UNDERSTANDING MICROWAVE IMAGING ACQUISITION MODES

When reading microwave imaging papers, two terms appear again and again: monostatic and bistatic. They are often presented as the fundamental acquisition modes, the building blocks of how microwave systems collect data. For many readers, they quickly become shorthand for understanding an entire imaging pipeline.

But here is the problem. Focusing only on monostatic and bistatic configurations gives an incomplete picture of how certain modern microwave imaging systems work. In reality, experimental platforms (and increasingly, AI-driven research pipelines) operate in a much richer design space, where multistatic arrays, scanning systems, and hybrid acquisition strategies play a central role.

Understanding this broader landscape is not just a matter of hardware design. It directly impacts the quality of reconstructed images, the structure of datasets and the performance of machine learning models

The principle of microwave data acquisition

Microwave imaging is often described in terms of reconstruction algorithms or AI models. But before any image can be reconstructed or any model trained there is a more fundamental question: How is the data actually collected?

In a typical setup, antennas are placed around the breast. These antennas emit electromagnetic waves and measure how those waves scatter inside the tissue. Because tumors exhibit different dielectric properties than surrounding tissue, they modify the signals in measurable ways.

However, the type of measurements you obtain depends entirely on how antennas are configured.

Monostatic imaging

In a monostatic configuration, a single antenna both transmits and receives the signal. The antenna sends a pulse into the breast and measures the signal that is reflected. This approach resembles radar: (1) the transmitted wave propagates into the tissue, (2) scattering occurs at dielectric boundaries, (3) the antenna records the backscattered signal.

Monostatic systems are appealing because they are simple and relatively easy to calibrate. They are often used in early-stage prototypes, low-cost experimental setups, and scanning systems with a single rotating antenna.

However, this simplicity comes at a cost. Because measurements are limited to backscattered signals, monostatic systems provide only a partial view of the electromagnetic interactions inside the breast.

Bistatic imaging

In bistatic configurations, one antenna transmits while another receives the scattered signal. This setup introduces a spatial separation between transmitter and receiver. This introduces spatial diversity: (1) signals travel along different paths through the tissue, (2) receivers capture more complex scattering patterns, (3) the system becomes sensitive to features that monostatic setups may miss.

Bistatic systems provide richer information, but they also introduce additional complexity on synchronization between antennas, increased calibration requirements and larger number of measurements.

For many years, monostatic and bistatic configurations were treated as the primary categories of microwave imaging systems. But modern systems rarely stop there.

Multistatic imaging

In a multistatic configuration, multiple antennas surround the breast, each antenna can act as transmitter or receiver while signals are recorded across many antenna pairs. If a system has N antennas, it can generate N monostatic measurements and N(N−1) bistatic measurements.

This leads to a dramatic increase in data volume and diversity. And this is where things become particularly interesting for AI. Multistatic datasets capture a much richer representation of electromagnetic interactions, making them especially valuable for 3D reconstruction algorithms, inverse scattering techniques and deep learning models.

More diverse measurements help capture the complex scattering patterns created by heterogeneous breast tissue. In other words, multistatic systems are not just a hardware upgrade, they fundamentally change the information content of the data.

Rotational and scanning systems

Another approach avoids large antenna arrays altogether. Instead, a single antenna is mechanically moved or rotated around the breast. This creates what is known as a synthetic aperture:

• the antenna collects measurements from multiple positions,
• each position acts like a virtual antenna in an array,
• the system reconstructs images as if multiple antennas were used.

Scanning systems are widely used in research because they reduce hardware complexity, enable high-resolution datasets and are well-suited for controlled phantom experiments. Many public datasets, including those used in early-stage AI research, are generated using this type of setup.

Hybrid and Emerging Acquisition Modes

The most important takeaway is that monostatic and bistatic are not the only acquisition modes and in modern systems, they are rarely used in isolation. Researchers are increasingly exploring hybrid strategies, such as multistatic radar combined with tomography, adaptive antenna configurations and multimodal systems (e.g., microwave + ultrasound).

These approaches aim to balance signal diversity, hardware complexity and reconstruction accuracy. They open new possibilities for both physics-based and AI-driven imaging pipelines.

Why This Matters for AI and Datasets

For projects like WaveCare, acquisition strategy is not just a hardware detail, it is a data problem. The way signals are collected determines the structure of the dataset, the type of features available to machine learning models, and the generalization ability of those models.

For example:

• monostatic datasets → simpler, but limited information
• multistatic datasets → richer, but more complex
• scanning datasets → high resolution, but potentially less realistic

This is particularly relevant when working with public datasets such as UMBMID, where measurements include a mix of acquisition modes. Designing robust AI models therefore requires understanding not only the data but also how that data was generated.

Unlike mature imaging technologies such as MRI or CT, microwave imaging has not yet converged on a standard acquisition architecture. Instead, it remains an active research field where multiple approaches coexist. Each configuration provides a different perspective on how electromagnetic waves interact with biological tissue. And each comes with its own trade-offs.

Toward Smarter Imaging Systems

As microwave imaging evolves, the boundary between hardware design and data science is becoming increasingly blurred. Future systems will likely be co-designed with AI in mind:

• acquisition strategies optimized for machine learning
• datasets tailored for detection and localization tasks
• hybrid pipelines combining physics-based models and deep learning

In this context, understanding acquisition modes is no longer just a technical detail. It is a key step toward building intelligent imaging systems capable of detecting cancer earlier, faster, and more reliably.

Sources and further reading

• Solis-Nepote, M. et al., An Air-Operated Bistatic System for Breast Microwave Radar Imaging: Pre-Clinical Validation. Annu Int Conf IEEE Eng Med Biol Soc, 2019. https://doi.org/10.1109/embc.2019.8857646
• Wang, L., Microwave Imaging and Sensing Techniques for Breast Cancer Detection. Micromachines, 2023. https://doi.org/10.1038/s44303-024-00012-8
• UMBMID: University of Manitoba Breast Microwave Imaging Dataset (GitHub). https://github.com/UManitoba-BMS/UM-BMID