormalized likelihood approximation f(x,y) as a function of y, with x fixed to one specific input image. Two segmentations y1 and y2 are marked on the horizontal axis, and as can be seen, the first segmentation f(x,y1) scores much lower than f(x,y2). This means that, according to this model, it is much more likely that the doctor manually depicted the patient as y2 rather than y1.

[0039] Using the values ​​shown in Figure 4, one of the possible segmentations can be selected as a starting point to arrive at a more plausible segmentation via a gradient method. The change in probability between the selected segmentation and the segmentations immediately to its right and left is determined, and the one that results in the greatest increase in probability is selected. The procedure then continues in the same direction until no more probability increases. In this way, a segmentation with a high probability should be arrived at, but as can be seen, it is not necessarily the segmentation with the highest possible probability. Therefore, starting from the first segmentation y1, segmentations to its right should be repeatedly selected until there is no longer a positive slope on the curve and the probability no longer increases. The segmentation selected by this procedure can be evaluated by the user, who can make further adjustments to the selected segmentation to provide a new starting point. The procedure can then be repeated, starting from the new starting point. In principle, any segmentation that matches the user-specified modifications can be used as a starting point. For example, the user-specified variables could be fixed and the remaining variables could be randomly drawn as independent U[0,1] variables. Another approach is to use the output of a classical U-net model to populate the free variables, i.e., again, freeze the variables related to the user input and use the U-net output as the starting point for the variables on which the optimization is performed.

[0040] The machine learning system used in accordance with the present invention can be any type of system capable of generating a score that reflects the probability that a particular segmentation was produced by a physician.

[0041] The machine learning system can be any suitable type of neural network trained in any suitable manner. For example, it can be part of a generative adversarial network (GAN), which is known per se. Such a network was first described in Goodfellow, Ian, et al. "Generative adversarial nets." Advances in neural information processing systems. 2014. A GAN uses two competing neural networks: a generator and a discriminator. During training, the discriminator is sent several training sets, each containing a medical image and one or more segmentations of one or more structures within the medical image. Some segmentations are genuine segmentations of medical images, usually created manually, while others are generated by the generator. The generator strives to create pictures, or segmentations in the context of the present invention, that are perceived as genuine, while the discriminator strives to distinguish between genuine segmentations and machine-generated segmentations. In this way, the generator and discriminator trigger each other to improve.

[0042] The GAN generator function is g w (Medical image, noise) -> Tumor and / or organ segmentation The noise is generated randomly to result in variations in the segmentation when the generator is run several times.

[0043] The GAN generator is g w (x)=y where g is a function of one or more weights w, where the weights are adjusted so that for each CT image x, a binary image y is generated in which each pixel is identified as belonging to a structure or not belonging to a structure. The process of training a GAN generator is known to those skilled in the art.

[0044] In this way, the generator is trained to generate segmentations that are increasingly similar to real segmentations made by humans, while the discriminator is trained to distinguish between real segmentations and machine-made segmentations with increasing accuracy. Formally, a discriminator is a classifier that approximates a probability distribution

number

number

[0045] Figure 5 is a flowchart of a possible method, known per se, for automatically generating segmentations using GANs. In a first step S61, a CT image and some general noise are input to a generator. In step S62, the generator creates a segmentation based on the input data from step S61. In step S63, the segmentation is output. This procedure can be repeated as many times as necessary if a set of segmentations is desired.

[0046] FIG. 6 is a flowchart of one embodiment of the method of the present invention, implemented by utilizing a discriminator. In step S71, a CT image and a segmentation of the relevant portion of the CT image are input to the discriminator. The segmentation can be obtained in any suitable manner, for example, manually or by the process of FIG. 1 or FIG. 5. In step S72, the discriminator generates a score S73 indicating how likely it is that the segmentation was made by a physician, given the CT image. If necessary, in step S74, the segmentation can be manually adjusted and sent to the discriminator again. This is typically done if the likelihood is too low, i.e., if the segmentation is too far from what a physician could have made, or if the physician determines that manual adjustment is desirable. It is also possible to perform manual adjustments such as in step S74 before step S72.

[0047] The discriminator is the problem argmax y The method is configured to generate an optimized segmentation by solving f(x,y), where f is a discriminator. The physician reviews the output from the discriminator and, if necessary, performs an adjustment step S74 and returns to step S72 to generate a new optimized segmentation. That is, if a correction is provided, y is split into y1 and y2, where y1 is the provided adjustment. Then, argmax y2 f(x,[y1,y2]). This is repeated until the result is satisfactory to the user.

[0048] Alternatively, the machine learning model can be a version of an autoencoder. Such networks, and how they can be configured, are known per se. One way to configure them can be described as follows with reference to Figure 7: An autoencoder is a network that stores data in a latent space approximately p ZThe distribution must be normalized to appear as a distribution, which is illustrated schematically by the dashed curve in the coordinate system on the right. Formally, given a dataset of n manually delineated medical images (x i ,y i )~p X,Y Consider again, where p X,Y An example of L is shown in the left-hand coordinate system. Let the neural network encoder be denoted by e∈E and the neural network decoder by d∈D, where E and D are two predefined sets of neural networks. In Figure 7, the encoder is illustrated by an arrow denoted e(x,y) pointing from the left-hand coordinate system to the right-hand coordinate system, and the decoder is illustrated by an arrow denoted d(x,y) pointing in the opposite direction. R Let be some reconstruction error in both x and y, and L Z Let the empirical distribution of data points in the latent space be the target distribution p Z This corresponds to the average of the Dirac measure, which takes the values ​​of several samples from the distribution depicted by the solid curve p(e(x,y)) in the right-hand coordinate system. The network then calculates

number

number

[0049] Furthermore, f Z If is chosen to be almost everywhere differentiable, e.g., normally distributed, then this optimization problem above for updating the segmentation given f can be solved by backpropagation.

[0050] For example, the network can be a fully convolutional encoder and decoder, with all activation functions being ReLU activations, and the output of the decoder related to segmentation being passed through a softmax activation as the last layer. The loss function can be defined such that the reconstruction of x is given by the minimum squared error, and the reconstruction of y is given by the mean cross-entropy error. Finally, the regularizer loss L Z is given by the Wasserstein loss. Since the regularizer term depends on every data point in the dataset, we cannot use mini-batch stochastic gradient descent in the standard way to train this model. To address this, we use the pair (x j ,y j ) mini-batch regularizer term

number

[0051] Figure 8 is a schematic diagram of a computer system in which the planning method of the present invention may be implemented. Computer 51 comprises a processor 53, a data memory 54, and a program memory 55. User input means 58 are also present, preferably in the form of a keyboard, mouse, joystick, voice recognition means, or any other available user input means.

[0052] The data memory 54 contains data such as images and possible segmentations, and typically one or more objective functions. The data in the data memory may be generated by the computer 51, entered by a user input means, or received from another storage means in any manner known in the art.

[0053] It will be appreciated that the data memory 54 is shown only diagrammatically, and there may be several data memory units, each holding one or more different types of data, for example one data memory for a set of values ​​and one data memory for an objective function.

[0054] The program memory 55 holds a computer program configured to control the processor to carry out the method of the present invention. It will be understood that not all method steps are necessarily performed by the computer 51.

Claims

1. Approximating non-normalized conditional probability distributions with neural networks [Equation 1] 1. A computer-implemented method for improving the determination of a segmentation of a medical image comprising a set X of pixels or voxels having a given pixel value x to a corresponding set Y of segmentation values ​​y using a machine learning system that: a. providing medical images to a neural network-based machine learning system that has been trained with a training dataset that includes manually created segmentations to determine whether the segmentations are similar to segmentations in the training dataset; b. Obtaining an initial segmentation of the medical image from the machine learning system in the form of a set Y of segmentation values ​​y indicating whether each pixel belongs to a structure; c. Manually adjust the initial segmentation and change at least one of the segmentation values ​​y to y １ By changing the segmentation value y′=[y １ , y ２ ], generating a conditioned set Y′ of y ２ indicates the portion of y that is not manually adjusted, and the adjustment includes marking one or more pixels in the segmentation as either "part of the structure," "not part of the structure," or "on the edge of the structure"; d. Machine learning systems can solve optimization problems [Equation 2] updating the initial segmentation by optimizing the set of segmentation values ​​Y based on the given pixel value x and the adjusted set of segmentation values ​​Y′ by solving Including, The machine learning system is a generative adversarial network (GAN) or an autoencoder. method.

2. The method of claim 1 , further comprising repeating steps c and d using the updated segmentation as the initial segmentation until the updated segmentation is found to be sufficiently good.

3. The optimization problem [Equation 3] The method of claim 1 or 2, wherein is solved by simulated annealing.

4. The optimization problem [Equation 4] 3. The method of claim 1, wherein is solved by successive approximations of the elementary problem.

5. 5. The method of any one of claims 1 to 4, wherein the machine learning system is a generative adversarial network (GAN) and the optimization is performed using a discriminator that is trained to recognize segmentations that are similar to segmentations in a training dataset of manually created or similar segmentations.

6. The method of any one of claims 1 to 3, wherein the machine learning system is an autoencoder and the optimization is performed using the encoder.

7. 1. A computer program for improving generation of a segmentation of a medical image comprising a set X of pixels having a given pixel value x using a neural network-based machine learning system, wherein the machine learning system is a generative adversarial network (GAN) or an autoencoder; When the computer program is executed on a computer, the computer is caused to: Obtaining an initial segmentation of the structures in the image in the form of a set Y of segmentation values ​​y indicating whether each pixel belongs to the structure; b. Receive manual input to adjust the initial segmentation, adjusting at least one of the segmentation values ​​y to y １ By changing the segmentation value y′=[y １ , y ２ ], generating a conditioned set Y′ of y ２ indicates the portion of y that is not manually adjusted, said manual input comprising marking one or more pixels in the segmentation as either "part of the structure", "not part of the structure", and "on the edge of the structure"; c. Machine learning systems can solve optimization problems [Equation 5] updating the initial segmentation by optimizing the set of segmentation values ​​Y based on the given pixel value x and the adjusted set of segmentation values ​​Y′ by solving Optionally, repeating steps b and c using the updated segmentation as the initial segmentation until the updated segmentation is found to be sufficiently good; A computer program that executes

8. 10. A computer system for generating a segmentation of a medical image, comprising: a processor; a program memory configured to hold at least one computer program executed by the processor; and a data memory, wherein the program memory holds the computer program of claim 7.

Images