2023年 · 第03卷第02期

Medical artificial intelligence (AI) and big data technology have rapidly advanced in recent years, and they are now routinely used for image-based diagnosis. China has a massive amount of medical data. However, a uniform criteria for medical data quality have yet to be established. Therefore, this review aimed to develop a standardized and detailed set of quality criteria for medical data collection, storage, annotation, and management related to medical AI. This would greatly improve the process of medical data resource sharing and the use of AI in clinical medicine.

Artificial intelligence (AI) is developing rapidly and is being used in several medical capacities, including assisting in diagnosis and treatment decisions. As a result, this raises the conceptual and practical problem of how to distribute responsibility when AI-assisted diagnosis and treatment have been used and patients are harmed in the process. Regulations on this issue have not yet been established. It would be beneficial to tackle responsibility attribution prior to the development of biomedical AI technologies and ethical guidelines.

In general, human doctors acting as superiors need to bear responsibility for their clinical decisions. However, human doctors should not bear responsibility for the behavior of an AI doctor that is practicing medicine independently. According to the degree of fault—which includes internal institutional ethics, the AI bidding process in procurement, and the medical process—clinical institutions are required to bear corresponding responsibility. AI manufacturers are responsible for creating accurate algorithms, network security, and insuring patient privacy protection. However, the AI itself should not be subjected to legal evaluation since there is no need for it to bear responsibility. Corresponding responsibility should be borne by the employer, in this case the medical institution.

After the outbreak of COVID-19, the interaction of infectious disease systems and social systems has challenged traditional infectious disease modeling methods. Starting from the research purpose and data, researchers improved the structure and data of the compartment model or used agents and artificial intelligence based models to solve epidemiological problems. In terms of modeling methods, the researchers use compartment subdivision, dynamic parameters, agent-based model methods, and artificial intelligence related methods. In terms of factors studied, the researchers studied 6 categories: human mobility, nonpharmaceutical interventions (NPIs), ages, medical resources, human response, and vaccine. The researchers completed the study of factors through modeling methods to quantitatively analyze the impact of social systems and put forward their suggestions for the future transmission status of infectious diseases and prevention and control strategies. This review started with a research structure of research purpose, factor, data, model, and conclusion. Focusing on the post-COVID-19 infectious disease prediction simulation research, this study summarized various improvement methods and analyzes matching improvements for various specific research purposes.

Understanding the interaction between the mitral valve (MV) and the left ventricle (LV) is very important in assessing cardiac pump function, especially when the MV is dysfunctional. Such dysfunction is a major medical problem owing to the essential role of the MV in cardiac pump function. Computational modelling can provide new approaches to gain insight into the functions of the MV and LV.

In this study, a previously developed LV–MV model was used to study cardiac dynamics of MV leaflets under normal and pathological conditions, including hypertrophic cardiomyopathy (HOCM) and calcification of the valve. The coupled LV–MV model was implemented using a hybrid immersed boundary/finite element method to enable assessment of MV haemodynamic performance. Constitutive parameters of the HOCM and calcified valves were inversely determined from published experimental data. The LV compensation mechanism was further studied in the case of the calcified MV.

Our results showed that MV dynamics and LV pump function could be greatly affected by MV pathology. For example, the HOCM case showed bulged MV leaflets at the systole owing to low stiffness, and the calcified MV was associated with impaired diastolic filling and much-reduced stroke volume. We further demonstrated that either increasing the LV filling pressure or increasing myocardial contractility could enable a calcified valve to achieve near-normal pump function.

The modelling approach developed in this study may deepen our understanding of the interactions between the MV and the LV and help in risk stratification of heart valve disease and in silico treatment planning by exploring intrinsic compensation mechanisms.

Breast cancer has the highest prevalence among all cancers in women globally. The classification of histopathological images in the diagnosis of breast cancers is an area of clinical concern. In computer-aided diagnosis, most traditional classification models use a single network to extract features, although this approach has significant limitations. Moreover, many networks are trained and optimized on patient-level datasets, ignoring lower-level data labels.

This paper proposed a deep ensemble model based on image-level labels for the binary classification of breast histopathological images of benign and malignant lesions. First, the BreaKHis dataset was randomly divided into training, validation, and test sets. Then, data augmentation techniques were used to balance the numbers of benign and malignant samples. Third, based on their transfer learning performance and the complementarity between networks, VGG16, Xception, ResNet50, and DenseNet201 were selected as base classifiers.

In a ensemble network model with accuracy as the weight, the image-level binary classification achieved an accuracy of 98.90%. To verify the capabilities of our method, it was experimentally compared with the latest transformer and multilayer perception (MLP) models on the same dataset. Our ensemble model showed a 5%–20% advantage, emphasizing its far-reaching abilities in classification tasks.

This research focuses on improving the performance of a classification model with an ensemble algorithm. Transfer learning has an essential role in classification of small datasets, improving training speed and accuracy. Our model may outperform many existing approaches with respect to accuracy and has applications in the field of auxiliary medical diagnosis.

Lack of depth perception from medical imaging systems is one of the long-standing technological limitations of minimally invasive surgeries. The ability to visualize anatomical structures in 3D can improve conventional arthroscopic surgeries, as a full 3D semantic representation of the surgical site can directly improve surgeons’ ability. It also brings the possibility of intraoperative image registration with preoperative clinical records for the development of semi-autonomous, and fully autonomous platforms. This study aimed to present a novel monocular depth prediction model to infer depth maps from a single-color arthroscopic video frame.

We applied a novel technique that provides the ability to combine both supervised and self-supervised loss terms and thus eliminate the drawback of each technique. It enabled the estimation of edge-preserving depth maps from a single untextured arthroscopic frame. The proposed image acquisition technique projected artificial textures on the surface to improve the quality of disparity maps from stereo images. Moreover, following the integration of the attention-ware multi-scale feature extraction technique along with scene global contextual constraints and multiscale depth fusion, the model could predict reliable and accurate tissue depth of the surgical sites that complies with scene geometry.

A total of 4,128 stereo frames from a knee phantom were used to train a network, and during the pre-trained stage, the network learned disparity maps from the stereo images. The fine-tuned training phase uses 12,695 knee arthroscopic stereo frames from cadaver experiments along with their corresponding coarse disparity maps obtained from the stereo matching technique. In a supervised fashion, the network learns the left image to the disparity map transformation process, whereas the self-supervised loss term refines the coarse depth map by minimizing reprojection, gradients, and structural dissimilarity loss. Together, our method produces high-quality 3D maps with minimum re-projection loss that are 0.0004132 (structural similarity index), 0.00036120156 (L1 error distance) and 6.591908 × 10^-5 (L1 gradient error distance).

Machine learning techniques for monocular depth prediction is studied to infer accurate depth maps from a single-color arthroscopic video frame. Moreover, the study integrates segmentation model hence, 3D segmented maps are inferred that provides extended perception ability and tissue awareness.

Ankylosing spondylitis (AS) generally occurs in young adults. The functional impairments resulting in limitation in activities and social participation might exert lifetime impacts. The present study investigated the preferences for mobile health (mHealth) features motivating the self-management behaviors in AS.

The present study was an internet-based, nationwide quantitative study based on the Chinese Ankylosing Spondylitis Prospective Imaging Cohort (CASPIC) study, which was a nationwide, ongoing, prospective cohort study launched in conjunction with Smart-phone SpondyloArthritis Management System (SpAMS) in China. Participants with AS from the CASPIC were invited to report their mHealth preferences from December 2019 to February 2020. The questionnaire was designed to determine the patient preferences for 28 mHealth features. Sociodemographic characteristics, including age, gender, and work status, were collected.

Among all the visitors to the online questionnaire (n = 872), 93.9% (819/872) respondents fully completed the questionnaire and were enrolled in the present study. The mean age was (33.4 ± 9.0) years, and 70.57% (578/819) of the respondents were males. The mean scores of 22 (78.57%) features were greater than 4 (like or strongly like). The mean standard deviation (SD) score of exercise instructions was 4.70 (0.63), which was the most preferred feature, whereas the social interaction features were preferred the least. Pain analysis was more preferred among female respondents (4.72 vs. 4.60, P = 0.012), whereas all items of the social interaction theme and social interaction as a whole (3.73 vs. 3.52, P < 0.001) were less preferred among female respondents. Additionally, the following themes were more preferred by respondents aged ≤ 40 years: credibility and styling (4.37 vs. 4.19, P < 0.001); disease action support (4.55 vs. 4.47, P = 0.007); and incentivization (4.35 vs. 4.24, P = 0.025).

AS patients show great interest for the majority of mHealth features. Exercise instructions and exercise scheduling are the most preferred features, whereas social interaction is the least preferred feature. In addition, gender-related and age-related differences exist in mHealth feature preferences.