11 May 2026 in Measurement Science and Technology
Area of breast cancer tissues has been clustered by newly-proposed electrical impedance tomography (EIT) implemented with Poisson and Gaussian mixture models (EIT-PGMM) for breast cancer identification. EIT-PGMM comprises two distinct methods for comparison: EIT-PMM (Poisson Mixture Model) and EIT-GMM (Gaussian Mixture Model). EIT-PGMM consists of two sections which are 1) reconstruction section by EIT to reconstruct conductivity distribution σ ∈R
(
F
×
N
)
( f : frequency 1≤ f ≤ F , n : area resolution 1≤ n ≤ N ) from impedance distribution Z ∈C
(
F
×
M
)
( f : frequency 1≤ f ≤ F , m : measurement number 1≤ m ≤ M ) on injection current i between electrodes on measurement boundary and measurement voltages v , and 2) clustering section by PGMM to cluster breast cancer area B with cluster assignment (one-hot vector) s ∈R
(
N
×
K
)
( k : cluster number 1≤ k ≤ K ) to sharpen the border between cancerous and normal breast tissues. The s is probabilistically determined through posterior p ( s|σ,Θ ) ( Θ : parameters of Poisson and Gaussian mixture models) assigns each spatial resolution to a specific cluster, which highest mean conductivity distribution 〈 σ
k
〉 among the K clusters is identified as the breast cancer region B . The EIT-PGMM was applied to the mastectomized breast tissues from 5 patients placed inside hemispherical sensor of EIT under the condition of injection current frequencies f ranging from 1 [kHz] to 3 [MHz]. Qualitatively, breast cancer region B remains similar across different values of K and between EIT-PMM and EIT-GMM. In order to quantitatively evaluate B , the results of EIT-PMM and EIT-GMM were compared with the diameter D measured clinically and area A of breast cancer segmented from the coronal plane of CT scan. EIT-GMM provides a more accurate estimation of breast cancer region B with accuracy of 90.2% in diameter D K GMM . and 79.3% in area A K GMM .