Accurate measurement of anterior segment parameters is crucial for diagnosing and managing ophthalmic conditions, such as glaucoma, cataracts, and refractive errors. However, traditional clinical measurement methods are often time-consuming, labor-intensive, and susceptible to inaccuracies. With the growing potential of artificial intelligence in ophthalmic diagnostics, this study aims to develop and evaluate a deep learning model capable of automatically extracting key points and precisely measuring multiple clinically significant anterior segment parameters from ultrasound biomicroscopy (UBM) images. These parameters include central corneal thickness (CCT), anterior chamber depth (ACD), pupil diameter (PD), angle-to-angle distance (ATA), sulcus-to-sulcus distance (STS), lens thickness (LT), and crystalline lens rise (CLR). A data set of 716 UBM anterior segment images was collected from Tianjin Medical University Eye Hospital. YOLOv8 was utilized to segment four key anatomical structures: cornea–sclera, anterior chamber, pupil, and iris–ciliary body—thereby enhancing the accuracy of keypoint localization. Only images with intact posterior capsule lentis were selected to create an effective data set for parameter measurement. Ten keypoints were localized across the data set, allowing the calculation of seven essential parameters. Control experiments were conducted to evaluate the impact of segmentation on measurement accuracy, with model predictions compared against clinical gold standards. The segmentation model achieved a mean IoU of 0.8836 and mPA of 0.9795. Following segmentation, the binary classification model attained an mAP of 0.9719, with a precision of 0.9260 and a recall of 0.9615. Keypoint localization exhibited a Euclidean distance error of 58.73 ± 63.04 μm, improving from the pre-segmentation error of 71.57 ± 67.36 μm. Localization mAP was 0.9826, with a precision of 0.9699, a recall of 0.9642 and an FPS of 32.64. In addition, parameter error analysis and Bland–Altman plots demonstrated improved agreement with clinical gold standards after segmentation. This deep learning approach for UBM image segmentation, keypoint localization, and parameter measurement is feasible, enhancing clinical diagnostic efficiency for anterior segment parameters.

Background:Ultrasound biomicroscopy (UBM) enables high-resolution imaging of the anterior segment, essential for accurate diagnosis and anatomical assessment. However, the quality evaluation of UBM currently relies on subjective judgment, which is time-consuming and inconsistent. This study proposes a deep learning (DL)-based no-reference method for objective and automated UBM image quality assessment (IQA), facilitating reliable selection of high-quality images for clinical use. Methods:A total of 1,154 clinical panoramic UBM images of the anterior segment were collected from Tianjin Eye Hospital. The YOLOv8s_DW_FOCUS model was employed to accurately extract the region of interest (ROI) and identify five key anatomical landmarks: the central corneal epithelium, central corneal endothelium, posterior lens capsule, left ciliary groove, and right ciliary groove. In collaboration with clinical ophthalmologists, eight key criteria for assessing anterior segment UBM image quality were established, integrating general medical image evaluation parameters and ophthalmic expertise. Based on these criteria, each frame was assigned a quality score. Images scoring 7 or higher were classified as high-quality, whereas those receiving a perfect score of 8 were considered standard. To validate the feasibility of our method, we conducted rigorous evaluations of its accuracy, focus on key regions, generalization capability, inter- and intra-class discrimination, and consistency in assessment results. Results:The target detection model achieved a mean average precision (mAP) of 0.935, a recall of 0.898, and a precision of 0.925. Additionally, it effectively focused on key regions, as demonstrated by the heatmap analysis. The t-distributed stochastic neighbor embedding (t-SNE) plot further highlighted the model's strong discriminative capability across different classes and its excellent generalization performance. To assess the consistency between our no-reference quality assessment method and expert evaluations, we analyzed 174 standard images that had been subjectively selected by clinical ophthalmologists. Among them, 146 images received a score of 8, whereas 26 images scored 7, indicating a high level of agreement with clinical experts in identifying high-quality images. Moreover, our method applies stricter criteria for defining standard images, enabling a more precise selection of high-quality anterior segment UBM images. Conclusions:The DL-based no-reference quality assessment method proposed in this study provides an objective evaluation of anterior segment UBM image quality. It effectively identifies high-quality images, significantly improving the efficiency of ophthalmic imaging professionals and demonstrating strong clinical potential for widespread adoption.

Tumor organoid-based drug sensitivity prediction is a new approach for precision medicine, which has wide applications in cancer treatment and attracts increasing attention. In the field of breast cancer, conventional organoid culture methods often require more than three weeks of culture period. The culture time greatly limits the further extension of the application scenarios of breast cancer organoids. We developed a fluid system that builds on the conventional organoid “dome” culture method, which continuously and stably supplies the nutrients for the growth of breast cancer organoids. We demonstrated that this is an effective optimization method, which can shorten the culture period of breast cancer organoids without significant changes in histological characteristics and drug sensitivity features.

Objectives To explore clinical value of miRNA‐18a, miRNA‐130a, and miRNA‐92a combined with transvaginal color Doppler ultrasound (TVCDS) in the diagnosis of cervical cancer (CC). Methods One hundred patients with pathologically confirmed CC (CC group), 100 patients with cervical epithelial neoplasia (disease group), and 100 patients with benign uterine lesions (control group) were selected. TVCDS was performed, and the levels of serum miRNA‐18a, miRNA‐130a, and miRNA‐92a were detected. Results The systolic blood velocity of the cancer group, the disease group, and the control group decreased sequentially, while the resistance index and pulsatility index increased sequentially. The serum miR‐18a, miR‐130a, miR‐92a, and expression levels of the patients' increased sequentially. Multivariate logistic regression analysis showed that age, high‐risk human papillomavirus (HR‐HPV) infection, menopause, blood flow RI, serum miRNA‐18a, miRNA‐92a, and miRNA‐130a were the influencing factors of CC. The receiver operating characteristic curve showed that the sensitivity, specificity, accuracy, and area under curve of transvaginal Doppler ultrasound in the diagnosis of CC were 86.43%, 88.01%, 84.32%, and 0.913; serum miR‐18a were, respectively, 76.56, 81.30, 80.36, and 0.839; serum miR‐130a were 77.88%, 76.97%, 78.32%, and 0.0.864; serum miR‐92a were 71.04%, 80.39%, 80.74%, and 0.894; 90.33%, 95.14%, 91.25%, and 0.947, the area under curve of the combined detection of the 3 was greater than that of the single detection. Conclusions Serum miRNA combined with TVCDS has the advantages of it being noninvasive, and having high sensitivity and high specificity in the diagnosis of CC.

Objective: This study aims to evaluate the impact of image factors on the performance of deep learning models used for ophthalmic ultrasound image diagnosis. Methods: A total of 3,373 ophthalmic ultrasound images are used to deeply evaluate the influence of image factors on the performance of deep learning classification models. Inceptionv3, Xception, and the fusion model Inceptionv3-Xception are used to explore how brightness, contrast, gain, noise, size, format, pseudo-color seven image-related factors affect the classification performance of the model. Results: Inceptionv3-Xception has advantages in the recognition accuracy of various image factors. When the image brightness changes, the model's performance shows a downward trend (0.5 vs. 1 vs. 1.8, ACC 95.73 vs. 97.06 vs. 93.54, P < 0.05). When the image contrast changes, the model's performance is comparable (0.5 vs. 1 vs. 1.2, ACC 96.23 vs. 96.95 vs. 97.45, P > 0.05). When the image gain drops to 50 dB, the model's accuracy decreases significantly (50 dB vs. 105 dB, ACC 96.49 vs. 97.57, P < 0.05). When Gaussian noise is added to the image, the model's performance gradually decreases (0.02 vs. 0, ACC 89.48vs97.06, P < 0.05). When the image size drops to 25% of the original image, the model's performance decreases significantly (25% vs. 100%, ACC 93.18 vs. 97.06, P < 0.01). When the image format changes, the model's recognition accuracy is similar (JPG vs. BMP vs. PNG, ACC 96.98 vs. 97.06 vs. 97.06, P > 0.05). The accuracy of the model in recognizing pseudo-color images decreases significantly compared to grayscale images (grayscale vs. pseudo-color, ACC 35.96 vs. 97.06). Conclusion: These results indicate that image quality greatly influences the model training process, and acquiring high-quality images is an important prerequisite for high recognition performance of the model. This study offers valuable insights for the improvement of other robust deep learning models for ophthalmic ultrasound image recognition.

Purpose: To evaluate and compare the accuracy of iTrace and CASIA2 in measuring the postoperative orientation of toric intraocular lens (IOL) without mydriasis. Setting: Tianjin Medical University Eye Hospital, Tianjin, China. Design: Prospective cohort study. Methods: Patients with SN6AT toric IOLs implanted after cataract surgery were enrolled. 1 month after surgery, the toric IOL orientation were measured by iTrace and CASIA2 in non-mydriatic, semi-dark conditions. Then, the toric axis was directly reviewed using the slit-lamp under full mydriasis. Axis measurement differences between each of the 2 devices and the slit-lamp, described as their relative differences (RDs), were calculated and compared. The percentage of RDs within 5 degrees, within 10 degrees and greater than 30 degrees were analyzed. Results: 77 eyes of 70 patients were included. Generally, the mean toric axis measurement RDs of CASIA2 and iTrace were 9.24 ± 10.53 degrees and 13.89 ± 15.47 degrees respectively (P = .04). For CASIA2 (72 eyes), 54.17% (39), 72.22% (52), and 4.17% (3) of eyes had RDs within 5 degrees, within 10 degrees and greater than 30 degrees, compared with 40.00% (28), 61.43% (43) and 12.86% (9) for iTrace (70 eyes). The 95% limits of agreements of CASIA2 with slit-lamp was narrower than that of iTrace with slit-lamp. The median RD of CASIA2 was significantly smaller in eyes with pupil ≥4 mm under dark condition compared with eyes with pupil <4 mm (P = .03). Conclusions: CASIA2 demonstrates greater precision in measuring toric IOL orientation under non-mydriatic conditions compared with iTrace. Moreover, the accuracy of CASIA2 is enhanced in cases of pupil >4 mm.

Objective: The goal of the work described here was to construct a deep learning-based intelligent diagnostic model for ophthalmic ultrasound images to provide auxiliary analysis for the intelligent clinical diagnosis of posterior ocular segment diseases.Methods: The InceptionV3-Xception fusion model was established by using two pre-trained network models- InceptionV3 and Xception-in series to achieve multilevel feature extraction and fusion, and a classifier more suitable for the multiclassification recognition task of ophthalmic ultrasound images was designed to classify 3402 ophthalmic ultrasound images. The accuracy, macro-average precision, macro-average sensitivity, macro-average F1 value, subject working feature curves and area under the curve were used as model evaluation metrics, and the credibility of the model was assessed by testing the decision basis of the model using a gradient-weighted class activation mapping method.Results: The accuracy, precision, sensitivity and area under the subject working feature curve of the InceptionV3 -Xception fusion model on the test set reached 0.9673, 0.9521, 0.9528 and 0.9988, respectively. The model decision basis was consistent with the clinical diagnosis basis of the ophthalmologist, which proves that the model has good reliability. Conclusion: The deep learning-based ophthalmic ultrasound image intelligent diagnosis model can accurately screen and identify five posterior ocular segment diseases, which is beneficial to the intelligent development of ophthalmic clinical diagnosis.

Background:Anisotropy which encodes rich structure and function information is one of the key and unique characteristics of tissues. Polarized photoacoustic imaging shows tremendous potential for the detection and quantification of the anisotropy of tissues. The existing polarized photoacoustic imaging methods cannot quantify anisotropy and detect the orientation of the optical axis in 3D imaging.Methods:We proposed a versatile polarized photoacoustic imaging method based on the detection of high-order harmonics of the photoacoustic signal, which can be used for both 2D and 3D polarized photoacoustic imaging, This method can detect and quantify the anisotropy and the orientation of the optical axis of the anisotropic objects by the amplitude and initial phase of the high-order harmonics. A double-focusing polarized photoacoustic microscopy was developed to validate the proposed method. Experiments were conducted on 2D and 3D anisotropic phantoms.Results:The results showed that the anisotropy and the orientation of the optical axis of the anisotropic object can be detected and quantified accurately by the amplitude and initial phase of the high-order harmonics, even at a depth of triple transport mean free path. The imaging depth of the polarized photoacoustic microscopy is mainly limited by laser energy attenuation rather than depolarization.Conclusions:Polarized photoacoustic microscopy based on high-order harmonics has tremendous potential for imaging the anisotropy of deep biological tissues in vivo. It also extends the capability of photoacoustic microscopy to image the anisotropy of tissues.

One major method for detecting retinopathy in clinics is optical coherence tomography. However, this manual diagnostic model is affected by strong subjectivity and low efficiency. Therefore, this paper proposes a lightweight convolutional neural network for the automatic detection of retinopathy. The proposed network consists of two modules. The first module combines atrous convolutions and depth wise separable convolutions to reduce the number of parameters; the second module uses the decomposition convolution method to extend the depth by decomposing the conventional convolution layer into multilayer asymmetric convolution. Both modules are combined to form a feature extractor, and the Softmax function is used as the classifier to obtain a lightweight model with 44 layers deep and 9.2 MB parameters. The accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve of the proposed network on the test set are 0.980, 0.954, 0.987, and 0.997, respectively. The visualization results show that the diagnostic basis of the model is consistent with that of ophthalmologists. These results show that the proposed network can accurately automate retinal disease detection.

目的 研究一种新型20 MHz眼科超声扫描成像方法,可在满足临床成像分辨力和声能安全性要求的前提下,显著提高图像的探测深度,拓展20 MHz超声频段的临床应用范围.方法 通过8位Golay互补序列,激励超声换能器产生超声波.回波信息经高速采集与匹配滤波后,采用相邻正、反编码扫描线数据复用的方法交替相加,完成解码运算.在保证图像扫描线数和扫描帧频的前提下,实现实时显像.最后通过钨丝靶线和仿组织超声体模实验,验证了新的成像方法在保持原有分辨能力与扫描帧频不变的前提下,提高了图像的探测深度.结果 与传统单脉冲模式相比,Golay互补编码模式成像中轴向分辨率与侧向分辨率分别达到80μm和150μm,小信号探测深度增加约0.5 cm,图像信噪比也得到显著改善.结论 基于Golay互补序列实现20 MHz眼部组织超声成像,相对于传统成像方式可极大改善图像质量,具有很好的临床应用前景.

Background: Pilomatricoma (PM) is one of the most common benign tumours in children. However, the inaccuracy of preoperative diagnosis and evaluation is high. Non-invasive examinations, including dermoscopy and ultrasound are helpful for diagnosing and evaluating PM. To date, ultra-high-frequency ultrasonographic features of PM have been rarely studied.Objective: We aimed to investigate the ultra-high frequency ultrasonographic features of PM in a large paediatric cohort and to determine the associations of these features with the clinical features of different histological subtypes of PM.Methods: This was a retrospective study. Patients who had both preoperative ultra-high-frequency ultrasonographic evaluation and pathological diagnosis of PM were enrolled. A series of infantile haemangiomas and cutaneous cysts during the same period were included as controls. Histological findings, including the stage, calcifying type, and ultrasonographic features of each lesion, were described.Results: A total of 133 patients with PM were included, and 147 PM lesions were analysed. The male-to-female ratio was 1:1.58, and the median age of onset was 91 (range: 10–188) months. On ultra-high-frequency ultrasonography, PM presented as heterogeneous (144/147, 98.0%), well-demarcated (143/147, 97.3%), and hypoechoic (126/147, 85.7%) tumours located between the deep dermis and subcutaneous tissue (139/147, 94.6%). The most common features were internal echogenic foci (135/147, 91.8%), hypoechoic rim (133/147, 90.5%), and posterior acoustic shadowing (94/147, 63.9%). Fourteen (9.5%) lesions were histologically categorized in the early stage, 58 (39.5%) in the fully developed stage, 65 (44.2%) in the early regressive stage and 10 (6.8%) in the late regressive stage. Three calcifying types, including scattered dots, clumps and arcs, were observed on histologic examination, which corresponded well with grey-scale imaging on ultra-high-frequency ultrasonography. Each calcifying type was significantly different in various histological stages (P = 0.001), among which scattered dots were mainly present in the early and fully developed stage and arc-shaped calcifying were present in the regressive stages. Calcification was observed in skin cysts, while there was more frequent posterior enhancement, less frequent posterior shadowing, and hypoechoic rim than PM. Haemangioma also presented as a hypoechoic tumour on grey-scale imaging. However, haemangioma was homogeneous and rarely calcifying.Conclusions: PM is a heterogeneous, well-demarcated, hypoechoic tumour located between the deep dermis and the subcutis on ultra-high-frequency ultrasonography. The most common features are internal echogenic foci (calcifying) and hypoechoic rim. Calcifying types can help in the staging of PM. Ultra-high-frequency ultrasound is a useful tool for the diagnosis and evaluation of PM.

目的 基于超声生物显微镜(UBM)图像,通过卷积神经网络(CNN)建立前房角开闭状态的自动识别方法.方法 数据集为我院及所属医联体医院采集的眼科疾病患者的UBM图像.由3名眼科医师手动分为房角开放和房角关闭两类,使用改进后的CNN VGG16模型实现前房角开闭状态的自动识别.结果 该CNN模型对前房角UBM图像的开闭状态识别准确率为0.9619,灵敏度为0.9618,AUC为0.9973.结论 基于CNN模型的前房角开闭状态的自动识别方法准确可靠,为原发性闭角型青光眼的UBM影像自动诊断奠定了实验基础.

Purpose To develop and assess a deep learning system that automatically detects angle closure and quantitatively measures angle parameters from ultrasound biomicroscopy (UBM) images using a deep learning algorithm. Methods A total of 3788 UBM images (2146 open angle and 1642 angle closure) from 1483 patients were collected. We developed a convolutional neural network (CNN) based on the InceptionV3 network for automatic classification of angle closure and open angle. For nonclosed images, we developed a CNN based on the EfficienttNetB3 network for the automatic localization of the scleral spur and the angle recess; then, the Unet network was used to segment the anterior chamber angle (ACA) tissue automatically. Based on the results of the latter two processes, we developed an algorithm to automatically measure the trabecular-iris angle (TIA500 and TIA750), angle-opening distance (AOD500 and AOD750), and angle recess area (ARA500 and ARA750) for quantitative evaluation of angle width. Results Using manual labeling as the reference standard, the ACA classification network's accuracy reached 98.18%, and the sensitivity and specificity for angle closure reached 98.74% and 97.44%, respectively. The deep learning system realized the automatic measurement of the angle parameters, and the mean of differences was generally small between automatic measurement and manual measurement. The coefficients of variation of TIA500, TIA750, AOD500, AOD750, ARA500, and ARA750 measured by the deep learning system were 5.77%, 4.67%, 10.76%, 7.71%, 16.77%, and 12.70%, respectively. The within-subject standard deviations of TIA500, TIA750, AOD500, AOD750, ARA500, and ARA750 were 5.77 degrees, 4.56 degrees, 155.92 µm, 147.51 µm, 0.10 mm2, and 0.12 mm2, respectively. The intraclass correlation coefficients of all the angle parameters were greater than 0.935. Conclusions The deep learning system can effectively and accurately evaluate the ACA automatically based on fully automated analysis of a UBM image. Translational Relevance The present work suggests that the deep learning system described here could automatically detect angle closure and quantitatively measure angle parameters from UBM images and enhancing the intelligent diagnosis and management of primary angle-closure glaucoma.

In this paper, a high-speed data transmission system for high-frequency ultrasonic radio frequency echo signals based on FPGA and a single-element high-frequency broadband transducer were designed. With the FPGA as the core control processor, it controlled the transmission and reception of ultrasound and the upload of radio frequency data through the USB3.0 interface. In the interactive display software interface of upper computer designed under the development environment of visual studio 2012, the waveform preview and storage of the original RF data received by the 50 MHz single element broadband transducer were completed. Through the sawtooth wave simulation experiment and the plexiglass ultrasonic echo radio frequency signal acquisition experiment, it was concluded that the radio frequency data transmitted by the experimental system to the host computer was correct. The experimental system had reasonable design, error-free signal transmission, and high integration. The written waveform preview interface of the upper computer application software could not only meet the experimental debugging of subsequent research, but also the stored raw radio frequency data in multiple formats satisfy the subsequent further scientific research.

目的 基于深度学习(deep learning,DL)和前房角超声生物显微镜(ultrasound biomicroscopy,UBM)图像进行前房角开闭状态的自动识别,为原发性闭角型青光眼的临床自动诊断提供辅助分析.方法 数据集为天津医科大学眼科医院采集的眼科疾病患者的前房角UBM图像,由眼科专家将UBM图像分为房角开放和房角关闭两类,按照6:2:2的比例随机设置训练集、验证集和测试集.为提高深度学习模型的鲁棒性和识别精度,对训练集图像随机进行了旋转、平移和反转等不影响房角形态的数据增强操作.比较VGG16、VGG19、DenseNet121、Xception和InceptionV3网络模型在本文数据集上的迁移学习结果,根据迁移学习结果对VGG16进行卷积层和全连接层的微调,用微调后的VGG16模型实现前房角开闭状态的自动识别.用接收者操作特征曲线下面积和准确率等评价指标对模型识别结果进行定量评价,用类激活热力图可视化模型识别前房角开闭状态时的主要关注区域.结果 类激活热力图表明微调后的VGG16模型识别前房角开闭状态的主要关注区域为房角中心区域,与眼科专家的识别依据一致.该模型的识别准确率为96.19％,接收者操作特征曲线下面积为0.9973.结论 基于深度学习和前房角UBM图像能够以较高的准确率实现前房角开闭状态的自动识别,有利于原发性闭角型青光眼自动诊断技术的发展.

High frequency ultrasonic imaging provides clinicians with high-resolution diagnostic images and more accurate measurement results. The technique is now widely used in ophthalmology, dermatology, and small animal imaging. However, since ultrasonic attenuation in tissue increases rapidly with increasing frequency, the depth of detection of high frequency ultrasound in tissue is limited to a few millimeters. In this paper, a novel method of using Golay-coded excitation as a replacement for conventional single-pulse excitation in high frequency ultrasound biomicroscopy was proposed, and real-time imaging was realized. While maintaining the transmission voltage and image resolution unchanged, the detection depth can be effectively improved. The ultrasonic transmission frequency is 30 MHz and the transmission voltage is ± 60 V p-p. In this study, 4-bit, 8-bit, and 16-bit coding sequences and decoding compression were used. To verify the effectiveness of the coding sequence in real-time imaging of ultrasound biomicroscopy, we designed a 10-μm diameter line target echo experiment, an ultrasound phantom experiment, and an in vitro porcine eye experiment. The experimental results show that the code/decode method of signal processing can not only maintain a resolution consistent with that of single-pulse transmission, but can also improve the detection depth and signal-to-noise ratio.

Purpose: The purpose of this study was to develop a convolutional neural network (CNN) for automated localization of the scleral spur in ultrasound biomicroscopy (UBM) images of open-angle eyes. Methods: UBM images were acquired, and one glaucoma specialist provided reference coordinates of scleral spur locations in all images. A CNN model based on the Efficient-NetB3 architecture was developed to detect the scleral spur in each image. The prediction errors and Euclidean distance were used to evaluate localization performance of the CNN model. Trabecular-iris angle 500 (TIA500) and angle-opening distance 500 (AOD500) were measured and analyzed using the scleral spur locations provided by the specialist and predicted by the CNN model. Results: The CNN was developed using a training dataset of 2328 images and tested using an independent dataset of 258 images. The mean absolute prediction errors of CNN model were 48.06 +/- 45.40 mu m for X-coordinates and 30.84 +/- 27.03 mu m for Y-coordinates. The mean absolute intraobserver variability was 47.80 +/- 44.45 mu m for X-coordinates and 29.50 +/- 25.77 mu m for Y-coordinates. The mean Euclidean distance of the CNN was 60.41 +/- 49.02 mu m and the intraobserver mean Euclidean distance was 59.78 +/- 47.12 mu m. The mean absolute error in TIA500 was 1.26 +/- 1.38 degrees for all test images and in AOD500 was 0.039 +/- 0.051 mm. Conclusions: A CNN can detect the scleral spur on UBM images of open-angle eyes with performance similar to that of a glaucoma specialist. Translational Relevance: Deep learning algorithms for automating scleral spur localization would facilitate the quantitative assessment of the opening of the angle and the risk in angle closure.

目的真皮和表皮可以反映人体皮肤的健康状况,是皮肤特征的一个重要评价指标。实验提出一种基于改进的U-Net网络的皮肤多类分割算法,利用U-Net深度神经网络强大的编码解码结构自动分割出人体皮肤中真皮层和表皮层区域。方法选择健康志愿者25例,其中男性15例,女性10例;年龄20～60岁,平均年龄30岁。采集部位为手背、前臂内侧、前臂外侧、面颊、额头、腹部、背部、大腿外侧、小腿内侧、脚面。利用50 MHz皮肤超声生物显微镜采集15例健康志愿者的150幅人体皮肤二维超声图像作为训练集和验证集,将另外10例健康志愿者采集得到的100幅图像作为测试集。150幅人体皮肤二维超声图像,数据来源于志愿者,皮肤状况为没有明显皮损。首先人工标注出真皮层、表皮层区域,并进行预处理操作,然后将图像输入到改进后的U-Net全卷积神经网络中进行训练,最后对测试集进行分割预测。对于U-Net网络的改进,实验提出了一种新的密集残差模块和残差空间金字塔池化模块,融合了空洞卷积结构,并加深了模型的深度。采用精确度（PRE）、灵敏度（SE）、平均交并比（MIOU）、相似性系数（DSC）、豪斯多夫距离（HD）作为评价指标。结果相比于U-Net网络,改进后的算法在皮肤图像分割中取得了良好的效果,正确识别出表皮层区域与真皮层区域,并且对于真皮层区域的下界面与结缔组织、脂肪等皮下组织分割更加精确,区域边缘更加圆滑精细。表皮层区域PRE达到97.45%,SE达到98.49%,MIOU达到95.99%,DSC达到97.94%,HD为13.45;真皮层区域PRE达到95.31%,SE达到86.44%,MIOU达到81.63%,DSC达到89.83%,HD为6.36。结论相比于原始U-Net网络,该方法有效地提升了皮肤超声图像的分割性能,证明了该方法的有效性。

Objective To develop a portable very-high-frequency ultrasound biomicroscope. Methods This system is primarily an ultrasonic transducer, ultrasonic transmission and receiving modules, imaging software on a host computer and peripheral equipment. A PVDF transducer with a frequency between 20 and 50 MHz was used for the ultrasonic transducer. In the transmission and receiving modules, the radio frequency echo signals were digitized by high-speed A/D. Then, the digital signals were transmitted, added, filtered, demodulated, log-amplified, double-sampled, and transferred to the host computer by USB interface for real-time display. Results The system was tested with a resolution test and an imaging experiment using a normal human eye, and improved experimental results and real-time images were obtained. Conclusions The system enabled real-time imaging using portable VHF ultrasound biomicroscope. The scheme was concise and clear. The overall design of the system was simple, and the overall performance and portability of the system were improved.

37 Background: Accumulating evidence indicates endothelial progenitor cells (EPCs) play a major role in regulating pulmonary vascular remodeling during pulmonary arterial hypertension (PAH) development. However, there is no reliable method of real-time trafficking and quantification of transferred EPCs. We aimed to detect the homing of EPCs in health and PAH rats by 89Zr-oxine labeling PET imaging. Methods: EPCs, isolated from human peripheral blood, was identified by specific EPCs biomarkers. The effect of 89Zr-oxine labeling on EPCs cell viability and proliferation was evaluated in vitro. 89Zr-labeled EPC cells (2 × 106 cells, 90 kBq/106 cells) were transferred intravenously to health and PAH rats and serial microPET/CT images were obtained. Results: EPCs were characterized and efficiently labeled. In vitro viability and proliferation were not significantly reduced when labeled with 90 kBq per million cells. Intravenously administered 89Zr-labeled EPCs distributed primarily to the lung at 1 h and then subsequently migrated to the liver and spleen. The liver and spleen showed moderate accumulation with the highest %ID/g value of 5.30 ± 1.10 and 2.95 ± 0.54 at 72 h, respectively. The kidney and heart were also slightly visualized. Furthermore, microPET/CT results showed significantly higher accumulation of EPCs in lung of PAH than control group. Conclusions: 89Zr-oxine can be used to delineate EPCs in lung of PAH by PET imaging and may provide a noninvasive EPCs monitoring tool. Figure legendA. Immunostaining and flow cytometric analysis of EPCs revealed expression of endothelial cell-specific markers CD31 (green), CD144 (green), and vWF (green), CD146 (red), KDR and the progenitor cell marker CD34. Nuclei were counterstained with DAPI (blue). Abbreviations: DAPI, 4,6-diamidino-2-phenylindole; vWF, von Willebrand factor; KDR, kinase insert domain receptor. B. Whole-body microPET/CT imaging of 89Zr-oxine EPCs in health rats. Maximum intensity projections (MIPs) of representative mice are shown at several time points after injection. C. Axial section showing the microPET/CT signal of 89Zr-oxine EPCs at the lung of the representative health and pulmonary arterial hypertension (PAH) rats. D. Quantification of PET images in the lung of health and PAH rats treated with 89Zr-oxine EPC cells over 10 days. Acknowledgments:This work was sponsored in part by the National Natural Science Foundation of China (Grant No. 81571713), Capital’s Funds for Health Improvement and Research(CFH) (Grant No. 2016-2-40115), CAMS Innovation Fund for Medical Sciences(CIFMS)(Grant No. 2016-I2M-4-003, 2017-I2M-3-001, 2018-I2M-3-001). No other potential conflicts of interest relevant to this article exist.