US 2026/0263160 A1 · ABLATION AUTOMATED NAVIGATION METHOD AND SYSTEM

Abstract

An ablation automated navigation method includes: setting hard constraint conditions and soft constraint conditions for an ablation path; initializing multiple preliminary ablation paths that meet the hard constraint conditions; calculating a risk index of each of the preliminary ablation paths; sorting each of the preliminary ablation paths in ascending order of the risk index, and selecting a first preset number of preliminary ablation paths with top rankings as initial solutions of a seagull optimization algorithm; optimizing the ablation path through the seagull optimization algorithm to determine multiple candidate ablation paths; sorting each of the candidate ablation paths in ascending order of fitness, and selecting a second preset number of candidate ablation paths with top rankings for organ damage assessment; determining an optimal ablation path from the candidate ablation paths; performing ablation navigation according to a needle insertion position, angle and depth determined by the optimal ablation path.

An ablation automated navigation system, comprising:

a processor;
a memory;
where computer-readable instructions are stored on the memory, and when the computer-readable instructions are executed by the processor to achieve a following ablation automated navigation method:
S1:
setting hard constraint conditions and soft constraint conditions for an ablation path;
S2:
based on the hard constraint conditions, initializing multiple preliminary ablation paths that meet the hard constraint conditions;
S3:
based on the soft constraint conditions, calculating a risk index of each of the preliminary ablation paths;
S4:
sorting each of the preliminary ablation paths in ascending order of the risk index, and selecting a first preset number of preliminary ablation paths with top rankings as initial solutions of a seagull optimization algorithm;
S5:
optimizing the ablation path through the seagull optimization algorithm to determine multiple candidate ablation paths;
S6:
sorting each of the candidate ablation paths in ascending order of fitness, and selecting a second preset number of candidate ablation paths with top rankings for organ damage assessment;
S7:
determining an optimal ablation path from the candidate ablation paths according to an organ damage assessment result;
S8:
performing ablation navigation according to a needle insertion position, angle and depth determined by the optimal ablation path;
a fitness function of the seagull optimization algorithm is specifically:
F
⁡
(
l
)
=
λ
d
⁢
d
l
+
λ
a
⁢
a
l
-
λ
h
⁢
h
l
-
λ
h
⁢
θ
l
;
wherein, F( ) represents the fitness function, l represents the ablation path, dl represents a distance between the ablation path l and a risk structure, λd represents a weight coefficient of a risk structure term, al represents a puncture angle of the ablation path l, λa represents a weight coefficient of a puncture angle term, hl represents a puncture depth of the ablation path l, λh represents a weight coefficient of a puncture depth term, θl represents a number of CT layers through which the ablation path l passes, and λθ represents a weight coefficient of a term of the number of CT layers passed through;
wherein, S5 specifically comprises:
S501:
sorting each preliminary ablation path in ascending order of the risk index, selecting the first preset number of preliminary ablation paths with top rankings as initial solutions of the seagull optimization algorithm, initializing seagull individuals, wherein each seagull individual represents a set of feasible model parameters, each seagull individual is composed of multiple dimensional components, and each component represents a model parameter;
S502:
in a global search phase, avoiding collision and move towards the optimal individual:
D
i
t
=
❘
"\[LeftBracketingBar]"
C
i
t
+
M
i
t
❘
"\[RightBracketingBar]"
C
i
t
=
A
×
P
i
t
M
i
t
=
B
×
(
P
b
⁢
e
⁢
s
⁢
t
t
-
P
i
t
)
;
wherein
D
i
t
represents a position of an i-th seagull individual after the global search phase in a t-th iteration,
C
i
t
represents a position of the i-th seagull individual after anti-collision processing in the t-th iteration, A represents the control factor,
P
i
t
represents a position of the i-th seagull individual in the t-th iteration,
M
i
t
represents a displacement of the i-th seagull individual moving towards an optimal individual in the t-th iteration, B represents a search balance factor, and
P
b
⁢
e
⁢
s
⁢
t
t
represents a position of an optimal individual in the t-th iteration;
A
=
[
1
-
lg
⁡
(
9
⁢
t
T
)
]
×
f
c
;
wherein t represents a current number of iterations, T represents a maximum number of iterations, and fc represents a linear descent frequency;
B
=
2
⁢
A
2
⁢
r
1
;
wherein r1 represents a random number between 0 and 1;
S503:
in a local search phase, generating a random number r2, and according to a random number r2, selecting in parallel between a spiral search strategy and an encirclement strategy to perform displacement in a spiral motion manner:
{
E
i
t
=
D
i
t
⁢
x
⁢
y
⁢
z
+
P
b
⁢
e
⁢
s
⁢
t
t
r
2
≤
0
.
5
E
i
t
=
(
D
i
t
+
P
b
⁢
e
⁢
s
⁢
t
t
)
⁢
x
⁢
y
⁢
z
r
2
>
0
.
5
x
=
r
⁢
sin
⁢
η
x
=
r
⁢
cos
⁢
η
z
=
r
⁢
η
r
=
u
⁢
e
η
⁢
v
;
wherein
E
i
t
represents a position of the i-th seagull individual after spiral motion in the t-th iteration, x represents a spiral flight coefficient in a x direction, y represents a spiral flight coefficient in a y direction, z represents a spiral flight coefficient in a z direction, r represents a radius of a spiral flight trajectory, η represents a random number between 0 and 2π, u and v represent spiral constants, and e represents a natural constant;
S504:
performing mutation operation on each seagull individual:
V
i
t
=
E
i
t
+
γ
⁡
(
P
b
⁢
e
⁢
s
⁢
t
t
-
P
r
)
;
wherein
V
i
t
represents a position of the i-th seagull individual after mutation in the t-th iteration, Pr represents a random individual, and γ represents a adaptive scale factor;
γ
=
γ
max
-
(
γ
max
-
γ
min
)
⁢
sin
[
π
2
⁢
(
t
T
)
2
]
;
wherein γmax represents a maximum scale factor, γmin represents a minimum scale factor, and sin represents a sine function;
S505:
judging whether a fitness value of a mutated position is greater than that of a position before mutation;
if yes, replacing the position before mutation with the mutated position;
otherwise, keeping the position before mutation unchanged;
S506:
updating the fitness value of each seagull individual and the global optimal individual;
S507:
judging whether a current number of iterations reaches a maximum number of iterations;
if yes, outputting a model parameter set represented by the seagull individual with a highest current fitness;
otherwise, returning to continue iteration;
wherein S8 specifically comprises:
S801:
pasting multiple markers on a surface of a patient's abdominal skin;
S802:
based on multiple markers, performing preliminary registration between a three-dimensional model constructed based on preoperative images and the patient's point cloud;
S803:
on a basis of preliminary registration, adopting an improved non-rigid ICP algorithm to perform forward point matching by minimizing an Euclidean distance;
performing backward point matching through global search;
S804:
setting a registration cost function based on a forward point matching result and a backward point matching result:
J
=
∑
v
i
Q
′
∈
Q
′
(
ω
i
⁢

v
i
Q
′
-
M
i
⁢
v
→
(
v
i
Q
′
)

2
+
ω
j
⁢

v
i
Q
′
-
M
j
⁢
v
←
(
v
i
Q
′
)

2
)
+
α
⁢
∑
v
i
P
,
v
j
P
∈
P

M
i
-
M
j

2
;
wherein J represents a registration cost function, Q′ represents data points in a deformed preoperative three-dimensional model, ωi represents a registration result parameter of the i-th data point in a deformed preoperative three-dimensional model, ωi is 1 when there is a corresponding data point for the i-th data point in the deformed preoperative three-dimensional model, otherwise 0, Mi represents a affine transformation matrix for transforming the i-th data point in the deformed preoperative three-dimensional model, ω1 represents a registration result parameter of a j-th data point in the patient's point cloud, ωj is 1 when there is a corresponding data point for the j-th data point in the patient's point cloud, otherwise 0, Mj represents an affine transformation matrix for transforming the j-th data point in the patient's point cloud, and a represents a regularization coefficient;
S805:
performing iterative optimization with a goal of reducing a registration cost function to accurately register a three-dimensional model constructed based on preoperative images with the patient's point cloud;
S806:
mapping the optimal ablation path to the patient's point cloud;
S807:
performing ablation navigation according to a needle insertion position, angle and depth determined after mapping.

416 more elements

Ablation automated navigation method and system

Life of this patent

Ablation automated navigation method and system

Life of this patent

Abstract

Claims1