Efficient Cancer Detection Using Multiple Neural Networks

Abstract

The inspection of live excised tissue specimens to ascertain malignancy is a challenging task in dermatopathology and generally in histopathology. We introduce a portable desktop prototype device that provides highly accurate neural network classification of malignant and benign tissue. The handheld device collects 47 impedance data samples from 1 Hz to 32 MHz via tetrapolar blackened platinum electrodes. The data analysis was implemented with six different backpropagation neural networks (BNN). A data set consisting of 180 malignant and 180 benign breast tissue data files in an approved IRB study at the Aurora Medical Center, Milwaukee, WI, USA, were utilized as a neural network input. The BNN structure consisted of a multi-tiered consensus approach autonomously selecting four of six neural networks to determine a malignant or benign classification. The BNN analysis was then compared with the histology results with consistent sensitivity of 100% and a specificity of 100%. This implementation successfully relied solely on statistical variation between the benign and malignant impedance data and intricate neural network configuration. This device and BNN implementation provides a novel approach that could be a valuable tool to augment current medical practice assessment of the health of breast, squamous, and basal cell carcinoma and other excised tissue without requisite tissue specimen expertise. It has the potential to provide clinical management personnel with a fast non-invasive accurate assessment of biopsied or sectioned excised tissue in various clinical settings.

Keywords: Biological neural networks; bioimpedance; biomedical engineering; cancer detection; feedforward neural networks.

FIGURE 1.

Vm (mV) for various tissue types provided courtesy of .

FIGURE 2.

Nernst calculation for K⁺ depicting depolarized Vm and the resulting Fcole, fc, probabilities.

FIGURE 3.

Smooth Frequency Response windowed for 256x256 Mel filter bank interpolation for 4 Tissue types displaying sub-band power variability.

FIGURE 4.

Direct Cosine Transfrom of inverted cepstrals windowed for 256x256 Mel filter bank interpolation for 4 Tissue types displaying sub-band power variability.

FIGURE 5.

47-sample dataset averages of 180 CA and 180 BE impedance files from 10 Hz – 5 MHz displaying potential conflicts in data differentiation.

FIGURE 6.

33-sample dataset averages of Z” for CA and BE from 10 Hz – 4 MHz displaying the Cole frequency, fc and the frequency range of interest.

FIGURE 7.

Kolmogorov-Smirnov nonparametric test of the comparison of the probability distribution for the 47 sample and 33 sample datasets.

FIGURE 8.

One way ANOVA for the 33 sample CA and BE datasets with a p-value of 0.00012 indicating significant difference in the datasets.

FIGURE 9.

Kruskal-Wallis nonparametric test displaying a rank comparison of the medians between the 33 samples BE and CA datasets.

FIGURE 10.

33-sample dataset averages of the imaginary CA and BE impedance files from 10 Hz – 5 MHz displaying Fcole (fc) and the frequency range of interest.

FIGURE 11.

Linear regression fits with correlation coefficients for initial Levenberg-Marquadt neural network.

FIGURE 12.

Neural network architectures depicting the number of hidden layers and neurons in each hidden layer for the six systems implemented.

FIGURE 13.

Simulation of CGF and RP neural network output response to correct classifications for BE and CA.

FIGURE 14.

Network performance based on mean squared error.

FIGURE 15.

Network performance based on mean squared error.