Objective:To develop and evaluate a deep learning-based automatic measurement model for panoramic ultrasound biomicroscopy (UBM) images.Methods:A diagnostic test study was conducted.Preoperative UBM examination results of 372 patients who underwent implantable collamer lens (ICL) implantation were collected at the Eye Center of Renmin Hospital of Wuhan University between February 2021 and March 2023.A total of 1 368 panoramic UBM images were obtained to establish an image database.The dataset was divided into a training set (760 images), a validation set (86 images) and an internal test set (522 images).An expert panel consisting of three ophthalmologists annotated the images.The UNet+ + network was used to automatically segment anterior segment tissues, such as the cornea, lens and iris.In addition, image processing techniques and geometric localization algorithms were developed to automatically identify the anatomical landmarks of pupil diameter (PD), anterior chamber depth (ACD), angle-to-angle distance (ATA) and sulcus-to-sulcus distance (STS) to complete the measurement of these parameters.Additionally, 480 panoramic UBM images of 135 patients (240 eyes) from Aier Eye Hospital of Wuhan University were used as an external test set to further evaluate the performance of the model in different centers.The consistency between the measurements from the model and expert panel, the Pentacam system was assessed.Finally, 150 images were randomly selected from the external test set for a human-machine comparison to further evaluate the model's performance.This study adhered to the Declaration of Helsinki.The study protocol was approved by the Ethics Committee of Renmin Hospital of Wuhan University (No.WDRY-2022-K109) and Aier eye Hospital of Wuhan University (No.2023IRBKY120903).Written informed consent was obtained from each subject.Results:In the internal test dataset and external test dataset, with manual labeling as the reference standard, the model achieved a mean Dice coefficient of not less than 0.882.At least 95.65% of the anatomical landmark localization results had Euclidean distance differences within 250 μm.The intraclass correlation coefficients (ICCs) for the measurements of PD, ACD, angle-to-angle ATA, and STS were at least 0.958, with mean relative errors not exceeding 2.407%.With the Pentacam measurements as the reference standard, the ICCs for PD in the internal and external test sets were 0.540 and 0.466, respectively, while the ICCs for ACD were 0.946 and 0.908, respectively.In the human-machine comparison, the ICCs between the model's measurements and those of senior experts were all not lower than 0.969.Conclusions:The deep learning-based model can automatically measure anterior segment parameters from preoperative panoramic UBM images of patients undergoing ICL surgery.The model demonstrates a consistency comparable to that of senior experts, while providing higher efficiency.In terms of ACD measurement, the model shows good agreement between the measurements obtained from the model and Pentacam system.

Objective:To develop and evaluate a deep learning-based automatic measurement model for panoramic ultrasound biomicroscopy (UBM) images.Methods:A diagnostic test study was conducted.Preoperative UBM examination results of 372 patients who underwent implantable collamer lens (ICL) implantation were collected at the Eye Center of Renmin Hospital of Wuhan University between February 2021 and March 2023.A total of 1 368 panoramic UBM images were obtained to establish an image database.The dataset was divided into a training set (760 images), a validation set (86 images) and an internal test set (522 images).An expert panel consisting of three ophthalmologists annotated the images.The UNet+ + network was used to automatically segment anterior segment tissues, such as the cornea, lens and iris.In addition, image processing techniques and geometric localization algorithms were developed to automatically identify the anatomical landmarks of pupil diameter (PD), anterior chamber depth (ACD), angle-to-angle distance (ATA) and sulcus-to-sulcus distance (STS) to complete the measurement of these parameters.Additionally, 480 panoramic UBM images of 135 patients (240 eyes) from Aier Eye Hospital of Wuhan University were used as an external test set to further evaluate the performance of the model in different centers.The consistency between the measurements from the model and expert panel, the Pentacam system was assessed.Finally, 150 images were randomly selected from the external test set for a human-machine comparison to further evaluate the model's performance.This study adhered to the Declaration of Helsinki.The study protocol was approved by the Ethics Committee of Renmin Hospital of Wuhan University (No.WDRY-2022-K109) and Aier eye Hospital of Wuhan University (No.2023IRBKY120903).Written informed consent was obtained from each subject.Results:In the internal test dataset and external test dataset, with manual labeling as the reference standard, the model achieved a mean Dice coefficient of not less than 0.882.At least 95.65% of the anatomical landmark localization results had Euclidean distance differences within 250 μm.The intraclass correlation coefficients (ICCs) for the measurements of PD, ACD, angle-to-angle ATA, and STS were at least 0.958, with mean relative errors not exceeding 2.407%.With the Pentacam measurements as the reference standard, the ICCs for PD in the internal and external test sets were 0.540 and 0.466, respectively, while the ICCs for ACD were 0.946 and 0.908, respectively.In the human-machine comparison, the ICCs between the model's measurements and those of senior experts were all not lower than 0.969.Conclusions:The deep learning-based model can automatically measure anterior segment parameters from preoperative panoramic UBM images of patients undergoing ICL surgery.The model demonstrates a consistency comparable to that of senior experts, while providing higher efficiency.In terms of ACD measurement, the model shows good agreement between the measurements obtained from the model and Pentacam system.

Objective:To construct an artificial intelligence (AI)-assisted system based on deep learning for corneal in vivo confocal microscopy (IVCM) image recognition and to evaluate its value in clinical applications. Methods:A diagnostic study was conducted.A total of 18 860 corneal images were collected from 331 subjects who underwent IVCM examination at Renmin Hospital of Wuhan University and Zhongnan Hospital of Wuhan University from May 2021 to September 2022.The collected images were used for model training and testing after being reviewed and classified by corneal experts.The model design included a low-quality image filtering model, a corneal image diagnosis model, and a 4-layer identification model for corneal epithelium, Bowman membrane, stroma, and endothelium, to initially determine normal and abnormal corneal images and corresponding corneal layers.A human-machine competition was conducted with another 360 database-independent IVCM images to compare the accuracy and time spent on image recognition by three senior ophthalmologists and the AI system.In addition, 8 trainees without IVCM training and with less than three years of clinical experience were selected to recognize the same 360 images without and with model assistance to analyze the effectiveness of model assistance.This study adhered to the Declaration of Helsinki.The study protocol was approved by the Ethics Committee of Renmin Hospital of Wuhan University (No.WDRY2021-K148).Results:The accuracy of this diagnostic model in screening high-quality images was 0.954.Its overall accuracy in identifying normal/abnormal corneal images was 0.916 and 0.896 in the internal and external test sets, respectively.Its accuracy reached 0.983, 0.925 in the internal test sets and 0.988, 0.929 in the external test sets in identifying corneal layers of normal and abnormal images, respectively.In the human-machine competition, the overall recognition accuracy of the model was 0.878, which was similar to the average accuracy of the three senior physicians and was approximately 300 times faster than the experts in recognition speed.Trainees assisted by the system achieved an accuracy of 0.816±0.043 in identifying corneal layers of normal and abnormal images, which was significantly higher than 0.669±0.061 without model assistance ( t=6.304, P<0.001). Conclusions:A deep learning-based assistant system for corneal IVCM image recognition is successfully constructed.This system can discriminate normal/abnormal corneal images and diagnose the corresponding corneal layer of the images, which can improve the efficiency of clinical diagnosis and assist doctors in training and learning.

Objective To investigate the multi-drug resistance of Klebsiella strains and its mechanism.Methods Twenty strains of Klebsiella were isolated from the Affiliated Hospital of Medical College,Ningbo University from October 2009 to March 2011,in which 18 isolates were Klebsiella pneumonia and 2 were Klebsiella planticola. Drug sensitivity was determined by K-B tests. Drug resistant genes gyrA,parC (chromosome mediated) and aac( 6′)-I b-Cr,qnrA,qnrB,qnrS,qepA (plasmid mediated) were amplified by PCR and verified by direct automated fluorogenic sequencing. Results Resistance to β-1actams,aminoglycosides and quinolones was observed in 20 strains,and resistant rates were all above 80％.Klebsiella planticola strains were sensitive to imipenem and meropenem.Mutations of gyrA and parC genes existed in 18 strains (90％),and the positive rates of aac (6') -I b-C r,qnrB and qnrS were 60％ (12/20),20％ (4/20) and 20％ (4/20),respectively.Conclusion The mutations ofgyrA and parC genes may be the main cause of the resistance to quinolones in these strains.