Cancer, Instrumentation, Computer Applications-Detection, diagnosis, Experimental, Breast
G. Tiberi1, L. Sani1, N. Ghavami2, M. Paoli1, A. Vispa1, G. Raspa1, E. Vannini1, A. Saracini1, M. Duranti1; 1Perugia/IT, 2London/UK
Fig 3 shows the microwave images obtained for subject 11R (a) and subject 08R (b); as pointed out before,
to allow intra-subject comparison,
the two images have been normalized to unitary average of the intensity.
Axes are given in meter.
Intensity is given in arbitrary unit,
with a scale from 0 to 2.
Microwave images are homogeneity maps of tissues' dielectric properties (both dielectric constant and conductivity),
and are given here as 2D images in the azimuthal plane i.e.
A certain level of in-homogeneity can be seen in both (a) and (b); however,
in-homogeneity is more pronounced.
Figs 3c and 3d show the mammography images for subject 11R and 08R,
mediolateral oblique views.
Even if a co-registration between Fig 3b and Fig 3d is not possible,
both figures detect an inclusion near the surface of the breast.
According to the radiologist study review,
8 healthy breasts and 12 non-healthy breasts underwent microwave imaging.
For each microwave imaging exam,
the parameter Max/Avg is given in Table 2; specifically,
the left part of Table 2 refers to healthy breasts,
while the right part of Table 2 refers to non-healthy breasts.
For the healthy breasts,
mean and standard deviation of Max/Avg are 1.64 and 0.04; for the non-healthy breasts,
mean and standard deviation of Max/Avg are 1.77 and 0.09.
The threshold T is equal to 1.68; T is then used to classify the microwave images of the non-healthy breasts.
By comparing the output of the radiologist study with the classification of microwaves images performed using the threshold T,
it is possible to note that 11/12 of non-healthy breasts have an altered microwave image; this turns out into a sensitivity of 91%.