This study has shown that a deep neural network can detect referable diabetic retinopathy with sensitivities and specificities close to or better than 80% from both an international and a domestic (New Zealand) database. We believe that deep neural networks can be integrated into community screening once they can successfully detect both diabetic retinopathy and diabetic macular oedema.

The increasing use of invoicing has created an unnecessary burden on labor and material resources in the financial sector. This paper proposes a method to intelligently identify invoice information based on template matching, which retrieves the required information by image preprocessing, template matching, an optical character recognizing, and information exporting. The origin invoice image is preprocessed first to remove the useless background information by secondary rotation and edge cutting. Then, the region of the required information in the obtained regular image is extracted by template matching, which is the core of the intelligent invoice information identification. The optical character recognizing is utilized to convert the image information into text so that the extracted information can be directly used. The text information is exported for backup and subsequent use in the last step. The experimental results indicate that the method using normalized correlation coefficient matching is the best choice, demonstrating a high accuracy of 95%, and the average running time of 14 milliseconds. INDEX TERMS Invoice information identification, template matching, contour extraction, image processing, convolutional neural network.

Signal modulation identification (SMI) plays a very important role in orthogonal frequency-division multiplexing (OFDM) systems. Currently, SMI methods are often implemented via feature extraction based on machine learning. However, the traditional methods encounter a bottleneck where the probability of correct classification (PCC) is very limited and hence it is hard to implement in practical OFDM systems due to the fact that traditional methods are difficult to extract feature of the OFDM signals. In order solve these problems, we propose a deep learning (DL) based SMI method for identifying OFDM signals. Specifically, convolutional neural network (CNN) is adopted to train in-phase and quadrature (IQ) samples for OFDM signals. Then we choose dropout layer to prevent overfitting and improve its identification accuracy. In addition, datasets with different modulation modes are adopted to verify our trained CNN. Experiments are conducted to show that our proposed method achieves higher accuracy and better consistency than traditional methods. Moreover, extensive results confirm that the proposed method performs robustly in different datasets. INDEX TERMS Orthogonal frequency-division multiplexing (OFDM), deep learning (DL), signal modulation identification (SMI), convolutional neural network (CNN).

No abstract

No abstract

Background: Over 700,000 New Zealanders (NZ), particularly elderly and Māori, live without timely access to specialist ophthalmology services. Teleophthalmology is a widely recognised tool that can assist in overcoming resource and distance barriers. Teleophthalmology gained unprecedented traction in NZ during the COVID-19 pandemic and subsequent lockdown. However, its provision is still limited and there are equity issues. The aim of this study was to conduct a systematic review identifying, describing and contrasting teleophthalmology services in NZ with the comparable countries of Australia, USA, Canada and the United Kingdom. Methods: The electronic databases Embase, PubMed, Web of Science, Google Scholar and Google were systemically searched using the keywords: telemedicine, ophthalmology, teleophthalmology/teleophthalmology. The searches were filtered to the countries above, with no time constraints. An integrative approach was used to synthesise findings. Results: One hundred and thirty-two studies were identified describing 90 discrete teleophthalmology services. Articles spanned from 1997 to 2020. Models were categorised into general eye care (n=21; 16%); emergency/trauma (n=6; 4.5%); school screening (n=25; 19%); artificial intelligence (AI) (n=23; 18%); and disease-specific models of care (MOC) (n=57; 43%). The most common diseases addressed were diabetic retinopathy (n=23; 17%); retinopathy of prematurity (n=9; 7%); and glaucoma (n=8; 6%). Programs were mainly centred in the US (n=72; 54.5%), followed by the UK (n=29; 22%), then Canada (n=16; 12%), Australia (n=13; 10%), with the fewest identified in NZ (n=3; 2%). Models generally involved an ophthalmologist consultative service, remote supervision and triaging. Most models involved local clinicians transmitting fed-forward or live images. Conclusion: Teleophthalmology will likely play a crucial role in the future of eye care. COVID-19 has offered a unique opportunity to observe the use of teleophthalmology services globally. Feed-forward and, increasingly, live-based teleophthalmology services have demonstrated feasibility and cost-effectiveness in similar countries internationally. New Zealand's teleophthalmology services, however, are currently limited. Investing in strategic partnerships and technology at a national level can advance health equities in ophthalmic care.

Importance Artificial intelligence (AI) algorithms are under development for use in diabetic retinopathy photo screening pathways. To be clinically acceptable, such systems must also be able to classify other fundus abnormalities and clinical features at the point of care. Background We aimed to develop an AI system that can detect several fundus pathologies and report relevant clinical features. Design Convolutional neural network training with retrospective data set. Participants Colour fundus photos were obtained from publicly available fundus image databases. Methods Images were uploaded to a web‐based AI platform for training and validation of AI classifiers. Separate classifiers were created for each fundus pathology and clinical feature. Main Outcome Measures Accuracy, sensitivity, specificity and area under receiver operating characteristic curve (AUC) for each classifier. Results We obtained 4435 images from publicly available fundus image databases. AI classifiers were developed for each disease state above. Although statistical performance was limited by the small sample size, average accuracy was 89%, average sensitivity was 75%, average specificity was 89% and average AUC was 0.58. Conclusion and Relevance This study is a proof‐of‐concept AI system that could be implemented within a diabetic photo‐screening pathway. Performance was promising but not yet at the level that would be required for clinical application. We have shown that it is possible for clinicians to develop AI classifiers with no previous programming or AI knowledge, using standard laptop computers.

The recent advances in mobile technology have made the smartphone a powerful and accessible tool. This article describe the development of a novel smartphone-based anterior segment microscope that is compatible with tele-manufacturing. The anterior segment microscope is equipped with both cobalt-blue and red-free filters that can be used for clinical photo-documentation. The digital files of the microscope are transferrable and compatible with additive-manufacturing. Therefore, the entire device can be locally manufactured with rapid prototyping techniques such as 3D printing.