Intelligent Microwave System for Breast Cancer Detection

近年来，微波已成为一种潜在的乳腺癌检测技术，可避免与X射线相关的不适、风险和成本以及MRI的过高成本。早期的基于微波的检测模式是基于介电特性的成像，其目标是将高对比度乳房内含物与癌组织相关联。 这些技术需要使用在超宽频率范围内工作的多个天线。然而，这些技术的挑战在于，典型的女性乳房是高度异质的，并且包含复杂的腺体脂肪和纤维结缔组织网络。此外，人体组织是高度分散的（频率相关的），因此使得宽带天线的使用极其困难。 由于乳腺组织的介电特性变化丰富，我们仍然相信微波成像具有很强的作用。为了解决早期基于微波的方法的两个基本挑战，我们建议开发近场天线探头。探头将被设计为在非常窄的频率范围内工作，同时保持高增益，与传统的近场探头形成鲜明对比。窄的频率范围将使乳腺组织的频率依赖性不相关。探头的近场方面，加上亚GHz频率，允许低功率非电离微波能量在乳房内的显著渗透。因此，异常的存在将产生差异化的共振轮廓，因为异常不是乳房的自然构成的一部分。为了使检测系统在实践中具有鲁棒性和适用性，我们将乳腺组织的异质性和不同女性之间的统计变化视为探针开发的一个组成部分。将开发神经网络形式的人工智能算法来训练探头，以实现最大的检测概率。

In recent years, microwaves have emerged as a potential technique for breast cancer detection which avoids the discomfort, risks and costs associated with x-rays and excessive cost of MRI. Early microwave-based detection modalities were based on imaging of the dielectric properties with the goal of correlating high contrast breast inclusions with cancerous tissues. These techniques required the use of a plurality of antennas operating over an ultra-wide range of frequencies. The challenge with these techniques, however, was that a typical female breast is highly heterogeneous and contains a complex network of glandular adipose and fibroconnective tissues. Also, human tissues are highly dispersive (frequency-dependent), thus making the use of wideband antennas extremely difficult. Because of the rich variation in the dielectric properties of breast tissues, we remain convinced that microwave imaging has a strong role to play. To address the two fundamental challenges of earlier microwaves-based methods, we propose to develop near-field antenna probes. The probes will be designed to operate over an exceptionally narrow range of frequencies while maintaining high gain, in sharp contrast to traditional near-field probes. The narrow frequency range will render the frequency-dependent nature of the breast tissues irrelevant. The near-field aspect of the probe, coupled with sub-GHz frequencies, allow significant penetration of low-power non-ionizing microwave energy within the breast. Therefore, the presence of an anomaly will create a differentiated resonance profile since the anomaly is not part of the natural makeup of the breast. For the detection system to be robust and suitable in practice, we consider the heterogeneity of the breast tissue and statistical variation amongst different woman as an integral part of the probe development. Artificial intelligence algorithms in the form of neural networks will be developed to train the probe to achieve maximum detection probability.