Recent advances in deep-learning-enhanced photoacoustic imaging

Advanced Photonics Nexus, Journal Year: 2023, Volume and Issue: 2(05)

Published: July 24, 2023

Photoacoustic imaging (PAI), recognized as a promising biomedical modality for preclinical and clinical studies, uniquely combines the advantages of optical ultrasound imaging. Despite PAI's great potential to provide valuable biological information, its wide application has been hindered by technical limitations, such hardware restrictions or lack biometric information required image reconstruction. We first analyze limitations PAI categorize them seven key challenges: limited detection, low-dosage light delivery, inaccurate quantification, numerical reconstruction, tissue heterogeneity, imperfect segmentation/classification, others. Then, because deep learning (DL) increasingly demonstrated ability overcome physical modalities, we review DL studies from past five years that address each challenges in PAI. Finally, discuss promise future research directions DL-enhanced

Language: Английский

---

Photoacoustic imaging plus X: a review

Journal of Biomedical Optics, Journal Year: 2023, Volume and Issue: 29(S1)

Published: Dec. 28, 2023

SignificancePhotoacoustic (PA) imaging (PAI) represents an emerging modality within the realm of biomedical technology. It seamlessly blends wealth optical contrast with remarkable depth penetration offered by ultrasound. These distinctive features PAI hold tremendous potential for various applications, including early cancer detection, functional imaging, hybrid monitoring ablation therapy, and providing guidance during surgical procedures. The synergy between other cutting-edge technologies not only enhances its capabilities but also propels it toward broader clinical applicability.AimThe integration advanced technology PA signal processing, image reconstruction, applications has significantly bolstered PAI. This review endeavor contributes to a deeper comprehension how can lead improved applications.ApproachAn examination evolving research frontiers in PAI, integrated technologies, reveals six key categories named "PAI plus X." encompass range topics, limited treatment, circuits design, accurate positioning system, fast scanning systems, ultrasound sensors, laser sources, deep learning, modalities.ResultsAfter conducting comprehensive existing literature on proposals have emerged advance development X. aim enhance system hardware, improve quality, address challenges effectively.ConclusionsThe progression innovative sophisticated approaches each category X is positioned drive significant advancements both applications. Furthermore, integrate above-mentioned broaden even further.

Language: Английский

---

Unsupervised denoising of photoacoustic images based on the Noise2Noise network

Biomedical Optics Express, Journal Year: 2024, Volume and Issue: 15(8), P. 4390 - 4390

Published: June 17, 2024

In this study, we implemented an unsupervised deep learning method, the Noise2Noise network, for improvement of linear-array-based photoacoustic (PA) imaging. Unlike supervised learning, which requires a noise-free ground truth, network can learn noise patterns from pair noisy images. This is particularly important in vivo PA imaging, where truth not available. developed method to generate pairs single set images and verified our approach through simulation experimental studies. Our results reveal that effectively remove noise, improve signal-to-noise ratio, enhance vascular structures at deeper depths. The denoised show clear detailed structure different depths, providing valuable insights preclinical research potential clinical applications.

Language: Английский

---

Development of a Software and Hardware Complex for Monitoring Processes in Production Systems

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1527 - 1527

Published: Feb. 28, 2025

The article presents a detailed exposition of hardware–software complex that has been developed for the purpose enhancing productivity accounting state production process. This facilitates automation identification parts in containers and utilisation supplementary markers. comprises mini computer (system unit industrial version) with connected cameras (IP or WEB), communication module LED signal lamps, software. cascade algorithm detection labels objects employs trained convolutional neural networks (YOLO VGG19), thereby recognition accuracy while concurrently reducing size training sample networks. efficacy system was assessed through laboratory experimentation, which yielded experimental results demonstrating 93% detail using algorithm, comparison to 72% achieved traditional approach employing single network.

Language: Английский

---

基于时域数据的深度学习光声内窥镜降噪算法

Acta Optica Sinica, Journal Year: 2025, Volume and Issue: 45(3), P. 0317001 - 0317001

Published: Jan. 1, 2025

---

Bessel Beams in Ophthalmology: A Review

Micromachines, Journal Year: 2023, Volume and Issue: 14(9), P. 1672 - 1672

Published: Aug. 27, 2023

The achievable resolution of a conventional imaging system is inevitably limited due to diffraction. Dealing with precise in scattering media, such as the case biomedical imaging, even more difficult owing weak signal-to-noise ratios. Recent developments non-diffractive beams Bessel beams, Airy vortex and Mathieu have paved way tackle some these challenges. This review specifically focuses on for ophthalmological applications. theoretical foundation beam discussed first followed by various applications utilizing beams. advantages disadvantages techniques comparison those existing state-of-the-art systems are discussed. concludes an overview current future perspectives ophthalmology.

Language: Английский

---

Singular value decomposition with weighting matrix applied for optical-resolution photoacoustic microscopes

Ultrasonics, Journal Year: 2024, Volume and Issue: 143, P. 107424 - 107424

Published: July 27, 2024

The prestige target selectivity and imaging depth of optical-resolution photoacoustic microscope (OR-PAM) have gained attentions to enable advanced intra-cellular visualizations. However, the broad-band nature signals is prone noise artifacts caused by inefficient light-to-pressure translation, resulting in poor image quality. present study foresees application singular value decomposition (SVD) effectively extract from these artifacts. Although spatiotemporal SVD succeeded ultrasound flow signal extraction, conventional multi frame model not suitable for data acquired with scanning OR-PAM due burden accessing multiple frames. To utilize on OR-PAM, this began exploring applied A-lines instead Upon explorations, an obstacle uncertain presence unwanted vectors was observed. tackle this, a data-driven weighting matrix designed relevant based analyses temporal-spatial vectors. Evaluation extraction capability showed superior quality efficient computation against past studies. In summary, contributes field providing exploration A-line as well its practical utilization distinguish recover artifact components.

Language: Английский

---

Decoupling thermal effects in GaN photodetectors for accurate measurement of ultraviolet intensity using deep neural network

Engineering Applications of Artificial Intelligence, Journal Year: 2023, Volume and Issue: 123, P. 106309 - 106309

Published: April 25, 2023

Language: Английский

---

Intelligent skin‐removal photoacoustic computed tomography for human based on deep learning

Journal of Biophotonics, Journal Year: 2024, Volume and Issue: 17(10)

Published: Aug. 2, 2024

Photoacoustic computed tomography (PACT) has centimeter-level imaging ability and can be used to detect the human body. However, strong photoacoustic signals from skin cover deep tissue information, hindering frontal display analysis of images regions interest. Therefore, we propose a 2.5 D learning model based on feature pyramid structure single-type annotation extract region, design mask generation algorithm remove automatically. PACT experiments periphery blood vessel verified correctness our proposed skin-removal method. Compared with previous studies, method exhibits high robustness uneven illumination, irregular boundary, reconstruction artifacts in images, errors decreased by 20% ~ 90% 1.65 dB improvement signal-to-noise ratio at same time. This study may provide promising way for high-definition tissues.

Language: Английский

---

A Review of Application of Deep Learning in Endoscopic Image Processing

Journal of Imaging, Journal Year: 2024, Volume and Issue: 10(11), P. 275 - 275

Published: Nov. 1, 2024

Deep learning, particularly convolutional neural networks (CNNs), has revolutionized endoscopic image processing, significantly enhancing the efficiency and accuracy of disease diagnosis through its exceptional ability to extract features classify complex patterns. This technology automates medical analysis, alleviating workload physicians enabling a more focused personalized approach patient care. However, despite these remarkable achievements, there are still opportunities further optimize deep learning models for including addressing limitations such as requirement large annotated datasets challenge achieving higher diagnostic precision, rare or subtle pathologies. review comprehensively examines profound impact on highlighting current strengths limitations. It also explores potential future directions research development, outlining strategies overcome existing challenges facilitate integration into clinical practice. Ultimately, goal is contribute ongoing advancement imaging technologies, leading accurate, personalized, optimized care patients.

Language: Английский

---