Journal of Biophotonics最新文献

Distinguishing osteosarcoma from bone callus remains a clinical challenge due to their morphological similarities. This study proposes J-CAN, a multimodal deep learning framework integrating hyperspectral imaging (HSI) and H&E-stained pathology for rapid and accurate classification. The HSI system captures 176 spectral bands (400–1000 nm), providing molecular-level insights. MobileNetV2 extracts spatial features, while 1D-CNN processes spectral signatures. A self-attention mechanism enhances feature selection, prioritizing key spectral and spatial characteristics to improve classification performance. Experimental results show that J-CAN outperforms conventional models, including LSTM, SVM, and 1D-CNN, achieving 87.33% accuracy, 89.07% sensitivity, and 85.49% specificity. These findings demonstrate the potential of HSI-driven deep learning for clinical pathology, enabling efficient, automated osteosarcoma diagnosis. This approach enhances diagnostic precision and provides a valuable tool for pathologists, addressing the limitations of traditional histopathological assessments and improving the differentiation between osteosarcoma and bone callus.

Assessing the biomechanical properties of tissues can provide important information for disease diagnosis and therapeutic monitoring. Optical coherence elastography (OCE) is an emerging technology for measuring the biomechanical properties of tissues. Clinical translation of this technology is underway, and steps are being implemented to streamline data collection and processing. OCE data can be noisy, data processing can require significant manual tuning, and a single acquisition may contain gigabytes of data. In this work, we introduce a convolutional neural network-based method to translate raw OCE phase data to strain for quasistatic OCE that is ~40X faster than the conventional least squares approach by bypassing many intermediate data processing steps. The results suggest that a machine learning approach may be a valuable tool for fast, efficient, and accurate extraction of biomechanical information from raw OCE data.

This study investigates the relationship between facial microcirculation and optical parameters using swept-source optical coherence tomography (SS-OCT). Facial microcirculation deterioration impacts skin health by causing blood stagnation and metabolic disorders, influencing optical properties like scattering and attenuation. High-resolution SS-OCT imaging revealed distinct correlations between blood flow and optical attenuation coefficient (OAC) across skin layers: epidermal OAC showed a negative correlation with blood flow (r = −0.29 ± 0.03), while dermal OAC demonstrated a positive correlation (r = −0.43 ± 0.06). Leveraging this finding, we developed a depth-encoded color OAC imaging mode that enhances facial microcirculation visualization without additional scanning. The work provides mechanistic insights into microcirculation-related skin deterioration and establishes a novel clinical tool for diagnosing and managing associated conditions.

Sickle cell disease (SCD) is a genetic blood disorder causing red blood cells to deform into a sickle shape, often leading to misdiagnosis. Early detection is crucial, but traditional screening is slow and labor-intensive. This paper introduces an intelligent microscope system for automated SCD screening, reducing manual intervention. The system uses an interferometric method to capture high-resolution 3D phase images, combined with a deep learning-based UNET model for semantic segmentation of sickle and healthy cells. Various machine-learning models classify RBCs, with the Gradient boosting model achieving 94.9% accuracy. The system is scalable, user-friendly, and well suited for resource-limited settings, offering a faster, more reliable diagnostic tool. This innovation not only improves SCD detection but also sets the stage for AI-driven haematological diagnostics. Future advancements will enhance system robustness and undergo extensive clinical validation.

Blood oxygen saturation measuring is crucial for the diagnosis of disease severity. Despite the great efforts in non-contact vital signs monitoring by image photoplethysmography (IPPG) technology, the high-efficiency acquisition of transient image data is still challenging and generally requires complicated data processing processes. In this paper, we demonstrated a novel method for rapid measurement of blood oxygen saturation. A snapshot multispectral imager was employed in the proposed solution to transiently capture changes in spectral image information from facial tissue skin.

Early identification of gastric adenomatous polyps and adenocarcinoma is vital for improving patient outcomes. This study proposes a hybrid CNN-GRU model to classify one-dimensional hyperspectral data from ex vivo gastric tissues, addressing limitations of traditional diagnostics. Our model innovatively combines convolutional neural networks (CNNs) and gated recurrent units (GRUs) to capture both spatial and sequential dependencies in spectral data. Experimental results demonstrate that our model achieves an accuracy of 86%, sensitivity of 88%, and specificity of 85%. Additionally, receiver operating characteristic analysis further underscores its robust performance with an area under the curve of 0.86, surpassing traditional methods and other baseline models. These findings highlight the potential of leveraging advanced machine learning techniques to enhance early diagnostic accuracy and treatment strategies. The proposed approach offers a promising tool for rapid, accurate differentiation of gastric lesions, underscoring the importance of integrating innovative technologies in clinical diagnostics.

In nailfold video recordings, the micro-shaking of the hand is amplified and interferes with physician observations and parameter measurement. We developed a fast and accurate registration method for large-field-of-view nailfold video images. Nailfold videos are first represented in the YCrCb color space, with the Cb spatial component replacing the original grayscale image to reduce sensitivity to illumination. The projection variance of each row/column is employed to improve registration accuracy and processing speed. The method was compared with Origin GrayDrop, feature point matching, unsupervised learning, and Adobe Premiere Pro in terms of the peak signal-to-noise ratio, structural similarity index, and mean squared error. The peak signal-to-noise ratio and structural similarity index are enhanced, and the mean squared error is reduced compared to the original projection method. Moreover, the proposed method is faster than the comparison methods and provides the best combination of registration accuracy and fast processing for nailfold video image registration.

We developed a stable, high-penetration arm-type swept-source optical coherence tomography (SS-OCT) system for visualizing retinal and choroidal structures in pediatric patients with congenital cataracts. The system features a compact OCT probe with an integrated iris camera and fixation target for easy alignment, mounted on a five-degree-of-freedom adjustable arm to reduce motion artifacts and operator fatigue. Feasibility was demonstrated through supine retinal imaging of healthy adults, congenital cataract children, and infants, achieving success rates of 100%, 97%, and 95%, respectively. The system captured abnormal retinal features (e.g., absent foveal structure) in congenital cataract patients, highlighting its clinical value for monitoring retinal development. High-speed (200 kHz) imaging and high-resolution (4.1 μm) further support its dual role in clinical diagnosis and scientific research, such as retinal development studies and visual prognosis modeling. This system demonstrates significant potential for routine use in clinical practice and research, offering a reliable tool for pediatric ophthalmic imaging.