Journal of Biomedical Optics最新文献

Significance: 5-Aminolevulinic acid (5-ALA) is a medical pro-drug used to induce the intracellular production of protoporphyrin IX (PpIX) via the heme synthesis pathway. Discoveries in mechanisms and developments in novel applications still continue with this uniquely endogenous intracellular optical system.

Aim: Understanding and exploiting the growing uses can be advanced through a survey of knowledge on the mechanisms and biokinetics of 5-ALA administration, partitioning, PpIX production, localization changes, clearance mechanisms, biological interactions, and methods for unique activation methods in both diagnostic and therapeutic applications.

Approach: The current medical uses of PpIX are reviewed, separating into therapeutic and diagnostic areas, and the expansion and lateral growth areas are outlined.

Results: Initially approved for photodynamic therapy of skin lesions, fluorescence diagnostic indications later developed to guide surgical resection in bladder cancer and glioma. Today, the 5-ALA-PpIX system's spatial-temporal complexity in photophysics and pharmacokinetics continues to lead to more uses, such as photodynamic priming to alter tissue, fast intracellular tissue oxygen sensing, infection, and burn imaging and therapy.

Conclusions: The 5-ALA-PpIX system has broad potential partly because of the ubiquity of the heme synthesis across many cell/tissue types, combined with natural selectivity, unique pharmacokinetics, bright fluorescence, and sufficiently strong singlet oxygen production.

Significance: Oral cancer surgery demands precise margin delineation to ensure complete tumor resection (healthy tissue margin $>5\mathrm{mm}$ ) while preserving postoperative functionality. Inadequate margins most frequently occur at the deep surgical margins, where tumors are located beneath the tissue surface; however, current fluorescent optical imaging systems are limited by their inability to quantify subsurface structures. Combining structured light techniques with deep learning may enable intraoperative margin assessment of 3D surgical specimens.

Aim: A deep learning (DL)-enabled spatial frequency domain imaging (SFDI) system is investigated to provide subsurface depth quantification of fluorescent inclusions.

Approach: A diffusion theory-based numerical simulation of SFDI was used to generate synthetic images for DL training. ResNet and U-Net convolutional neural networks were developed to predict margin distance (subsurface depth) and fluorophore concentration from fluorescence images and optical property maps. Validation was conducted using in silico SFDI images of composite spherical harmonics, as well as simulated and phantom datasets of patient-derived tongue tumor shapes. Further testing was done in ex vivo animal tissue with fluorescent inclusions.

Results: For oral cancer optical properties, the U-Net DL model predicted the overall depth, concentration, and closest depth with errors of $1.43\pm 1.84\mathrm{mm}$ , $2.26\pm 1.63\mu g/\mathrm{ml}$ , and $0.33\pm 0.31\mathrm{mm}$ , respectively, using in silico patient-derived tongue shapes with closest depths below 10 mm. In PpIX fluorescent phantoms of inclusion depths up to 8 mm, the closest subsurface depth was predicted with an error of $0.57\pm 0.38\mathrm{mm}$ . For ex vivo tissue, the closest distance to the fluorescent inclusions with depths up to 6 mm was predicted with an error of $0.59\pm 0.53\mathrm{mm}$ .

Conclusions: A DL-enabled SFDI system trained with in silico images demonstrates promise in providing margin assessment of oral cancer tumors.

Significance: As fluorescence-guided surgery (FGS) gains clinical adoption, robust and experimentally validated computational models for tissue fluorescence are increasingly essential. Although there have been several developments in modeling fluorescence with Monte Carlo simulations, the scope of the experimental validation has been limited in the parameters tested and phantoms used.

Aim: We aim to present and experimentally validate a graphics processing unit (GPU)-accelerated, voxel-based Monte Carlo fluorescence framework capable of modeling varying fluorophore concentrations, optical properties, and complex three-dimensional (3D) geometries.

Approach: A two-step approach (MCX-ExEm) based on Monte Carlo eXtreme was developed for simulating fluorescence. Both commercial reference targets and custom 3D-printed phantoms with well-characterized optical properties were imaged for varying parameters-including absorption, scattering, fluorophore concentrations, and geometries-and compared against simulations.

Results: Strong agreement is observed between simulated and experimental fluorescence across all tested parameters. MCX-ExEm accurately captures nonlinear quenching at high fluorophore concentrations, variations driven by scattering and absorption, intensity scaling with volume, and depth-dependent attenuation and resolution. Minor deviations occur primarily under low-scattering or low-absorption regimes, where optical characterization presents greater uncertainties.

Conclusions: By integrating experimentally validated simulations with a broad range of solid phantoms, this framework establishes a foundation for developing fluorescence digital twins, enabling faster and more systemic testing of fluorescence imaging systems. These findings can help accelerate the design and optimization of FGS and other fluorescence-based biomedical applications.

Significance: Molecular photoacoustic (PA) imaging with exogenous dyes faces a significant challenge due to the photobleaching of the dye that can compromise tissue visualization, particularly in 3D imaging. Addressing this limitation can revolutionize the field by enabling safer, more reliable imaging and improve real-time visualization, quantitative analysis, and clinical decision-making in various molecular PA imaging applications such as image-guided surgeries.

Aim: We tackle photobleaching in molecular PA imaging by introducing a platform-flexible deep learning framework that enhances SNR from single-laser pulse data, preserving contrast and signal integrity without requiring averaging of signals from multiple laser pulses.

Approach: The generative deep learning network was trained with an LED-illuminated PA image dataset and tested on acoustic resolution PA microscopy images obtained with single-laser pulse illumination. In vitro and ex vivo samples were first tested for demonstrating SNR improvement, and then, a 3D-scanning experiment with an ICG-filled tube was conducted to depict the usability of the technique in reducing the impact of photobleaching during PA imaging.

Results: Our generative deep learning model outperformed traditional nonlearning, filter-based algorithms and the U-Net deep learning network when tested with in vitro and ex vivo single pulse-illuminated images, showing superior performance in terms of signal-to-noise ratio ( $93.54\pm 6.07$ , and $92.77\pm 10.74$ compared with $86.35\pm 3.97$ , and $84.52\pm 11.82$ with U-Net for kidney, and tumor, respectively) and contrast-to-noise ratio ( $11.82\pm 4.42$ , and $9.9\pm 4.41$ compared with $7.59\pm 0.82$ , and $6.82\pm 2.12$ with U-Net for kidney, and tumor respectively). The use of cGAN with single-pulse rapid imaging has the potential to prevent photobleaching ( $9.51\pm 3.69\%$ with cGAN, and $35.14\pm 5.38\%$ with long-time laser exposure by averaging 30 pulses), enabling accurate, quantitative imaging suitable for real-time implementation, and improved clinical decision support.

Conclusions: We demonstrate the potential of a platform-flexible generative deep learning-based approach to mitigate the effects of photobleaching in PA imaging by enhancing signal-to-noise ratio from single pulse-illuminated data, the

Significance: In previous studies, we reported on a new untargeted contrast agent, TMR-PEG1k, that exhibits high and durable diagnostic performance in glioma models starting within minutes of administration. This agent uses a fluorophore in the visible regime, which, although it helps ensure high-resolution imaging of superficial tissue, precludes the detection of subsurface structures due to the limited optical penetration. Thus, development of a near-infrared (NIR) version of the agent that retains the properties of TMR-PEG1k would combine the favorable diagnostic performance of TMR-PEG1k with sensitivity to subsurface pathologies enabled by NIR imaging.

Aim: We aim to determine whether exchanging the fluorophore on TMR-PEG1k from tetramethylrhodamine to cyanine 7 would retain the highly favorable imaging properties exhibited by TMR-PEG1k.

Approach: Mice ( $N=5$ ) with orthotopic gliomas expressing green fluorescent protein (GFP) were co-administered TMR-PEG1k and Cy7-PEG1k. After 30 min, animals were euthanized, and the whole-body specimens were imaged using fluorescence cryotomography to recover co-registered three-dimensional (3D) fluorescence distributions of all fluorescent reporters. We evaluated the two agents using tumor-to-background ratio (TBR), contrast-to-noise ratio (CNR), area under the receiver operating characteristic curve (ROC-AUC), and normalized cross-correlation with GFP fluorescence (CC to GFP).

Results: Although the imaging system exhibited higher sensitivity to Cy7-PEG1k in phantoms, the 3D cryotomography results showed dramatic differences in the properties of the two agents in vivo. TMR-PEG1k produced highly selective tumor enhancement and concordance with GFP, whereas Cy7-PEG1k showed much lower selectivity, lower signal intensity, and produced no enhancement in many tumor regions. These observations were confirmed by the evaluation metrics, which were found to be (1) TBR = 5.2 versus 1.7 ( $p=0.0037$ ), (2) CNR = 17.7 versus 3.8 ( $p=0.046$ ), (3) ROC-AUC = 0.999 versus 0.821, and (4) CC to GFP = 0.90 versus 0.52, for TMR-PEG1k versus Cy7-PEG1k, respectively.

Conclusions: Cy7-PEG1k did not retain the favorable properties exhibited by TMR-PEG1k and thus is not a suitable NIR analog for this agent.

Significance: Confocal Raman spectroscopy (CRS) is a noninvasive technique that enables detailed biochemical analysis of the skin, providing insights into its structural and molecular composition. Atopic dermatitis and hand eczema, both characterized by impaired skin barrier function, require effective monitoring tools to assess treatment efficacy.

Aim: We aimed to quantitatively evaluate changes in skin physiological and biochemical parameters following the application of a ceramide-based moisturizer (test cream) compared with an aqueous moisturizer (control cream) in healthy volunteers and eczema patients.

Approach: Skin physiological assessments and CRS measurements were performed to evaluate water content and ceramide levels in both superficial and deeper skin layers after application of the two moisturizers.

Results: CRS revealed a significant increase in water content following test cream application, indicating improved skin hydration, whereas physiological measurements detected no statistically significant changes, underscoring the greater sensitivity of CRS. The test cream also significantly enhanced ceramide levels in both superficial and deeper skin layers, whereas the control cream increased ceramide levels only in the superficial stratum corneum (SC).

Conclusions: These findings suggest that the test cream has a substantial positive impact on hand eczema, potentially improving both skin barrier function and biochemical properties.

Significance: Airway wall elastography (AWE) is promising for evaluating upper airway obstructive disorders and airway injuries. Technologies for AWE based on endoscopic optical coherence tomography (OCT) provide micron-scale resolution to capture airway wall deformations during tidal breathing. Combined with an intraluminal pressure probe, these technologies can provide quantitative AWE as part of a routine bronchoscopy exam. However, scan times must be of short duration to mitigate risk.

Aim: Our objective is to reduce the scan time necessary to perform OCT elastography over a 50 mm length of the airway wall to less than 1 min.

Approach: We introduce an innovative, 4D OCT imaging technique that scans in a sawtooth pattern to revisit each axial position of the airway over a diversity of respiratory phases. An anatomical (long-range) OCT system capable of capturing cross-sections of the upper airway was employed in conjunction with an intraluminal pressure catheter. Scanned data are retrospectively sorted into axial bins with high- and low-pressure thresholds used to compute cross-sectional compliance (CC) within each bin across the length of the upper airway.

Results: 4D OCT was tested in simulation, on rigid and deformable samples, and on in vivo pigs undergoing bronchoscopy. A precise CC measurement with a 0.5 mm sampling resolution over a 50 mm scan length in under 42 s was achieved.

Conclusions: The retrospective, respiratory-gated 4D aOCT scanning method is a minimally invasive technique for measuring airway wall CC. The method exhibited high precision in controlled models, effectively detected elastic heterogeneity, and yielded clinically relevant results in in vivo pigs.

The editorial introduces the special issue honoring Brian C. Wilson as a pioneer of biomedical optics.

Significance: Photodynamic therapy (PDT) for the treatment of oral cancers and oral potentially malignant lesions can be enhanced by the capability of the photosensitizer to serve as a fluorescence contrast agent for treatment guidance. The development of image-based dosimetry reporters can inform treatment progress in real time to avoid under-treatment, leading to incomplete response and recurrence.

Aim: We investigate the hypothesis that imaging of protoporphyrin IX (PpIX) photoproduct (PP) accumulation may be leveraged as an implicit PDT dosimetry reporter for PDT using 5-aminolevulinic acid (ALA)-induced PpIX photosensitization.

Approach: In initial spectroscopy studies, we investigate dose-dependent changes in absorption and fluorescence spectra of PpIX corresponding to PP accumulation during red light (635 nm) delivery. We use spectral analysis to select fluorescence excitation and spectral filtering components for PP imaging during treatment. We evaluate the capability for imaging PP accumulation concomitant with PpIX photobleaching in tissue phantoms, 3D oral squamous cell carcinoma (OSCC) models, and in murine xenografts.

Results: Spectroscopy shows fluence-dependent changes in PpIX optical properties, and that excitation of photobleached PpIX with 450 nm light produces fluorescence emission associated with PpIX PPs. An existing handheld intraoral probe is shown to be capable of imaging dose-dependent PP accumulation with the addition of a spectral filter to isolate fluorescence emission longer than 650 nm. PP signal increases concomitant with PpIX photobleaching in a fluence-dependent manner and correlates with the extent of cytotoxic response in 3D cultures. PP accumulation is also observed to occur concomitantly with photobleaching in OSCC subcutaneous xenografts.

Conclusions: Overall, the results show that imaging of PP accumulation is feasible by adapting traditional photodiagnosis optical components and may serve as a dosimetry reporter for ALA-PDT, which is complementary to the measurement of PpIX photobleaching.

Significance: Linear-array-based photoacoustic imaging (PAI) combines functional imaging with structural imaging from ultrasound. However, it suffers from depth-dependent optical attenuation due to surface illumination, resulting in decreased signal amplitude and image contrast with depth. Existing attenuation compensation methods often amplify noise, creating a trade-off between depth enhancement and image quality.

Aim: We aim to develop a deep learning method that addresses the coupled problem of optical attenuation compensation and denoising for linear-array-based PAI and test the applicability in vivo.

Approach: We propose a vision-transformer-based generative model to address this coupled problem. A diverse dataset was created using simulated data and experimental twin phantoms. Vascular twin phantoms were made by printing digital images onto polyurethane films to test the performance of the model. We trained and compared three deep learning architectures, Pix2Pix, Residual U-Net, and the proposed Transformer U-Net, using various loss functions, including adversarial, MSE, PSNR, SSIM, and a combined PSNR + SSIM. We tested the model on small animal tumor images.

Results: Quantitative evaluation shows that PSNR + SSIM loss is robust in preserving structural details and suppressing noise. Under the pre-specified SSIM + PSNR training objective, Trans U-Net achieves the highest SSIM and PSNR across noise levels on both datasets. In vivo validation using murine breast tumor models and in vivo breast imaging confirmed the model's ability to enhance visualization of deep vascular structures without introducing noise amplification.

Conclusions: The proposed Trans U-Net effectively addresses the coupled problem of attenuation correction and denoising in handheld PAI. This method improves depth-resolved vascular imaging and is potentially useful in clinical and preclinical photoacoustic applications.