How demanding are medical imaging systems?

The physical principle of digital radiography (DR) is the same . X-rays passing through the human body are attenuated by human tissues with different radiation penetration and projected onto a flat-panel detector system, as shown in Figure 1.

Figure 1. Digital X-ray detector signal chain.

The detector converts X-ray photons into an electrical charge proportional to the energy of the incident particle. The resulting electrical signal is amplified and converted into the digital domain to produce an accurate digital representation of the X-ray image. The image quality depends on the sampling of the signal in both the spatial and intensity dimensions. In the spatial dimension, the minimum sampling rate is defined by the pixel matrix size of the detector and the update rate of the real-time fluoroscopic imaging.

Flat panel detectors with millions of pixels and typical update rates up to 25 fps to 30 fps use channel multiplexing and multiple ADCs with sampling rates up to tens of MSPS to meet the minimum conversion time requirements without sacrificing accuracy. In the intensity dimension, the digital output signal of the ADC represents the integrated amount of X-ray photons absorbed by a given pixel during a specific exposure time. This value is grouped into a finite number of discrete levels defined by the bit depth of the ADC.

Another important parameter is the signal-to-noise ratio (SNR), which defines the system's intrinsic ability to faithfully represent the anatomical features of the imaged body. Digital X-ray systems use 14-bit to 18-bit ADCs with SNR levels ranging from 70 dB to 100 dB, depending on the type of imaging system and its requirements. There are a wide variety of discrete ADCs and integrated analog front ends that enable various types of DR imaging systems to have higher dynamic range, finer resolution, higher detection efficiency, and lower noise.

Computed tomography (CT) also uses ionizing radiation technology, but unlike digital X-ray technology, it is based on a fan-shaped detector system that rotates synchronously with the X-ray source and uses more complex processing technology to generate high-resolution 3D images of blood vessels, soft tissues, etc.

The CT detector is the core component of the entire system architecture. It is actually the heart . It consists of multiple modules, as shown in Figure 2. Each module converts the incident X-rays into electrical signals and routes them to a multi-channel analog data acquisition system (ADAS). Each module contains a scintillator crystal array, a photodiode array, and an ADAS with multiple integrator channels multiplexed to an ADC. The ADAS must have very low noise performance to maintain good spatial resolution, reduce X-ray dose, and have very low current output to achieve high dynamic range performance.

Figure 2. CT detector module signal chain.

To avoid image artifacts and ensure good contrast, the converter front end must have excellent linearity performance and provide low-power operating modes to reduce cooling requirements for heat-sensitive detectors. The ADC must have a high resolution of at least 24 bits to obtain better and clearer images, while also having a fast sampling rate (as short as 100 μs) to digitize the detector readings. The ADC sampling rate must also support multiplexing, so that a smaller number of converters can be used and the size and power consumption of the entire system can be reduced.

Positron emission tomography (PET) involves ionizing radiation produced by a radionuclide introduced into the body. It emits positrons that collide with electrons in tissue, producing pairs of gamma rays radiating in roughly opposite directions.

These high-energy photon pairs simultaneously strike opposing PET detectors, which are arranged in a ring around the port of the stent. The PET detector (shown in Figure 3) consists of an array of scintillation crystals and photomultiplier tubes (PMTs), which convert gamma rays into current, which is then converted into voltage, which is then amplified and compensated for amplitude variations by a variable gain amplifier (VGA). The resulting signal is then split between ADC and comparator paths to provide energy and timing information that the PET coincidence processor uses to reconstruct a 3D image of the radiotracer concentration in the body.

Figure 3. PET electronic front-end signal chain.

If two photons have an energy of about 511 keV and their detection times differ by less than one nanosecond, they can be classified as correlated. The energy of the photons and the difference in detection time place stringent requirements on the ADC, which must have high resolution of 10 to 12 bits and fast sampling rates, typically greater than 40 MSPS. Low noise performance maximizes dynamic range, while low power operation reduces heat dissipation, both of which are important for PET imaging.

Magnetic resonance imaging (MRI) is a non-invasive medical imaging technique that relies on the phenomenon of nuclear magnetic resonance and does not require the use of ionizing radiation, which distinguishes it from DR, CT, and PET systems.

The carrier frequency of the MR signal is directly proportional to the main magnetic field strength and ranges from 12.8 MHz to 298.2 MHz for commercial scanners. The signal bandwidth is defined by the field of view in the frequency encoding direction and can vary from a few kHz to tens of kHz.

This places special demands on the receiver front end, which is typically based on a superheterodyne architecture with a relatively slow SAR ADC (see Figure 4). However, recent advances in analog-to-digital conversion have enabled fast, low-power multichannel pipeline ADCs to directly digitize MR signals at conversion rates exceeding 100 MSPS with 16-bit depth over the most common frequency ranges. The dynamic range requirements are very stringent, typically exceeding 100 dB.

Figure 4. MRI superheterodyne receiver signal chain.

Oversampling the MR signal can improve the resolution, increase the SNR, and eliminate aliasing artifacts in the frequency encoding direction, thereby enhancing image quality. To achieve fast scan acquisition time, compressed detection technology based on undersampling can be applied.

Sonography, or medical ultrasound, uses a different physical principle than all other imaging modalities discussed in this article. It uses sound wave pulses with a frequency range of 1 MHz to 18 MHz. These sound waves scan the internal tissues of the human body and reflect as echoes of varying intensities. These echoes are acquired in real time and displayed as a sonogram, which may contain different types of information, such as acoustic impedance, blood flow, the activity of the tissue over time, or its stiffness.

The key functional blocks of the medical ultrasound front end (shown in Figure 5) are represented by an integrated multi-channel analog front end (AFE), which includes a low noise amplifier, a variable gain amplifier, an anti-aliasing filter (AAF), an ADC, and a demodulator. One of the most important requirements for the AFE is the dynamic range. Depending on the imaging mode, this requirement may need to reach 70 dB to 160 dB in order to distinguish the blood signal from the background noise generated by the motion of the probe and body tissue.

Figure 5. Medical ultrasound front-end signal chain.

Therefore, the ADC must have high resolution, high sampling rate, and low total harmonic distortion (THD) to maintain the dynamic fidelity of the ultrasound signal. The high channel density of the ultrasound front end also requires low power consumption. A range of integrated AFEs available for medical ultrasound equipment can achieve optimal image quality while reducing power consumption, system size, and cost.

Medical imaging places extremely demanding demands on electronic design. Low power, low noise, high dynamic range, and high resolution performance at low cost and in compact packages are trends driven by the requirements of modern medical imaging systems discussed in this article. ADI meets these requirements, providing highly integrated solutions for key signal chain functional blocks, driving the realization of state-of-the-art clinical imaging equipment that is increasingly becoming an integral part of today's international healthcare system.

Finally, let me share ADI’s products that are suitable for the various medical imaging modes mentioned in this article:

ADAS1256 : This highly integrated analog front end contains 256 channels with low noise integrators, low-pass filters, and correlated double samplers multiplexed into a high speed 16-bit ADC. It is a complete charge-to-digital conversion solution designed for DR applications that can be mounted directly on a digital X-ray panel.

For discrete DR systems, the 18-bit PulSAR® ADC AD7960 offers 99 dB SNR and 5 MSPS sampling rate, providing unmatched performance to meet the highest dynamic range noise and linearity requirements. The 16-bit, dual-channel AD9269 and 14-bit, 16-channel AD9249 pipeline ADCs offer sampling rates up to 80 MSPS and 65 MSPS, respectively, to enable high-speed fluoroscopy systems.

ADAS1135 and ADAS1134 : These highly integrated 256-channel and 128-channel data acquisition systems consist of a low noise/low power/low input current integrator, a synchronous sample-and-hold device, and two high speed ADCs with configurable sampling rates and up to 24-bit resolution, providing excellent linearity to maximize image quality in CT applications.

AD9228 , AD9637 , AD9219 , and AD9212 : These 12-bit and 10-bit multichannel ADCs, with sampling rates from 40 MSPS to 80 MSPS, are optimized for excellent dynamic performance and low power consumption to meet PET requirements.

AD9656 : This 16-bit, quad-channel pipeline ADC offers conversion rates up to 125 MSPS and is optimized for traditional direct digital conversion MRI system architectures, with excellent dynamic performance and low power consumption.

AD9671 : This 8-channel integrated receiver front end is designed for low-cost, low-power medical ultrasound applications and uses a 14-bit ADC with a sampling rate of up to 125 MSPS. Each channel is optimized for high dynamic performance of 160 dBFS/√Hz and low power of 62.5 mW in continuous wave mode for applications requiring small package size.