Ultrasonic detection of brachytherapy seeds based on singular spectrum analysis
The present invention provides a method for detecting and locating brachytherapy seeds using Singular Spectrum Analysis (SSA). The method exploits the presence of seed-specific repetitive echo signals in detecting seeds and determining their location. Radio frequency (rf) echo signals received from a scanned area of interest are converted to zero-mean envelope-detected signals and then processed using SSA to produce a spectral-power line (a P line). The P line indicates the relative likelihood of a seed being present. A colored P-mode image derived from a set of P lines is superimposed on a conventional ultrasound B-mode image to depict seed location.
1. A method for detecting brachytherapy seeds and determining their location in tissue, comprising:
imaging a brachytherapy seed with an ultrasonic transducer to obtain radio-frequency (rf) echo signals;
digitizing the if echo signals and converting them to zero-mean envelope-detected signals, s;
performing a singular spectrum analysis on the envelope-detected signals in a computing unit to obtain a periodic reconstructed signal s r , wherein the analyzed envelope-detected signals are a time signal, s, composed of N samples, and that a sliding window of M samples is used to estimate a correlation matrix A where M N, wherein A is determined by:
A
i
,
j
=
1
N
-
M
∑
k
=
1
N
-
M
s
(
k
)
s
(
k
+
i
-
j
)
;
determining a P value indicative of the likelihood of seed presence based on the spectral power of s r ;
determining a maximum amplitude of s r ; and
using the maximum amplitude of s r to locate a brachytherapy seed in tissue.
2. The method in accordance with claim 1 further including diagonalization of A to obtain selected eigenvalues and ordering the eigenvalues from the largest to the smallest.
3. The method in accordance with claim 2 wherein certain eigenvalue pairs are selected from the ordered eigenvalues with the eigenvalue pairs being selected based upon predetermined heuristic criteria.
4. The method in accordance with claim 3 wherein a new time signal, s r , is constructed from the selected eigenvalue pairs.
5. The method in accordance with claim 4 wherein s r is computed according to the following relationship:
s
r
(
i
+
j
)
=
∑
k
∈
Γ
a
i
k
E
j
k
.
6. The method in accordance with claim 5 where the Fourier transform S r (ƒ) of s r is computed.
7. The method in accordance with claim 6 wherein a spectral power, P, is computed based on S r (ƒ) where the value of P is a number representing the relative likelihood of a seed being present in a particular location in the tissue.