Ultrasound imaging is frequently used for guiding minimally invasive percutaneous procedures such as peripheral nerve blocks, tumour biopsy and fetal blood sampling. Accurate and efficient identification of the proceduretarget and the needle is of paramount importance to ensurethe efficacy and safety of the procedures. Despite a number of prominent advantages associated with ultrasound imaging such as its real-time imaging capability, high affordability and accessibility, it suffers from intrinsically low soft tissue contrast that sometimes results in insufficient visibility of critical tissue structures such as nerves and small blood vessels. Moreover, visibility of clinical needles with ultrasound imaging is strongly dependent on the insertion angle and depth of the needle.  With steep insertion angles, ultrasound reflections can be readily outside of the transducer aperture, leading to poor ultrasound visibility.Loss of visibility of tissue targets or the needle can cause significant complications. At PURL, we investigate the combination of ultrasound and photoacoustic imaging for guiding minimally invasive procedures by offering complementary information to each other, with ultrasound imaging providing tissue structural information and photoacoustic imaging identifying critical tissue structures and invasive surgical devices such as metal-lic needles.


The imaging depths of surface-illumination-based photoacoustic imaging systems have been suffering from rapid attenuation of light with depth, which limits the clinical applicability of photoacoustic imaging. To address this challenge, we have developed an interventional photoacoustic imaging system with multiple excitation wavelengths for guiding minimally invasive procedures (Xia et al, 2017). A light delivery optical fibre was placed within a needle cannula and inserted into tissue for exciting photoacoustic signals. US detection was provided by a commercial US linear array imaging probe. The feasibility of the system for guiding fetal surgical procedures was demonstrated by visualising surface vasculature a freshly excised human placenta. 

Screenshot 2021-11-04 at 20.49.38.png


Laser diodes and light emitting diodes (LEDs) have shown promise as an alternative to commonly used solid-state laser excitation sources with their compact size and low cost, further benefiting the clinical transition of PA imaging (Xia et al, 2018). However, the pulse energy of LEDs is much lower than that of the laser, leading to degraded PA images with lower signal-to-noise (SNR) ratio.


Deep learning (DL) as a powerful tool for signal and image processing tasks in medical imaging could be explored for improving the image quality of the low-fluence based PA imaging system. Recently, we developed a DL-based framework based on semi-synthetic dataset for enhancing photoacoustic visualisation of clinical needles (Shi et al, 2021). The validations using blood-vessel-mimicking phantoms, ex vivo tissue and in vivo data show significant improvements by the proposed DL model compared to conventional reconstruction, which would be of great help on improving the outcomes of PA/US-guided minimally invasive procedures in the future.