System-level optimization in spectroscopic photoacoustic imaging of prostate cancer

This study presents a system-level optimization of spectroscopic photoacoustic (PA) imaging for prostate cancer (PCa) detection in three folds. First, we present a spectral unmixing model to segregate spectral system error (SSE). We constructed two noise models (NMs) for the laser spectrotemporal fluctuation and the ultrasound system noise. We used these NMs in linear spectral unmixing to denoise and to achieve high temporal resolution. Second, we employed a simulation-aided wavelength optimization to select the most effective subset of wave-lengths. NMs again were considered so that selected wavelengths were not only robust to the collinearity of optical absorbance, but also to noise. Third, we quantified the effect of frame averaging on improving spectral unmixing accuracy through theoretical analysis and numerical validation. To validate the whole framework, we performed comprehensive studies in simulation and an in vivo experiment which evaluated prostate-specific membrane antigen (PSMA) expression in PCa on a mice model. Both simulation analysis and in vivo studies confirmed that the proposed framework significantly enhances image signal-to-noise ratio (SNR) and spectral unmixing accuracy. It enabled more sensitive and faster PCa detection. Moreover, the proposed framework can be generalized to other spectroscopic PA imaging studies for noise reduction, wavelength optimization, and higher temporal resolution.

Automated summary

This study presents a system-level optimization of spectroscopic photoacoustic imaging for prostate cancer detection, which improves the image signal-to-noise ratio and spectral unmixing accuracy through strategies such as spectral unmixing model, wavelength optimization, and frame averaging. Simulation and in vivo experiments confirm the effectiveness and generalizability of this method.