Artificial neural networks for breast cancer detection using micro antennas

The American Cancer Society estimates that over 170,000 women will be diagnosed with invasive breast cancer in 2007 in the United States alone. The survival rate of these women is inversely proportional to the stage of the cancer at the time of detection. This makes detection vital in the early stages, when treatment is readily available and effective. Current detection methods include X-ray mammography, magnetic resonance imaging (MRI), and ultrasound. Unfortunately mammography and ultrasound screenings yield high false-negative and false-positive rates. MRI, on the other hand, can be cost-prohibitive and is not widely available.; Microwave imaging has emerged as a promising solution to overcoming these problems. This technology has the potential to offer a low-cost, safe, and effective diagnostic tool. The strength of microwave imaging relies on the significant contrast between the dielectric properties of healthy breast tissue and cancer cells at microwave frequencies. This contrast means that the tumor scatters the impinging electromagnetic waves considerably. Taking advantage of this significant scattering allows for the detection of small tumors inside dense breast tissue.; The design of micromachined antennas operating at microwave frequencies in conjunction with the artificial neural networks detection tool was investigated as a preprocessing detection system. The antennas were designed to be adequately small to allow an array to be positioned around the breast to focus the beam on the tumor. The artificial neural network used the scattered fields from the tumor and provided a statistical indication of whether a tumor is present. The accuracy of the network depended mainly on the training procedure of the network as was investigated in this work.; The effect of using the forward or backward scatter was determined by changing the receiver location around the breast. The forward scattering mechanism represented waves traveling through the breast tissue and then received at the antenna. The backscatter mechanism represented receiving the waves at the same location of the transmitting antennas. Additionally, the effect of tumor orientation in regards to polarization of the incident wave was investigated using a designed planar dual-linear polarized broadband antenna.; The ability of this system to provide real time results makes it an ideal candidate as a preprocessing system prior to the time intensive image processing techniques. The proof of concept is presented using synthetic data simulated using a 3D breast model.