Journal of Biomedical Optics最新文献

Significance: Optical coherence tomography angiography (OCTA) offers dye-free, three-dimensional views of skin microvasculature, yet progress in developing reliable and quantitative solutions for vessel architecture analysis is slowed by heterogeneous preprocessing practices, scarce annotated data, and limited evaluation metrics.

Aim: We assess how typical OCTA preprocessing steps influence the accuracy of deep learning vessel segmentation, and we identify network designs and metrics best suited to OCTA dermatological data.

Approach: Experiments use the open DERMA-OCTA dataset containing 330 volumes from different skin conditions; each volume is additionally provided in five progressively pre-processed versions: original, Bscan normalization, projection artifact attenuation, contrast enhancement, and vesselness filtering. Segmentation is performed with representative 2D and 3D deep learning approaches. Besides standard segmentation metrics, evaluation includes the connectivity-area-length index, which proved particularly effective for assessing dermatological vessel segmentation.

Results: The analysis shows that Bscan normalization, projection artifact attenuation, and contrast enhancement incrementally improve segmentation accuracy, whereas vesselness enhancement can impair segmentation performance. Among the tested architectures, 2D models achieved the highest segmentation performance, although 3D approaches proved more effective for deeper tissue layers. Testing across different pathologies revealed challenges in model generalization to varied vascular patterns.

Conclusions: Combining 2D and 3D models and using topology-aware indices provide a full, clinically relevant evaluation of algorithm performance.

Significance: Accurate intraoperative assessment of intestinal tissue viability is critical in determining the extent of resection in cases of intestinal ischemia. Current evaluation methods are largely subjective and lack the precision required for reliable decision-making during surgery.

Aim: We aim to develop and validate a hyperspectral imaging (HSI) system combined with machine learning (ML) to objectively assess intestinal wall viability and differentiate between reversible and irreversible ischemia.

Approach: A portable HSI system was used to acquire spectral data from rat models with induced intestinal ischemia at different time points (1, 6, and 12 h). Tissue oxygen saturation was calculated using a two-wavelength algorithm. Spectral data were classified using an ML pipeline based on principal component analysis (PCA) and the XGBoost algorithm, trained on histologically validated tissue classes.

Results: Tissue saturation decreased with prolonged ischemia (from 66% in healthy tissue to 21% after 12 h). Classification accuracy using PCA features reached 98% for intact tissue, 95% for possibly reversible ischemia, and 97% for irreversible ischemia. Classification maps closely matched tissue saturation distributions and histological findings. Initial clinical testing confirmed the system's sensitivity to ischemic changes in human subjects, although further training on human data is required for ML application.

Conclusions: HSI combined with ML provides an effective, non-invasive tool for real-time intraoperative assessment of intestinal viability. This approach improves the objectivity of surgical decision-making and may reduce unnecessary resections.

Significance: Understanding thermal effects on tissue optical properties is fundamental for optimizing laser-based medical interventions. We address the critical knowledge gap of temperature-dependent changes in porcine dermis optical properties.

Aim: We explore the thermal damage influence on the excised dermis optical properties at wavelengths from 400 to 1100 nm.

Approach: Using a double-integrating-sphere system and inverse adding-doubling, we determined absorption, ${\mu }_a$ , and reduced scattering, ${\mu }_s^{\text{'}}$ , coefficients before and after a 2.5-min thermal exposure.

Results: We observed non-linear changes in both ${\mu }_a$ and ${\mu }_s^{\text{'}}$ across temperature regimes. Minimal changes occurred at 37°C and 43°C. At 50°C, slight increases in both coefficients were observed. Significant alterations occurred at 60°C, with substantial increases in ${\mu }_s^{\text{'}}$ and variable changes in ${\mu }_a$ depending on wavelength region. At 70°C, ${\mu }_s^{\text{'}}$ values remained elevated, whereas ${\mu }_a$ showed mixed responses, with some wavelength regions decreasing, indicating progressive structural breakdown. The Arrhenius damage model showed an exponential increase with temperature.

Conclusions: We reveal complex thermal-induced changes in tissue optical properties, particularly at higher temperatures. Findings reinforce a critical threshold between 50°C and 60°C where significant changes occur. The non-linear, wavelength-dependent responses emphasize the need for comprehensive data in laser-tissue interaction modeling, with important implications for optimizing laser-based medical treatments.

Significance: Lumen segmentation in intravascular optical coherence tomography (IVOCT) images is essential for quantifying vascular stenosis severity, location, and length. Current methods relying on manual parameter tuning or single-frame spatial features struggle with image artifacts, limiting clinical utility.

Aim: We aim to develop a temporal residual U-Net (TR-Unet) leveraging spatiotemporal feature fusion for robust IVOCT lumen segmentation, particularly in artifact-corrupted images.

Approach: We integrate convolutional long short-term memory networks to capture vascular morphology evolution across pullback sequences, enhanced ResUnet for spatial feature extraction, and coordinate attention mechanisms for adaptive spatiotemporal fusion.

Results: By processing 2451 clinical images, the proposed TR-Unet model shows a well performance as Dice coefficient = 98.54%, Jaccard similarity (JS) = 97.17%, and recall = 98.26%. Evaluations on severely blood artifact-corrupted images reveal improvements of 3.01% (Dice), 1.3% (ACC), 5.24% (JS), 2.15% (recall), and 2.06% (precision) over competing methods.

Conclusions: TR-Unet establishes a robust and effective spatiotemporal fusion paradigm for IVOCT segmentation, demonstrating significant robustness to artifacts and providing architectural insights for temporal modeling optimization.

Significance: The effects of optogenetic stimulation (OS) on in vitro neural network behavior were studied through a reservoir computing-based obstacle avoidance task, revealing its impact on the task-processing capabilities of the network. Furthermore, it is demonstrated that a minimal output of signals from 15 neurons in the network is sufficient to achieve stable task control, with a success rate exceeding 95%. The optogenetically enhanced biological reservoir computing frame could find applications in neuro-robotic control and brain-inspired intelligence.

Aim: We aim to utilize optogenetically controlled in vitro neural networks and the first-order reduced and controlled error (FORCE) learning algorithm to achieve obstacle avoidance in neuro-robotic systems.

Approach: We presented an all-optical biological reservoir computing framework that leverages optogenetics and calcium imaging to precisely regulate and record neuronal activities. A closed-loop system was developed incorporating the FORCE learning algorithm, which guided a virtual car through obstacle avoidance tasks.

Results: The system demonstrated high accuracy and efficiency in navigating obstacles, achieving optimal performance after $\sim 150s$ of training. OS significantly improved the obstacle avoidance success rate, enhancing the system's adaptability and accuracy.

Conclusions: The results highlight the potential of optogenetically controlled biological neural networks in neuro-robotic systems, showcasing their capability to achieve accurate and efficient obstacle avoidance through physical reservoir computing.

Significance: Cervical cancer is the fourth most common cancer among women globally. Hence, it is crucial to develop a noninvasive and portable optical imaging modality for the early detection of premalignant cervical lesions.

Aim: We present the development and clinical validation of GynoSight v2.0, an indigenously developed multispectral, non-chip-on-tip source, hand-held, portable transvaginal imaging probe, for evaluating tissue health and identifying anomalies, such as those linked to precancerous cervical lesions.

Approach: GynoSight v2.0 houses a 16 LEDs, 5-megapixel camera, and a Raspberry Pi 5 module. A comparative shadowing effect analysis was performed between GynoSight v2.0 and colposcopy by evaluating statistical metrics such as mean pixel intensity (MPI), shadow area percentage (SAP), entropy, and contrast-to-noise ratio. In addition, the relative oxygen saturation maps of the cervical tissue were computed from the multispectral registered image using the proposed discrete Fourier transform-based image registration technique.

Results: The images of $N=6$ ( $N=2$ normal, $N=1$ premalignant, and $N=3$ high-grade squamous intraepithelial lesion) subjects were acquired. A comparative shadowing analysis shows a 50-unit gray level value separation between the colposcope and GynoSight v2.0 images. The pixel values of the colposcope are skewed to the lower pixel values, and the pixel values of GynoSight v2.0 are spread uniformly over 0 to 255 gray-level pixel values. In addition, the statistical analysis showed that MPI, SAP, and entropy are significant metrics for shadowing effect quantification.

Conclusions: The colposcopy images showed more shadowing effects than the GynoSight v2.0 images and hence provide better illumination to aid in better diagnosis.

Significance: The use of tissue attenuation coefficients as biomarkers for disease detection is rising. However, especially for ex vivo studies, sample handling methods can notably impact tissue optical attenuation properties, and these effects have yet to be studied in detail.

Aim: We aim to compare and evaluate common methods for sample handling and assess their impact on the optical attenuation and structural properties of ex vivo colon tissue.

Approach: Six different handling methods were tested: Direct freezing at $-80°C$ , slow freezing in a cryobox with and without cryopreservation media, snap freezing in isopentane, formalin fixation, and fresh tissue stored directly in phosphate-buffered saline. All samples were imaged using optical coherence tomography; images were assessed qualitatively for morphological changes and quantitatively by extracting the tissue attenuation coefficient using the Lambert-Beer law. All handling methods were compared with representative histology (hematoxylin and eosin staining and periodic acid-Schiff staining).

Results: All sample handling methods showed a significant difference in tissue attenuation and morphology relative to the fresh tissue ( $p\ll 0.0001$ ), with frozen samples generally showing a lower attenuation coefficient, e.g., directly frozen ( $2.0\pm 1.0{\mathrm{mm}}^{-1}$ ) compared with formalin-fixed ( $2.5\pm 1.3{\mathrm{mm}}^{-1}$ ) and fresh tissue ( $2.5\pm 1.0{\mathrm{mm}}^{-1}$ ). Formalin-fixed and snap frozen samples had the smallest effect size ( $\delta =0.002$ and $-0.09$ , respectively). Macroscopic structural changes were also observed, including alterations to the epithelial layer and indications of goblet cell degradation for all methods but formalin fixation.

Conclusions: Understanding the impact of sample handling methods is critical to the accurate interpretation of morphology-based analysis. In the case of fresh tissue being unavailable, formalin-fixed and snap frozen tissue samples yield the best alternative with negligible effect sizes for colon tissue.

Significance: Digital holographic microscopy (DHM) has proven effective for particle segmentation within a given volume, making it well-suited for rapid monitoring of bacterial growth in microfluidic cartridges-such as in single-cell-based antimicrobial susceptibility testing assays. However, the development of optimal assays depends on a range of factors related to the instrument, consumables, and the sample itself. Despite this, comprehensive investigations into how these parameters influence the quality of the resulting phase images remain limited.

Aim: To address this problem, we systematically explore the effect of these factors, including the microfluidic chamber height and its material properties, the density of the suspension, and other sample-inherent properties, on the signal-to-noise ratio (SNR) of the reconstructed phase image.

Approach: We constructed an off-axis digital holographic microscope and defined a robust numerical processing pipeline allowing for the numerical reconstruction, refocusing and counting of suspended particles in a measurement volume spanning roughly $120\times 120\times 400\mu m^3$ , at 50× magnification. We analyzed the performance of this system using various dilution steps of silica microspheres, Gram-positive spherical Staphylococcus warneri and Gram-negative rod-shaped Escherichia coli bacteria, filled in commercial microfluidic chips with different chamber heights.

Results: Experimental results demonstrated the system's capability in reflecting the dilution steps over 2 to 3 orders of magnitude. Our SNR analysis highlighted the microfluidic chamber height and the density of the suspension as key contributors to the background noise, whereas the particles themselves seemed to have a negligible effect. From this insight, we were able to derive an analytical function to predict the SNR of a given DHM system for various concentrations, chamber heights, and particle types.

Conclusions: We successfully built a DHM system for counting suspended particles over a wide concentration range and for various microfluidic chamber heights. We also derived an initial framework for predicting and optimizing the performance of a given DHM system.

Significance: Noninvasive monitoring of blood components is important in daily health management. Conventional optical techniques such as attenuated total reflection (ATR) have limited penetration depth and sensitivity. Photoacoustic spectroscopy using a piezoelectric transducer (PZT-PAS) can detect components in the interstitial fluid beneath the stratum corneum with a relatively simple device.

Aim: We explored the feasibility of PZT-PAS for noninvasive analysis of blood components.

Approach: Biomimetic phantom experiments were conducted to evaluate depth sensitivity. A technique for enhancing signal strength by generating standing acoustic waves in tissue was validated by tuning the laser modulation frequency. Human trials were conducted to assess the capability of the method for predicting blood glucose levels.

Results: PZT-PAS successfully detected signals from depths beyond 10 to $20\mu m$ , outperforming ATR. Signal enhancement was achieved in a 2-mm-thick interdigital membrane using resonant acoustic conditions. In human trials, discriminant analysis to determine blood glucose levels relative to a threshold of 140 mg/dL showed an accuracy rate of 85.3%.

Conclusions: These results highlight the potential of PZT-PAS combined with signal processing for future wearable, noninvasive health care monitoring applications.