果冻传媒

Ischemic stroke, caused by an interruption of the blood supply to the brain, is a major cause of death and disability worldwide. Atherosclerotic disease of the carotid artery is a major cause of ischemic stroke and transient ischemic attack. Carotid plaques are accumulations of fatty deposits, cellular debris, calcium, fibrin and other substances in the carotid arteries, which leads to stenosis (narrowing of these arteries). The presence and composition of these plaques are critical indicators for cardiovascular health, because certain types of plaques have a greater chance of rupturing and causing strokes.

Current guidelines recommend surgery if the stenosis exceeds 70%, but this often leads to overtreatment, because many plaques that are removed during surgery are stable. Vulnerable plaques, which have a greater chance of causing adverse events, often have characteristics such as a thin fibrous cap, intraplaque hemorrhages and a large lipid core. Ultrasound imaging (US) is often used to assess stenosis, but cannot reliably distinguish between stable and vulnerable plaques due to limited acoustic contrast. In photoacoustic imaging (PAI), light pulses are used to generate ultrasonic waves in the tissue, which provides contrast based on the optical absorption of different tissue compositions. PAI has great potential for identifying the composition of plaques.

However, the signal-to-noise ratio (SNR) is limited with US and especially with PAI, resulting in a low penetration depth and low contrast. A new technology for US transducers, capacitive machine ultrasound transducers (CMUTs), offers a broader frequency range and better integration possibilities at a relatively low cost, making them ideal for advanced imaging applications. In addition, multi-aperture imaging systems, which use multiple transducers to transmit or receive signals, offer enormous improvements in SNR, contrast and field of view (FOV).

Multi-perspective photoacoustic imaging (MP-PAI)

In his dissertation, Gholampour developed a prototype for multi-perspective photoacoustic imaging (MP-PAI) based on spatially distributed CMUTs to improve angle coverage and FOV. This system showed improvements in FOV, resolution and contrast, especially with a larger number of transducers. Experimental results of a plaque-mimicking phantom with three CMUTs confirmed the improvements, with a 20% increase in contrast and twice the structural coverage compared to a single CMUT. However, in some cases, the PSNR evaluations of MP-PAI scored almost the same or lower than those of single-perspective imaging.

To better understand SNR behavior in multi-aperture imaging systems, Gholampour conducted an extensive study, introducing a simplified k-space model to analyze the influence of system configuration on signal strength in both multi-aperture ultrasound and photoacoustic imaging geometries. The noise in multi-aperture imaging systems was analyzed for both coherent and incoherent aggregation methods, and analytical SNR estimates were proposed for each method. The theory was validated by numerical and experimental approaches.

Despite these findings, Gholampour noted that SNR must be further improved for better PAI of carotid plaques. A multi-aperture encoding scheme (MAES) was introduced to further increase the SNR of an MP-PAI system. MAES involves the simultaneous reception of multi-aperture photoacoustic signals in an encoded series, whereby the signals are restored by a decoding step prior to image reconstruction. The findings indicated that the SNR improved considerably for a larger number of apertures and when bipolar coding was applied. In addition, he also applied deep learning (DL) techniques to classify spectral photoacoustic imaging (sPAI) of carotid plaques. A convolutional neural network (CNN) model effectively distinguished constituent regions within plaques, which improved detection and classification.

Gholampour concludes his dissertation with a look into the future of imaging carotid plaques, discussing the further development of CMUT-based array prototypes and DL-based methods for image fusion. These advances contribute to an improved characterization and analysis of carotid plaques, with the goal of patient-specific in vivo imaging.

Title of PhD thesis: 鈥�鈥�

Supervisors: Richard Lopata, and Hans-Martin Schwab