SignificancePhotoacoustic imaging (PAI) promises to measure spatially resolved blood oxygen saturation but suffers from a lack of accurate and robust spectral unmixing methods to deliver on this promise. Accurate blood oxygenation estimation could have important clinical applications from cancer detection to quantifying inflammation.AimWe address the inflexibility of existing data-driven methods for estimating blood oxygenation in PAI by introducing a recurrent neural network architecture.ApproachWe created 25 simulated training dataset variations to assess neural network performance. We used a long short-term memory network to implement a wavelength-flexible network architecture and proposed the Jensen–Shannon divergence to predict the most suitable training dataset.ResultsThe network architecture can flexibly handle the input wavelengths and outperforms linear unmixing and the previously proposed learned spectral decoloring method. Small changes in the training data significantly affect the accuracy of our method, but we find that the Jensen–Shannon divergence correlates with the estimation error and is thus suitable for predicting the most appropriate training datasets for any given application.ConclusionsA flexible data-driven network architecture combined with the Jensen–Shannon divergence to predict the best training data set provides a promising direction that might enable robust data-driven photoacoustic oximetry for clinical use cases.
Longitudinal characterisation of the tumour vascular response to radiotherapy is essential for understanding the role of oxygenation and microvascular disruption in response to therapy. Using multi-scale in vivo photoacoustic imaging (PAI), we assessed early response to two hypofractionated radiotherapy schemes in two human breast cancer models. Mesoscopic and multispectral tomographic photoacoustic imaging was performed 24h pre-, post-radiotherapy, and at endpoint. PAI biomarkers were validated ex vivo with multiplex immunofluorescence using a 20-plex panel developed specifically for vascular response assessment at sub-cellular resolution. PAI captured radiotherapy response, revealing the differential effect between radiotherapy schemes and models with different hypoxia phenotypes.
Longitudinal mesoscopic photoacoustic imaging of vascular networks requires accurate image co-registration to assess local changes in growing tumours, but remains challenging due to sparsity of data and scan-to-scan variability. Here, we compared a set of 5 curated co-registration methods applied to 49 pairs of vascular images of mouse ears and breast cancer xenografts. Images were segmented using a generative adversarial network and pairs of images and/or segmentations were fed into the 5 tested algorithms. We show the feasibility of co-registering vascular networks accurately using a range of quality metrics, taking a step towards longitudinal characterization of those complex structures.
Biomedical optical imaging and sensing techniques are known to be confounded by skin tone, typically leading to worse outcomes in people with darker skin tones. Here, we present a healthy volunteer study to evaluate the effects of skin tone in photoacoustic imaging. We recruited 42 people, 6 from each Fitzpatrick skin type and 6 people with vitiligo. Our preliminary analysis shows increased reconstruction artefacts and changes in blood oxygen estimates in higher Fitzpatrick types. The results suggest that equitable application of quantitative photoacoustic imaging in the clinic will require improved methods to account for changing light fluence and acoustic artefacts.
Photoacoustic imaging holds promise in clinical applications, but lacks standardized testing methods. To overcome this, the International Photoacoustic Standardization Consortium (IPASC) assessed the fabrication reproducibility of a stable tissue-mimicking phantom material in an international multicenter study (n>15 centers). The material consists of mineral oil, polymer, ink, and titanium dioxide. Participating centers followed a recipe set up by the main site (Cambridge, UK) and returned samples for characterization. The results demonstrate promising reproducibility for acoustic, photoacoustic and optical properties. By performing this study, IPASC hopes to broaden the uptake of a stable phantom material, supporting system validation and testing.
SignificancePhotoacoustic imaging (PAI) provides contrast based on the concentration of optical absorbers in tissue, enabling the assessment of functional physiological parameters such as blood oxygen saturation (sO2). Recent evidence suggests that variation in melanin levels in the epidermis leads to measurement biases in optical technologies, which could potentially limit the application of these biomarkers in diverse populations.AimTo examine the effects of skin melanin pigmentation on PAI and oximetry.ApproachWe evaluated the effects of skin tone in PAI using a computational skin model, two-layer melanin-containing tissue-mimicking phantoms, and mice of a consistent genetic background with varying pigmentations. The computational skin model was validated by simulating the diffuse reflectance spectrum using the adding-doubling method, allowing us to assign our simulation parameters to approximate Fitzpatrick skin types. Monte Carlo simulations and acoustic simulations were run to obtain idealized photoacoustic images of our skin model. Photoacoustic images of the phantoms and mice were acquired using a commercial instrument. Reconstructed images were processed with linear spectral unmixing to estimate blood oxygenation. Linear unmixing results were compared with a learned unmixing approach based on gradient-boosted regression.ResultsOur computational skin model was consistent with representative literature for in vivo skin reflectance measurements. We observed consistent spectral coloring effects across all model systems, with an overestimation of sO2 and more image artifacts observed with increasing melanin concentration. The learned unmixing approach reduced the measurement bias, but predictions made at lower blood sO2 still suffered from a skin tone-dependent effect.ConclusionPAI demonstrates measurement bias, including an overestimation of blood sO2, in higher Fitzpatrick skin types. Future research should aim to characterize this effect in humans to ensure equitable application of the technology.
Machine learning-based approaches have shown promise for quantitative photoacoustic oximetry, however, the impact of learned methods is hampered by challenges of usability and generalisability, caused by the strong dependence of learned methods on the training data sets. To address these issues we developed a deep learning-based approach with higher flexibility. The method is trained on a suite of training data sets representing a range of general assumptions. The performance is systematically compared to linear unmixing methods and is validated on in silico, in vitro, and in vivo data representing different use cases.
Here, we assess the capabilities of photoacoustic imaging (PAI) biomarkers to shed light into perfusion-limited hypoxia, a key driver of tumor malignancy. Using two breast cancer xenograft models, we found that photoacoustic tomography could detect higher fluctuations in oxygen saturation (sO2MSOT) in models with higher disease aggressiveness, supported by an overall lower sO2MSOT and greater spatial heterogeneity in sO2MSOT. Photoacoustic mesoscopy revealed differences in vascular architecture and perfusion dynamics between the models. The results were validated using immunohistochemistry and RNA sequencing, highlighting the potential of PAI to provide non-invasive insight on dynamic phenomena associated with perfusion-limited hypoxia in vivo.
We developed an open-source python toolkit for photoacoustic image (PAI) reconstruction and processing. The toolkit implements GPU-accelerated processing algorithms including preprocessing, image reconstruction (backprojection and model-based) and multispectral analysis (linear spectral unmixing and learned spectral decolouring). We implemented methods for the advanced analysis of longitudinal PA data, including standardised analysis of oxygen-enhanced and dynamic contrast enhanced MSOT data. The toolkit currently works with pre-clinical, clinical and simulated PA systems, integrating with the IPASC open data format, simulated datasets from the SIMPA toolkit and iThera Medical MSOT devices. It can easily be extended to support other algorithms and systems.
KEYWORDS: Photoacoustic imaging, Monte Carlo methods, Skin, Tissues, Tissue optics, Blood, Systems modeling, In vivo imaging, In vitro testing, Chromophores
Spectral colouring severely affects quantification in photoacoustic imaging, impacting the biological interpretation of the output imaging biomarkers. Melanin is a particularly strong optical absorber in the near infrared wavelength range that exhibits substantial variation in concentration according to skin tone. Here, Monte-Carlo simulations of light transport in a computational skin phantom were carried out to establish the effects of quantifying blood oxygenation at different melanin concentrations. These results were validated with a tissue-mimicking phantom. The results demonstrated that raised melanin concentration in the epidermis significantly affects quantification of haemoglobin concentration and oxygenation with photoacoustic imaging.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.