Breast photoacoustic tomography (PAT) enables noninvasive vascular imaging with high spatial resolution and optical contrast, but most prior studies have focused on static anatomy rather than functional dynamics. We introduce hypoxia-correlation functional PAT (hc-PAT), which leverages the abnormal vascular reactivity of malignant tumors. Participants underwent alternating breath-holding and normal breathing to induce transient systemic hypoxia, monitored by pulse oximetry. Simultaneous hc-PAT imaging was performed, and voxel-wise correlation with the oxygenation waveform generated functional maps. Malignant tumors showed significantly higher correlation coefficients than surrounding tissue, clearly delineating lesions consistent with magnetic resonance imaging (MRI) and tumor pathophysiology. Quantitative analysis revealed greater inter-breast asymmetry in malignant versus benign cases (p∼0.014). hc-PAT also sensitively tracked response to neoadjuvant chemotherapy, showing reduced hypoxia reactivity alongside MRI-confirmed tumor shrinkage. This pilot study demonstrates that hc-PAT can enhance breast cancer detection using endogenous functional contrast, suggesting the potential for sensitive, injection-free functional imaging that warrants validation in larger cohorts.
Photoacoustic tomography (PAT) is widely utilized in biomedical imaging due to its high optical contrast and deep acoustic penetration. However, the scarcity of high-quality datasets hinders the application of deep learning in this field. To address this, we propose a latent diffusion model (LDM)-based dual-domain data generation framework, which consists of two independently-trained synthesis pathways, that synthesizes photoacoustic data in both image and projection domains. Our method maps raw data into a downsampled latent space, preserving essential structural and semantic information while significantly reducing modeling dimensionality. The two pathways are trained separately to model their corresponding data distributions. To validate the framework, the generated data was used to train a reconstruction network for sparse-view, limited-view, and bandwidth-limited tasks. Experimental results show that models trained with synthetic data achieve performance comparable to reference data training in terms of structural fidelity, artifact suppression, and overall image quality. Quantitative evaluation via PSNR and SSIM metrics further confirms the high usability of the generated data. This study represents the first systematic exploration of dual-domain generation in PAT, offering a robust strategy for deep learning research under medical data constraints.
Detection and continuous monitoring of circulating tumor cells (CTCs) using photoacoustic imaging remains challenging because of weak transient signals and strong blood-background absorption. This proof-of-concept study investigated the lower lip as a potential photoacoustic imaging window for melanoma CTC detection. A tissue-mimicking lower-lip phantom incorporating vessel-mimicking channels, physiologically relevant flow conditions, and donated human blood was developed. Melanoma-cell targets, including HS936.T and melanin-enhanced Mel624+α-MSH clusters, were imaged using photoacoustic tomography. Singular value decomposition (SVD)-based background suppression and kymograph-based temporal analysis were applied to enhance and characterize transient target-passage events. A composite kymograph-derived Target Score was developed for both automated classification and temporal guidance of visual inspection. Detectability depended on target size, melanin content, and surrounding medium. Mel624+α-MSH clusters remained detectable under blood-containing conditions, whereas similarly sized HS936.T clusters were not detected. Single HS936.T cells were also not detected, and the smallest detected cluster at the tested melanin concentrations was 500μm. Operator-referenced automated classification achieved an F1 score of 0.800 and an AUC of 0.733, while temporal guidance captured 83.3% of visually identified passages and reduced the temporal search space by 91.79%. Limited blood and cell availability constrained systematic evaluation and replication. At this proof-of-concept stage, the framework provides a semi-automated approach combining automatic scoring with focused visual inspection. These findings support further investigation of the lower lip as a potential site for photoacoustic monitoring of circulating melanoma cells and provide a foundation for developing automated and semi-automated CTC detection approaches in future studies.
Photoacoustic imaging (PAI) under limited-view conditions often suffers from severe reconstruction artifacts due to incomplete data, compromising image interpretation and quantitative analysis. In this paper, we propose DPAT-Net, an end-to-end reconstruction network that combines known acquisition geometry with learned measurement- and image-domain priors. The network employs a “dual-domain processing” strategy: in the data domain, a hybrid sinogram enhancement module with bottleneck axial attention processes limited-view sensor-time data by modeling long-range sensor-axis dependencies; subsequently, a fixed-geometry Differentiable Back-Projection (DBP) layer maps both the enhanced sensor-time data and the raw limited-view measurements into the image domain, while allowing image-domain gradients to reach the upstream enhancement module; finally, an Image Refinement Network fuses the two preliminary images to produce the final reconstruction. Validation on a vascular simulation dataset, an in vivo mouse-brain dataset, and the public PATATO preclinical phantom dataset demonstrates higher PSNR and SSIM than the evaluated baseline methods under limited-view sampling. Notably, on the in vivo dataset, the proposed method achieves the highest PSNR and SSIM relative to the full-view reference reconstruction among the evaluated methods using only one-quarter of the measurement data (32 channels), with an inference time of 9.97 ms per image on an RTX 4090 GPU.
Optical-resolution photoacoustic microscopy (OR-PAM) enables label-free, subcellular-resolution microvascular imaging, but the speed-resolution trade-off remains a bottleneck. While sparse sampling accelerates acquisition, the disparity between OR-PAM’s tight optical focus and enlarged scanning steps causes irreversible information loss. To mitigate the limitations of ultra-sparse sampling, a single-shot dual-scale PAM framework is proposed. An independent fiber-based optical foci-steering configuration enables synchronized OR and acoustic-resolution (AR) acquisition, which ensures scanning flexibility while harnessing AR signals to enhance reconstruction. The AR modality, governed by its broader acoustic point-spread function (PSF), acts as an intrinsic physical low-pass filter that robustly captures the target’s dominant vascular architecture. Guided by the theoretical AR-PSF, a deblurring operation is performed to obtain a sharpened structural prior. Subsequently, a dual-scale super-resolution generative adversarial network fuses this prior with ultra-sparse OR measurements to reconstruct high-fidelity 2D vascular MAP images. For reconstruction evaluation, the ultra-sparse inputs were generated by retrospectively decimating densely sampled in vivo data. Experimental results demonstrate that the proposed method significantly outperforms representative leading methods. By overcoming extreme spatial information scarcity under high-ratio sparse sampling up to 16 × 16 (0.4% sampling rate), the proposed method achieves a 16.1% SSIM improvement and a 2.18 dB PSNR gain over the best-performing baseline, together with a 33.7% reduction in LPIPS.
