Comparative study of deep neural networks with unsupervised Noise2Noise strategy for noise reduction of optical coherence tomography images

As a powerful diagnostic tool, optical coherence tomography (OCT) has been widely used in various clinical setting. However, OCT images are susceptible to inherent speckle noise that may contaminate subtle structure information, due to low-coherence interferometric imaging procedure. Many supervised learning-based models have achieved impressive performance in reducing speckle noise of OCT images trained with a large number of noisy-clean paired OCT images, which are not commonly feasible in clinical practice. In this article, we conducted a comparative study to investigate the denoising performance of OCT images over different deep neural networks through an unsupervised Noise2Noise (N2N) strategy, which only trained with noisy OCT samples. Four representative network architectures including U-shaped model, multi-information stream model, straight-information stream model and GAN-based model were investigated on an OCT image dataset acquired from healthy human eyes. The results demonstrated all four unsupervised N2N models offered denoised OCT images with a performance comparable with that of supervised learning models, illustrating the effectiveness of unsupervised N2N models in denoising OCT images. Furthermore, U-shaped models and GAN-based models using UNet network as generator are two preferred and suitable architectures for reducing speckle noise of OCT images and preserving fine structure information of retinal layers under unsupervised N2N circumstances.     

N2NSR-OCT: Simultaneous denoising and super-resolution in optical coherence tomography images using semisupervised deep learning

Optical coherence tomography (OCT) imaging shows a significant potential in clinical routines due to its noninvasive property. However, the quality of OCT images is generally limited by inherent speckle noise of OCT imaging and low sampling rate. To obtain high signal-to-noise ratio (SNR) and high-resolution (HR) OCT images within a short scanning time, we presented a learning-based method to recover high-quality OCT images from noisy and low-resolution OCT images. We proposed a semisupervised learning approach named N2NSR-OCT, to generate denoised and super-resolved OCT images simultaneously using up- and down-sampling networks (U-Net (Semi) and DBPN (Semi)). Additionally, two different super-resolution and denoising models with different upscale factors (2× and 4× ) were trained to recover the high-quality OCT image of the corresponding down-sampling rates. The new semisupervised learning approach is able to achieve results comparable with those of supervised learning using up- and down-sampling networks, and can produce better performance than other related state-of-the-art methods in the aspects of maintaining subtle fine retinal structures.     

A deep‐learning‐based approach for noise reduction in high‐speed optical coherence Doppler tomography

Optical coherence Doppler tomography (ODT) increasingly attracts attention because of its unprecedented advantages with respect to high contrast, capillary‐level resolution and flow speed quantification. However, the trade‐off between the signal‐to‐noise ratio of ODT images and A‐scan sampling density significantly slows down the imaging speed, constraining its clinical applications. To accelerate ODT imaging, a deep‐learning‐based approach is proposed to suppress the overwhelming phase noise from low‐sampling density. To handle the issue of limited paired training datasets, a generative adversarial network is performed to implicitly learn the distribution underlying Doppler phase noise and to generate the synthetic data. Then a 3D based convolutional neural network is trained and applied for the image denoising. We demonstrate this approach outperforms traditional denoise methods in noise reduction and image details preservation, enabling high speed ODT imaging with low A‐scan sampling density.     

Quantitative assessment of optical coherence tomography angiography algorithms for neuroimaging

Optical coherence tomography (OCT) angiography can noninvasively map microvascular networks and quantify blood flow in a cerebral cortex with a resolution of 1 to 10 μm and a penetration depth of 2 to 3 mm incorporating OCT signals and angiography algorithms. Different angiography algorithms have been developed in recent years; however, the performance of the algorithms has not been assessed quantitatively for neuroimaging applications. In this paper, we developed four metrics including vascular connectivity, contrast‐to‐noise ratio, signal‐to‐noise ratio and processing time to quantitatively assess the performance of OCT angiography algorithms in image quality and computation speed. After the imaging of a rat cortex using an OCT system, the cerebral microvascular networks were visualized by seven algorithms, and the performance of the algorithms was quantified and compared. Quantitative performance assessment of the algorithms can provide suggestions for the selection of appropriate OCT angiography algorithms in neuroimaging.     

Jones matrix‐based speckle‐decorrelation angiography using polarization‐sensitive optical coherence tomography

We show that polarization‐sensitive optical coherence tomography angiography (PS‐OCTA) based on full Jones matrix assessment of speckle decorrelation offers improved contrast and depth of vessel imaging over conventional OCTA. We determine how best to combine the individual Jones matrix elements and compare the resulting image quality to that of a conventional OCT scanner by co‐locating and imaging the same skin locations with closely matched scanning setups. Vessel projection images from finger and forearm skin demonstrate the benefits of Jones matrix‐based PS‐OCTA. Our study provides a promising starting point and a useful reference for future pre‐clinical and clinical applications of Jones matrix‐based PS‐OCTA.     

Long‐term and in vivo assessment of Aβ protein‐induced brain atrophy in a zebrafish model by optical coherence tomography

In this study, a neurotoxicity model of zebrafish induced by amyloid beta (Aβ) protein was developed and evaluated in vivo by optical coherence tomography (OCT). Aβ protein and phosphate buffer saline (PBS) were separately injected into the head of two groups of adult zebrafish (n = 6 per group). Congo‐red staining results confirmed that Aβ protein had penetrated into brain tissue. All zebrafish were imaged with OCT on the 0th, 5th, 10th, 15th and 20th day postinjection. OCT images showed that PBS is not toxic to brain tissue. However, significant brain atrophy could be seen in the OCT images of zebrafish injected with Aβ‐protein that was verified by histological consequences. In addition, zebrafish in the model group showed memory decline in behavioral tests. This study verified the feasibility of in vivo long‐term assessment of Aβ protein‐induced brain atrophy in adult zebrafish by OCT that has great potential to be applied in the neurological diseases research.     

An automatic diagnostic system of coronary artery lesions in Kawasaki disease using intravascular optical coherence tomography imaging

Intravascular optical coherence tomography (IV‐OCT) is a light‐based imaging modality with high resolution, which employs near‐infrared light to provide tomographic intracoronary images. Morbidity caused by coronary heart disease is a substantial cause of acute coronary syndrome and sudden cardiac death. The most common intracoronay complications caused by coronary artery disease are intimal hyperplasia, calcification, fibrosis, neovascularization and macrophage accumulation, which require efficient prevention strategies. OCT can provide discriminative information of the intracoronary tissues, which can be used to train a robust fully automatic tissue characterization model based on deep learning. In this study, we aimed to design a diagnostic model of coronary artery lesions. Particularly, we trained a random forest using convolutional neural network features to distinguish between normal and diseased arterial wall structure. Then, based on the arterial wall structure, fully convolutional network is designed to extract the tissue layers in normal cases, and pathological tissues regardless of lesion type in pathological cases. Then, the type of the lesions can be characterized with high precision using our previous model. The results demonstrate the robustness of the model with the approximate overall accuracy up to 90%.      

Preclinical mouse model of optical coherence tomography for subcortical brain imaging without dissection

The purpose of this study was to investigate the feasibility of using optical coherence tomography (OCT) to identify internal brain lesions, specifically intracerebral hemorrhage, without dissection. Mice with artificially injected brain hematomas were used to test the OCT system, and the recorded images were compared with microscopic images of the same mouse brains after hematoxylin and eosin staining. The intracranial structures surrounding the hematomas were clearly visualized by the OCT system without dissection. These images reflect the ability of OCT to determine the extent of a lesion in several planes. OCT is a useful technology, and these findings could be used as a starting point for future research in intraoperative imaging.     

Liver tissue classification of en face images by fractal dimension‐based support vector machine

Full‐field optical coherence tomography (FF‐OCT) has been reported with its label‐free subcellular imaging performance. To realize quantitive cancer detection, the support vector machine model of classifying normal and cancerous human liver tissue is proposed with en face tomographic images. Twenty samples (10 normal and 10 cancerous) were operated from humans and composed of 285 en face tomographic images. Six histogram features and one proposed fractal dimension parameter that reveal the refractive index inhomogeneities of tissue were extracted and made up the training set. The other different 16 samples (8 normal and 8 cancerous) were imaged (190 images) and employed as the test set with the same features. First, a subcellular‐resolution tomographic image library for four histopathological areas in liver tissue was established. Second, the area under the receiver operating characteristics of 0.9378, 0.9858, 0.9391, 0.9517 for prediction of the cancerous hepatic cell, central vein, fibrosis, and portal vein were measured with the test set. The results indicate that the proposed classifier from FF‐OCT images shows promise as a label‐free assessment of quantified tumor detection, suggesting the fractal dimension‐based classifier could aid clinicians in detecting tumor boundaries for resection in surgery in the future.     

Phase unwrapping for Doppler spectral domain optical coherence tomography flow measurement

Doppler optical coherence tomography (OCT) offers additional flow velocity information, which extends the application of OCT. Phase wrapping is the inherent problem that limits measureable range of Doppler OCT. We propose a phase unwrapping method which is suitable for correcting phase in Doppler OCT images. Points (pixels) in flow region are divided into groups according to the radial distance. Points in the same group are supposed to have close velocity. Phase unwrapping algorithm begins at the boundary layer group and is performed sequentially toward the center. Using the proposed criterion, points in a group are separated into two categories, signal points and noise points. Wrapping rounds are determined for signal points phase unwrapping. Mean value of the corrected signal points replaces the noise points for noise reduction. The method is validated with capillary tube flow phantom and in vivo blood flow.     

Diseased thyroid tissue classification in OCT images using deep learning: Towards surgical decision support

Intraoperative guidance tools for thyroid surgery based on optical coherence tomography (OCT) could aid distinguish between normal and diseased tissue. However, OCT images are difficult to interpret, thus, real-time automatic analysis could support the clinical decision-making. In this study, several deep learning models were investigated for thyroid disease classification on 2D and 3D OCT data obtained from ex vivo specimens of 22 patients undergoing surgery and diagnosed with several thyroid pathologies. Additionally, two open-access datasets were used to evaluate the custom models. On the thyroid dataset, the best performance was achieved by the 3D vision transformer model with a Matthew's correlation coefficient (MCC) of 0.79 (accuracy = 0.90) for the normal-versus-abnormal classification. On the open-access datasets, the custom models achieved the best performance (MCC > 0.88, accuracy > 0.96). Results obtained for the normal-versus-abnormal classification suggest OCT, complemented with deep learning-based analysis, as a tool for real-time automatic diseased tissue identification in thyroid surgery.     

The use of optical coherence tomography and convolutional neural networks to distinguish normal and abnormal oral mucosa

Incomplete surgical resection of head and neck squamous cell carcinoma (HNSCC) is the most common cause of local HNSCC recurrence. Currently, surgeons rely on preoperative imaging, direct visualization, palpation and frozen section to determine the extent of tissue resection. It has been demonstrated that optical coherence tomography (OCT), a minimally invasive, nonionizing near infrared mesoscopic imaging modality can resolve subsurface differences between normal and abnormal head and neck mucosa. Previous work has utilized two‐dimensional OCT imaging which is limited to the evaluation of small regions of interest generated frame by frame. OCT technology is capable of performing rapid volumetric imaging, but the capacity and expertise to analyze this massive amount of image data is lacking. In this study, we evaluate the ability of a retrained convolutional neural network to classify three‐dimensional OCT images of head and neck mucosa to differentiate normal and abnormal tissues with sensitivity and specificity of 100% and 70%, respectively. This method has the potential to serve as a real‐time analytic tool in the assessment of surgical margins.     

Three‐dimensional mapping of the attenuation coefficient in optical coherence tomography to enhance breast tissue microarchitecture contrast

Effective intraoperative tumor margin assessment is needed to reduce re‐excision rates in breast‐conserving surgery (BCS). Mapping the attenuation coefficient in optical coherence tomography (OCT) throughout a sample to create an image (attenuation imaging) is one promising approach. For the first time, three‐dimensional OCT attenuation imaging of human breast tissue microarchitecture using a wide‐field (up to ~45 × 45 × 3.5 mm) imaging system is demonstrated. Representative results from three mastectomy and one BCS specimen (from 31 specimens) are presented with co‐registered postoperative histology. Attenuation imaging is shown to provide substantially improved contrast over OCT, delineating nuanced features within tumors (including necrosis and variations in tumor cell density and growth patterns) and benign features (such as sclerosing adenosis). Additionally, quantitative micro‐elastography (QME) images presented alongside OCT and attenuation images show that these techniques provide complementary contrast, suggesting that multimodal imaging could increase tissue identification accuracy and potentially improve tumor margin assessment.     

Interplay of temperature,thermal‐stresses and strains in laser‐assisted modification of collagenous tissues: Speckle‐contrast and OCT‐based studies

Moderate heating of collagenous tissues such as cartilage and cornea by infrared laser irradiation can produce biologically nondestructive structural rearrangements and relaxation of internal stresses resulting in the tissue reshaping. The reshaping results and eventual changes in optical and biological properties of the tissue strongly depend on the laser‐irradiation regime. Here, a speckle‐contrast technique based on monochromatic illumination of the tissue in combination with strain mapping by means of optical coherence elastography (OCE) is applied to reveal the interplay between the temperature and thermal stress fields producing tissue modifications. The speckle‐based technique ensured en face visualization of cross correlation and contrast of speckle images, with evolving proportions between contributions of temperature increase and thermal‐stresses determined by temperature gradients. The speckle‐technique findings are corroborated by quantitative OCE‐based depth‐resolved imaging of irradiation‐induced strain‐evolution. The revealed relationships can be used for real‐time control of the reshaping procedures (e.g., for laser shaping of cartilaginous implants in otolaryngology and maxillofacial surgery) and optimization of the laser‐irradiation regimes to ensure the desired reshaping using lower and biologically safer temperatures. The figure of waterfall OCE‐image demonstrates how the strain‐rate maximum arising in the heating‐beam center gradually splits and drifts towards the zones of maximal thermal stresses located at the temperature‐profile slopes.     

Cross‐scanning optical coherence tomography angiography for eye motion correction

We propose a cross‐scanning optical coherence tomography (CS‐OCT) system to correct eye motion artifacts in OCT angiography images. This system employs a dual‐illumination configuration with two orthogonally polarized beams, each of which simultaneously perform raster scanning in perpendicular direction with each other over the same area. In the reference arm, a polarization delay unit is used to acquire the two orthogonally polarized interferograms with a single photo detector by introducing different optical delay lines. The two cross‐scanned volume data are affected by the same eye motion but in two orthogonal directions. We developed a motion correction algorithm, which removes artifacts in the slow axis of each angiogram using the other and merges them through a nonrigid registration algorithm. In this manner, we obtained a motion‐corrected angiogram within a single volume scanning time without additional eye‐tracking devices.     

Retinal imaging with optical coherence tomography and low‐loss adaptive optics using a 2.8‐mm beam size

As data acquisition for retinal imaging with optical coherence tomography (OCT) becomes faster, efficient collection of photons becomes more important to maintain image quality. One approach is to use a larger aperture at the eye's pupil to collect more photons that have been reflected from the retina. A 2.8‐mm beam diameter system with only seven reflecting surfaces was developed for low‐loss retinal imaging. The larger beam size requires defocus and astigmatism correction, which was done in a closed loop adaptive optics method using a Shack‐Hartmann wavefront sensor and a deformable mirror (DM) with 140 actuators and a ±2.75 μm stroke. This DM facilitates defocus correction ranging from approximately ?3 D to +3 D. Comparing the new system with a standard 1.2‐mm system on a model eye, a signal‐to‐noise gain of 4.5 dB and a 2.3 times smaller speckle size were measured. Measurements on the retinas of five subjects showed even better results, with increases in dynamic range up to 13 dB. Note that the new sample arm only occupies 30 cm × 60 cm, which makes it highly suitable for imaging in a clinical environment. Figure: B‐scan images obtained over a width of 8 deg from the right eye of a 31‐year‐old Caucasian male. While the left side was imaged with a standard 1.2‐mm OCT system, the right side was imaged with the 2.8‐mm system. Both images were collected with the same integration time and incident power, after correction of aberrations. Using the dynamic range within the images, which is determined by comparing the highest pixel value to the noise floor, a difference in dynamic range of 10.8 dB was measured between the two systems.      

Comparison of the Iowa Reference Algorithm to the Heidelberg Spectralis optical coherence tomography segmentation algorithm

For spectral‐domain optical coherence tomography (SD‐OCT) studies of neurodegeneration, it is important to understand how segmentation algorithms differ in retinal layer thickness measurements, segmentation error locations and the impact of manual correction. Using macular SD‐OCT images of frontotemporal degeneration patients and controls, we compare the individual and aggregate retinal layer thickness measurements provided by two commonly used algorithms, the Iowa Reference Algorithm and Heidelberg Spectralis, with manual correction of significant segmentation errors. We demonstrate small differences of most retinal layer thickness measurements between these algorithms. Outer sectors of the Early Treatment Diabetic Retinopathy Study grid require a greater percent of eyes to be corrected than inner sectors of the retinal nerve fiber layer (RNFL). Manual corrections affect thickness measurements mildly, resulting in at most a 5% change in RNFL thickness. Our findings can inform researchers how to best use different segmentation algorithms when comparing retinal layer thicknesses.     

Contrast‐enhanced optical coherence tomography for melanoma detection: An in vitro study

Optical coherence tomography (OCT), with a high‐spatial resolution (<10 microns), intermediate penetration depth (~1.5 mm) and volumetric imaging capability is a great candidate to be used as a diagnostic‐assistant modality in dermatology. At this time, the accuracy of OCT for melanoma detection is lower than anticipated. In this letter, we studied for the first time, the use of a novel contrast agent consist of ultra‐small nanoparticles conjugated to a melanoma biomarker to improve the accuracy of OCT for differentiation of melanoma cells from nonmelanoma cells, in vitro. We call this approach SMall nanoparticle Aggregation‐enhanced Radiomics of Tumor (SMART)‐OCT imaging. This initial proof of concept study is the first step toward the broad utilization of this method for high accuracy all types of tumor detection applications.     

Cross-domain retinopathy classification with optical coherence tomography images via a novel deep domain adaptation method

Deep learning based retinopathy classification with optical coherence tomography (OCT) images has recently attracted great attention. However, existing deep learning methods fail to work well when training and testing datasets are different due to the general issue of domain shift between datasets caused by different collection devices, subjects, imaging parameters, etc. To address this practical and challenging issue, we propose a novel deep domain adaptation (DDA) method to train a model on a labeled dataset and adapt it to an unlabelled dataset (collected under different conditions). It consists of two modules for domain alignment, that is, adversarial learning and entropy minimization. We conduct extensive experiments on three public datasets to evaluate the performance of the proposed method. The results indicate that there are large domain shifts between datasets, resulting a poor performance for conventional deep learning methods. The proposed DDA method can significantly outperform existing methods for retinopathy classification with OCT images. It achieves retinopathy classification accuracies of 0.915, 0.959 and 0.990 under three cross-domain (cross-dataset) scenarios. Moreover, it obtains a comparable performance with human experts on a dataset where no labeled data in this dataset have been used to train the proposed DDA method. We have also visualized the learnt features by using the t-distributed stochastic neighbor embedding (t-SNE) technique. The results demonstrate that the proposed method can learn discriminative features for retinopathy classification.     

LV-GAN: A deep learning approach for limited-view optoacoustic imaging based on hybrid datasets

The optoacoustic imaging (OAI) methods are rapidly evolving for resolving optical contrast in medical imaging applications. In practice, measurement strategies are commonly implemented under limited-view conditions due to oversized image objectives or system design limitations. Data acquired by limited-view detection may impart artifacts and distortions in reconstructed optoacoustic (OA) images. We propose a hybrid data-driven deep learning approach based on generative adversarial network (GAN), termed as LV-GAN, to efficiently recover high quality images from limited-view OA images. Trained on both simulation and experiment data, LV-GAN is found capable of achieving high recovery accuracy even under limited detection angles less than 60^°. The feasibility of LV-GAN for artifact removal in biological applications was validated by ex vivo experiments based on two different OAI systems, suggesting high potential of a ubiquitous use of LV-GAN to optimize image quality or system design for different scanners and application scenarios.