Volume 03 · Issue 03 · 2023

With the popularity and development of artificial intelligence (AI), disease screening systems based on AI algorithms are gradually emerging in the medical field. Such systems can be used for primary screening of diseases to relieve the pressure on primary health care. In recent years, AI algorithms have demonstrated good performance in the analysis and identification of lesion signs in the macular region of fundus color photography, and a screening system for fundus lesion signs applicable to primary screening is bound to emerge in the future. Therefore, to standardize the design and clinical application of macular region lesion sign screening systems based on AI algorithms, the Ocular Fundus Diseases Group of Chinese Ophthalmological Society, in collaboration with relevant experts, developed this guideline after investigating issues, discussing production evidence, and holding guideline workshops. It aimed to establish uniform standards for the definition of the macular region and lesion signs, AI adoption scenarios, algorithm model construction, dataset establishment and labeling, architecture and function design, and image data acquisition for the screening system to guide the implementation of the screening work.

Background Tuberculosis (TB) continues to be prevalent in China also among children and adolescents in China. We built a dynamic mathematical model for TB transmission in China, and applied it to compare the epidemic trends 2021-2030 under a range of screening interventions focusing on children and adolescents. Methods We developed a dynamic mathematical model with a flexible structure. The model can be applied either stochastically or deterministically, and can encompass arbitrary age structure and resistance levels. In the present version, we used the deterministic version excluding resistance but including age structure with six groups: 0-5, 6-11, 12-14, 15-17, 18-64, and 65 years and above. We parameterized the model by literature data and fitting it to case and death estimates provided by the World Health Organization. We compared the new TB cases and TB-related deaths in each age group over the period 2021-2030 in 10 scenarios that involved intensified screening of particular age groups of children, adolescents, or young adults, or decreased or increased diagnostic accuracy of the screening. Results Screening the entire age class of 18-year-old persons would prevent 517,000 TB cases and 14,600 TB-related deaths between years 2021 and 2030, corresponding to 6.6% and 5.5% decrease from the standard of care projection, respectively. Annual screening of children aged 6-11 and, to a lesser extent, 0-5 years, also reduced TB incidence and mortality, particularly among children of the respective ages but also in other age groups. In contrast, intensified screening of adolescents did not have a major impact. Screening with a simpler and less accurate method resulted in worsened outcomes, which could not be offset by more intensive screening. More accurate screening and better sensitivity to detect latent TB could prevent 2.3 million TB cases and 68,500 TB deaths in the coming 10 years. Conclusion Routine screening in schools can efficiently reduce the burden of TB in China. Screening should be intensified particularly among children in primary school age.

Background Manual segmentation of thymoma is an onerous, labor-intensive, and subjective task for radiologists. Accordingly, the development of an automatic and efficient method for thymoma segmentation can be valuable for the early detection and diagnosis of this malignancy. Methods Three hundred and ten subjects were enrolled in this retrospective study and all underwent CECT scans. All the scans were manually labeled by four experienced radiologists. The successful application of convolution neural networks (CNNs) and Transformer in computer vision led us to propose a hybrid CNN-Transformer architecture, named transformer attention Net (TA-Net), that would allow the utilization of both local information from CNN features and the global information encoded by Transformers. U-Net was used as the basic structure and Transformers were inserted into convolution blocks in the encoder. In addition, attention gates were embedded in skip connections to highlight salient features. Comparison of the accuracy, intersection over Union (IoU), Dice score, and Boundary F1 contour matching score (BFScore) between the predicted segmentation and the manual labels were utilized to evaluate segmentation performance. Results For thymoma segmentation using TA-Net, the accuracy, Dice score, IoU, and BFScore were 92.49%, 89.92%, 83.80%, and 0.8945, respectively, and no significant differences were detected among tumor types and enhanced phases. Our proposed method achieved the best performance when compared with state-of-the-art methods. Conclusion The proposed method, which combines CNNs with Transformer, achives outstanding performance in thymoma segmentation compared with previous methods. TA-Net may provide consistent and reproducible delineation, thereby assisting radiologists in clinical applications.

Background Ferroptosis, a pathologic state induced by lipid-driven oxidative stress, is associated with the development of human cancers. Calycosin, a natural compound with antioxidant and anti-inflammatory activities, has promising antitumor effects. However, the ferroptosis-related mechanism of calycosin in the treatment of hepatic carcinoma has not been reported. Methods This study applied network pharmacology and bioinformatic approaches (including Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis) to investigate the pharmacologic targets and mechanism of action of calycosin in the treatment of hepatic carcinoma through targeting ferroptosis. By searching online databases including The Cancer Genome Atlas, FerrDb, GeneCards, SwissTargetPrediction, SuperPred, BindingDB, TargetNet, BATMAN-TCM, and Drugbank, we identified 13 ferroptosis-related putative target genes of calycosin against hepatic carcinoma including IL-6, PTGS2, SRC, HRAS, NQO1, NOX4, PGK1, G6PD, GPI, MIF, NOS2, ALDOA, and SQSTM1. Results Molecular docking analysis revealed that calycosin potentially binded directly with the target proteins IL-6, PTGS2, and SRC. Functional enrichment analysis of these proteins indicated that they were involved in gluconeogenesis and apoptosis through regulation of ERK1, ERK2, and MAPK activities (P < 0.05). Conclusion Calycosin exerts antitumor effects in hepatic carcinoma by targeting ferroptosis through regulation of IL-6, PTGS2, and SRC.

Background Standard views in two-dimensional echocardiography are well established but the qualities of acquired images are highly dependent on operator skills and are assessed subjectively. This study was aimed at providing an objective assessment pipeline for echocardiogram image quality by defining a new set of domain-specific quality indicators. Consequently, image quality assessment can thus be automated to enhance clinical measurements, interpretation, and real-time optimization. Methods We developed deep neural networks for the automated assessment of echocardiographic frames that were randomly sampled from 11,262 adult patients. The private echocardiography dataset consists of 33,784 frames, previously acquired between 2010 and 2020. Unlike non-medical images where full-reference metrics can be applied for image quality, echocardiogram’s data are highly heterogeneous and requires blind-reference (IQA) metrics. Therefore, deep learning approaches were used to extract the spatiotemporal features and the image’s quality indicators were evaluated against the mean absolute error. Our quality indicators encapsulate both anatomical and pathological elements to provide multivariate assessment scores for anatomical visibility, clarity, depth-gain and foreshortedness. Results The model performance accuracy yielded 94.4%, 96.8%, 96.2%, 97.4% for anatomical visibility, clarity, depth-gain and foreshortedness, respectively. The mean model error of 0.375±0.0052 with computational speed of 2.52 ms per frame (real-time performance) was achieved. Conclusion The novel approach offers new insight to the objective assessment of transthoracic echocardiogram image quality and clinical quantification in A4C and PLAX views. It also lays stronger foundations for the operator’s guidance system which can leverage the learning curve for the acquisition of optimum quality images during the transthoracic examination.

Objective The spread of the COVID-19 disease has caused great concern around the world and detecting the positive cases is crucial in curbing the pandemic. One of the symptoms of the disease is the dry cough it causes. It has previously been shown that cough signals can be used to identify a variety of diseases including tuberculosis, asthma, etc. In this paper, we proposed an algorithm to diagnose the COVID-19 disease via cough signals. Methods The proposed algorithm was an ensemble scheme that consists of a number of base learners, where each base learner used a different feature extractor method, including statistical approaches and convolutional neural networks (CNNs) for automatic feature extraction. Features were extracted from the raw signal and some transforms performed it, including Fourier, wavelet, Hilbert-Huang, and short-term Fourier transforms. The outputs of these base-learners were aggregated via a weighted voting scheme, with the weights optimised via an evolutionary paradigm. This paper also proposed a memetic algorithm for training the CNNs in the base-learners, which combined the speed of gradient descent (GD) algorithms and global search space coverage of the evolutionary algorithms. Results Experiments were performed on the proposed algorithm and different rival algorithms which included a number of CNN architectures in the literature and generic machine learning algorithms. The results suggested that the proposed algorithm achieves better performance compared to the existing algorithms in diagnosing COVID-19 via cough signals. Conclusion COVID-19 may be diagnosed via cough signals and CNNs may be employed to process these signals and it may be further improved by the optimization of CNN architecture.

Skin cancer is among the most common and lethal cancer types, with the number of cases increasing dramatically worldwide. If not diagnosed in the nascent stages, it can lead to metastases, resulting in high mortality rates. Skin cancer can be cured if detected early. Consequently, timely and accurate diagnosis of such cancers is currently a key research objective. Various machine learning technologies have been employed in computer-aided diagnosis of skin cancer detection and malignancy classification. Machine learning is a subfield of artificial intelligence (AI) involving models and algorithms which can learn from data and generate predictions on previously unseen data. The traditional biopsy method is applied to diagnose skin cancer, which is a tedious and expensive procedure. Alternatively, machine learning algorithms for cancer diagnosis can aid in its early detection, lowering the workload of specialists while simultaneously enhancing skin lesion diagnostics. This article presented a critical review of select state-of-the-art machine learning techniques used to detect skin cancer. Several studies had been collected, and an analysis of the performance of k-nearest neighbors, support vector machine, and convolutional neural networks algorithms on benchmark datasets was conducted. The shortcomings and disadvantages of each algorithm were briefly discussed. Challenges in detecting skin cancer were highlighted and the scope for future research was proposed.