SYSTEM AND METHOD FOR ESTIMATING MOTION OF TARGET INSIDE TISSUE BASED ON SURFACE DEFORMATION OF SOFT TISSUE

Abstract

Provided is a system and method for estimating the motion of a target inside a tissue based on surface deformation of the soft tissue. The system consists of an acquisition unit, a reference input unit, two surface extraction units, a target position extraction unit, a feature calculation unit, and a target motion estimation unit. The method includes: the acquisition unit acquires an image I.sub.i of the soft tissue; the surface extraction unit extracts a surface f.sub.i of the soft tissue from I.sub.i; the reference input unit acquires a reference image I.sub.ref of the soft tissue; the surface extraction unit and the target position extraction unit respectively extract a reference surface f.sub.ref of the soft tissue and a target reference position t.sub.ref from I.sub.ref, the feature calculation unit calculates deformation feature Ψ.sub.i of f.sub.i relative to f.sub.ref, the target motion estimation unit estimates the target displacement based on Ψ.sub.i and t.sub.ref.

Claims

1. A system for estimating the motion of a target inside a soft tissue based on surface deformation of the issue, wherein the system is composed of an acquisition unit, a reference input unit, two surface extraction units, a target position extraction unit, a feature calculation unit, and a target motion estimation unit.

2. The system for estimating the motion of a target inside a soft tissue based on surface deformation of the issue according to claim 1, wherein the acquisition unit is used for obtaining an actually captured image of the soft tissue; the reference input unit is used for inputting a reference image of the soft tissue; the surface extraction units are used for extracting soft tissue surfaces from the actually captured image and the reference image of the soft tissue; the target position extraction unit is used for extracting a reference position of the target from the reference image of the soft tissue; the feature calculation unit is used for calculating a deformation feature of the surface of the soft tissue in the actually captured image relative to the surface in the reference image; and the target motion estimation unit is used for calculating and outputting a motion displacement estimation of the target based on the deformation feature and the reference position of the target.

3. A method for estimating the motion of a target inside a soft tissue based on surface deformation of the tissue, wherein the method is achieved through the following steps: S1: a reference input unit capturing a soft tissue image as a reference image I.sub.ref by using medical imaging equipment; S2: a target position extraction unit identifying, calculating and outputting a reference position (denoted as t.sub.ref) of the target from I.sub.ref, S3: a surface extraction unit extracting a soft tissue surface (denoted as f.sub.ref) from I.sub.ref, wherein a method of extracting the soft tissue surface comprises an automatic edge recognition algorithm by applying adaptive threshold segmentation or a fully convolutional neural network model; S4: an acquisition unit capturing an actually captured image (denoted as I.sub.i) of the soft tissue by using the medical imaging equipment; S5: the surface extraction unit extracting a soft tissue surface (denoted as f.sub.i) from I.sub.i; S6: a feature calculation unit calculating and outputting a deformation feature (denoted as Ψ.sub.i) of f.sub.i relative to f.sub.ref, a method of calculating and outputting the deformation feature of f.sub.i relative to f.sub.ref is: inputting f.sub.i and f.sub.ref into Ñ neural network models M={M.sub.j|j=1, . . . , Ñ}, and output results of the Ñ models together constitute the deformation feature Ψ.sub.i={Ψ.sub.i,j|j=1, . . . , Ñ}; and S7: inputting Ψ.sub.i and t.sub.ref into a target motion estimation unit, and a target motion estimation model (m) in the unit calculating and outputting a motion estimation (denoted as {tilde over (t)}.sub.i) of the target based on Ψ.sub.i and t.sub.ref.

4. The method according to claim 3, wherein the medical imaging equipment in the steps S1 and S4 comprises: computed tomography (CT), cone-beam CT (CBCT), and ultrasonography, and in the step S1, before the image is captured, one or more markers are implanted into a target area, and then the soft tissue image is captured by using the medical imaging equipment.

5. The method according to claim 3, wherein a design method of the Ñ neural network models M={M.sub.j|j=1, . . . , Ñ} in the feature calculation unit in the step S6 is: (1) modeling: establishing a fully convolutional neural network model (denoted as FCN), wherein an input layer of FCN is 2 pieces of surface data, hidden layers comprise {l.sub.1, . . . , l.sub.N−1}, and an output layer is l.sub.N; and a deformation vector field (denoted as ϕ.sub.k) of the 2 pieces of surface data is output by l.sub.N; (2) collecting training data: using the medical imaging equipment to collect multiple groups of soft tissue images I.sub.k, extracting changing surfaces f.sub.k (k=1, 2, . . . , n) of the soft tissue from I.sub.k, taking any one of the surfaces {f.sub.k|k=1, . . . , n} as a reference surface (denoted as f.sub.ref), and taking the rest surfaces as changing surfaces to form training sample pairs {(f.sub.k, f.sub.ref)|k=1, . . . ,n and k≠ref} together; (3) training and optimizing FCN: inputting {(f.sub.k, f.sub.ref)|k=1, . . . , n and k≠ref} into FCN, performing iterative optimization on the model by using unsupervised learning, setting a loss function as the difference between f.sub.k and a generated surface. The generated surface (ϕ.sub.kf.sub.ref) is achieved by applying ϕ.sub.k on f.sub.ref, and when the loss function is optimal, terminating the optimization; wherein an index for measuring the difference between ϕ.sub.kf.sub.ref and f.sub.k is a sum of minimum distances from all points in ϕ.sub.k f.sub.ref to f.sub.k; (4) constructing M={M.sub.j|j=1, . . . , Ñ}: in the layer structure {l.sub.1, . . . , l.sub.N} of the trained FCN, and taking Ñ layers ({I.sub.k.sub.j|j=1,2, . . . , Ñ, k.sub.j ∈[1, N], Ñ≤N}) as the output layers of Ñ M.sub.js, respectively, wherein the input layer is consistent with that of FCN, both of which are two pieces of surface data, and the hidden layers of each M.sub.j is formed by sequentially ordering {l.sub.1, . . . , l.sub.k.sub.j−1}.

6. The method according to claim 5, wherein the medical imaging equipment in the step (2) comprises four-dimension (4D) CT, 4D CBCT, and three-dimensional ultrasonography, and the soft tissue surface is directly delineated or identified by using automatic threshold segmentation or the neural network from the collected images.

7. The method according to claim 5, wherein in the step (4), if k.sub.j=1, then M.sub.j is only composed of an input layer and an output layer (l.sub.1), and the {l.sub.k.sub.j|j=1,2, . . . , Ñ, k.sub.j ∈[1,N], Ñ≤N} is preferably a convolutional layer in the trained FCN.

8. The method according to claim 3, wherein a design method of the target motion estimation model (m) in the step S7 is: (1) data collection: using medical imaging equipment to capture multiple groups of soft tissue images I.sub.p, and identifying and calculating a soft tissue surface f.sub.p and a target position t.sub.p from I.sub.p, where p=1, 2, . . . ,n′; (2) calculating a deformation feature Ψ.sub.pto form training data: randomly taking a group from the collected {(f.sub.p, t.sub.p)|p=1,2, . . . , n′} to serve as reference samples denoted as (f.sub.ref, t.sub.ref), taking the rest as change samples {(f.sub.p,t.sub.p)|p=1,2, . . . ,n′ and p≠ref}, and inputting {(f.sub.p, f.sub.ref)|p=1,2, . . . , n′ and p≈ref} into the feature calculation unit to generate the deformation feature gi.sub.p , wherein the deformation feature constitutes training data {(Ψ.sub.p, t.sub.p,t.sub.ref)|p=1,2, . . . ,n′ and p≠ref} together with t.sub.p and t.sub.ref, and (3) fitting the target motion estimation model (m): the input data of the model is t.sub.ref and Ψ.sub.p, the output is the displacement estimation (denoted as {circumflex over (t)}.sub.p) of the target; through iterative optimization, a difference between {circumflex over (t)}.sub.p output by the model and a true value t.sub.p thereof is minimized, the difference between {circumflex over (t)}.sub.p output by the model and the true value t.sub.p thereof is measured by an Euclidean distance between {circumflex over (t)}.sub.p and t.sub.p.

9. The method according to claim 8, wherein the medical imaging equipment in the step (1) comprises 4D CT and three-dimensional ultrasonography, the soft tissue images acquired by 4D CT are 10 groups of 3D CT scans which are uniformly sampled within one breathing cycle; each 3D CT image contains a 3D image of the soft tissue surface and the target, or, by implanting one or more markers into the target area of the soft tissue, surface changes of the soft tissue and the displacement of the markers are captured by using the 3D ultrasonography, and the soft tissue surface and the target position are directly delineated, or identified by using automatic threshold segmentation or the neural network from the collected images.

10. The method according to claim 8, wherein the target motion estimation model (m) in the step (3) comprises: a neural network model, a fully convolutional neural network model, a linear model, and a support vector machine model.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0038] FIG. 1 is a schematic diagram of a system for estimating the motion of a target inside a tissue based on surface deformation of the soft tissue.

[0039] FIG. 2 is a flow diagram of a method for estimating the motion of a target inside a tissue based on surface deformation of the soft tissue.

[0040] FIG. 3 is a schematic diagram of a design method of a plurality of neural network models in a feature calculation unit.

[0041] FIG. 4 is a schematic diagram of a method of a target motion estimation model.

DESCRIPTION OF EMBODIMENTS

[0042] The present application will be further explained in conjunction with the drawings and examples.

Example 1

[0043] A system for estimating the motion of a target inside a tissue based on surface deformation of the soft tissue, as shown in FIG. 1 includes: an acquisition unit, a reference input unit, two surface extraction units, a target position extraction unit, a feature calculation unit, and a target motion estimation unit, wherein the acquisition unit is used for obtaining an actually captured image I.sub.i of the soft tissue; the reference input unit is used for inputting a reference image I.sub.ref of the soft tissue; the surface extraction units are used for extracting soft tissue surfaces from the actually captured image I.sub.i and the reference image I.sub.ref of the soft tissue, and the extracted surfaces are respectively expressed as f.sub.i and f.sub.ref, the target position extraction unit is used for extracting a reference position t.sub.ref of the target from the reference image I.sub.ref of the soft tissue; the feature calculation unit is used for calculating a deformation feature (denoted as Ψ.sub.i) of the surface f.sub.i of the soft tissue in the actually captured image relative to the surface f.sub.refin the reference image; and the target motion estimation unit is used for calculating and outputting motion displacement estimation {circumflex over (t)}.sub.i of the target based on the deformation feature Ψ.sub.i and the reference position t.sub.ref of the target.

Example 2

[0044] A method for estimating the motion of a target inside a tissue based on surface deformation of the soft tissue, as shown in FIG. 2, is achieved through the following steps:

[0045] S1: a reference input unit uses medical imaging equipment to capture a soft tissue image as a reference image I.sub.ref.

[0046] In at least one embodiment of the present application, the “medical imaging equipment” includes: CT, CBCT and ultrasonography.

[0047] In at least one embodiment of the present application, before the image is captured, one or more markers are implanted into a target area, and then the soft tissue image is captured by using the medical imaging equipment.

[0048] S2: a target position extraction unit identifies, calculates and outputs a reference position (denoted as t.sub.ref) of the target from I.sub.ref.

[0049] S3: a surface extraction unit extracts a soft tissue surface (denoted as f.sub.ref) from I.sub.ref.

[0050] In at least one embodiment of the present application, the method of “extracting the soft tissue surface” includes: an automatic edge recognition algorithm based on adaptive threshold segmentation or a fully convolutional neural network model.

[0051] S4: an acquisition unit uses medical imaging equipment to capture an actually captured image (denoted as of the soft tissue.

[0052] In at least one embodiment of the present application, the “medical imaging equipment” includes: CT, CBCT, and ultrasonography.

[0053] S5: the surface extraction unit extracts a soft tissue surface (denoted as f.sub.i) from I.sub.i.

[0054] S6: a feature calculation unit calculates and outputs a deformation feature (denoted as Ψ.sub.i) of f.sub.i relative to f.sub.ref.

[0055] In at least one embodiment of the present application, the method of “calculating and outputting the deformation feature of f.sub.i relative to f.sub.ref” is: inputting f.sub.i and f.sub.ref into Ñ neural network models M={M.sub.j|j=1, . . . , Ñ}, and the output results of the Ñ models together constitute the deformation feature Ψ.sub.i={Ψ.sub.i,j|j=1, . . . , Ñ}.

[0056] S7: Ψ.sub.i and t.sub.ref are input into a target motion estimation unit, and a target motion estimation model (m) in the unit calculates and outputs motion estimation (denoted as {circumflex over (t)}.sub.i) of the target based on Ψ.sub.i and t.sub.ref.

[0057] The design method of the “Ñ neural network models M={M.sub.j|j=1, . . . , Ñ}” in the “feature calculation unit” in the step S6 is:

[0058] (1) Modeling: establishing a fully convolutional neural network model (denoted as FCN), wherein an input layer of FCN is 2 pieces of surface data, the hidden layers include {l.sub.1, . . . l.sub.N−1} and an output layer is l.sub.N. A deformation vector field ϕ.sub.k of the 2 pieces of surface data is output by I.sub.N.

[0059] (2) Collecting training data: using medical imaging equipment to collect multiple groups of soft tissue images I.sub.k, extracting changing surfaces f.sub.k(k=1, 2, . . . , n) of the soft tissue from I.sub.k, taking any one of the surfaces { f.sub.k|k=1, . . . ,n} as a reference surface (denoted as f.sub.ref), and taking the rest surfaces as changing surfaces to form training sample pairs {(f.sub.k, f.sub.ref)|k=1, . . . ,n and k≈ref} together.

[0060] In at least one embodiment of the present application, the “medical imaging equipment” includes 4D CT, 4D CBCT, and three-dimensional ultrasonography. The soft tissue surface is directly delineated or identified by using automatic threshold segmentation or the neural network from the collected images.

[0061] (3) Training and optimizing FCN: in the embodiment shown in FIG. 3, inputting {(f.sub.k, f.sub.ref)|k=1, . . . ,n and k≈ref} into FCN, performing iterative optimization on the model by using unsupervised learning, setting a loss function as the difference between f.sub.k and a generated surface. The generated surface (ϕ.sub.kf.sub.ref) is achieved by applying ϕ.sub.k on f.sub.ref, and when the loss function is optimal, terminating the optimization.

[0062] In at least one embodiment of the present application, the index for measuring the difference between ϕ.sub.kf.sub.ref and f.sub.k is the sum of minimum distances from all points in ϕ.sub.kf.sub.ref to f.sub.k.

[0063] (4) Constructing M={M.sub.j|j=1, . . . , Ñ}: in the embodiment shown in FIG. 3, Ñ=N, that is, all layers in the trained FCN construct N neural network models M={M.sub.j|j=1, . . . , N} The input layer of each M.sub.j is consistent with that of FCN, both of which are 2 pieces of surface data (f.sub.k, f.sub.ref), the output layer is l.sub.j, and the hidden layers are formed by sequentially ordering {l.sub.1, . . . , l.sub.j−1}.

[0064] If j=1, then M.sub.1 is only composed of an input layer and an output layer (l.sub.i).

[0065] The design method of the “target motion estimation model (m)” in the step S7 is:

[0066] (1) data collection: using medical imaging equipment to capture multiple groups of soft tissue images I.sub.p, and identifying and calculating a soft tissue surface f.sub.p and a target position t.sub.p from I.sub.p, wherein p=1,2, . . . , n′.

[0067] In at least one embodiment of the present application, the “medical imaging equipment” includes 4D CT and three-dimensional ultrasonography. The soft tissue images acquired by 4D CT are 10 groups of 3D CT scans which are uniformly sampled within one breating cycle. Each 3D CT image contains a 3D image of the soft tissue surface and the target. Or, by implanting one or more markers into the target area of the soft tissue, surface changes of the soft tissue and the displacement of the markers are captured by using the 3D ultrasonography. The soft tissue surface and the target position are directly delineated or identified by using automatic threshold segmentation or the neural network from the collected images.

[0068] (2) Calculating a deformation feature Ψ.sub.p to form training data: randomly taking a group from the collected {(f.sub.p, t.sub.p)|p=1,2, . . . n′} as reference samples, marking the same as (f.sub.ref, t.sub.ref), taking the rest as change samples {(f.sub.p,t.sub.p)|p=1,2, . . . ,n′ and p≠ref}, and inputting {(f.sub.p,f.sub.ref)|p=1,2, . . . , n′ and p≠ref} into the feature calculation unit to generate the deformation feature Ψ.sub.p. The deformation feature constitutes training data {(Ψ.sub.p, t.sub.p, t.sub.ref)|p=1,2, . . . , n′ and p≠ref} together with t.sub.p and t.sub.ref.

[0069] (3) Fitting the target motion estimation model (m): in the embodiment shown in FIG. 4, the input data of the model is t.sub.ref and Ψ.sub.p, the output is the displacement estimation (denoted as {circumflex over (t)}.sub.p) of the target. Through iterative optimization, the difference between {circumflex over (t)}.sub.p output by the model and a true value t.sub.p thereof is minimized.

[0070] In at least one embodiment of the present application, the “difference between {circumflex over (t)}.sub.p output by the model and the true value t.sub.p thereof” is measured by the Euclidean distance between {circumflex over (t)}.sub.p and t.sub.p.

[0071] In at least one embodiment of the present application, the “target motion estimation model (m)” includes: a neural network model, a fully convolutional neural network model, a linear model, and a support vector machine model.

[0072] The core of the technical solution of the present application for realizing target displacement estimation is fitting, that is, correlating the target displacement with the deformation feature of the soft tissue surface using a mathematical expression. Since the speed at which the mathematical expression gives an estimated value is only related to the own calculation speed of a computer, the present application can ensure the timeliness of estimation. The source of the motion of the target in the soft tissue is the deformation of the soft tissue, and the most obvious deformation is a change on its surface, therefore, in the solution provided by the present application, the feature describing the surface deformation of the soft tissue is used as an association object, and the deformation feature is extracted by reconstructing the output of the hidden layers of the neural network, the function of the neural network is to realize the matching of the actually measured surface and the reference surface, and in order to avoid the influence of the gray scale of the image, the matched object adopts a surface contour extracted from the image.