Biometric Information Research Articles

Abstract 4131796: Non-Contact Biometric System for Early Detection of Hypertension and Diabetes Using AI and RGB Imaging

Introduction: In recent years, the development of wearable devices and bioinformation analysis applications utilizing smartphone sensors has seen significant advancements. Despite these technological strides, their adoption and sustained use remain limited to health-conscious individuals. The general populace, particularly those indifferent to health management, stands to miss out on the potential benefits of wearable technologies due to a lack of engagement. This disconnect underscores the necessity for more accessible health monitoring solutions. Research Questions and Goals: This study aims to bridge the gap in health monitoring accessibility by introducing a system capable of non-invasive, non-contact acquisition of biometric information. Such a system seeks to democratize the early detection and prevention of diseases across various demographics. Methods and Results: A prospective cohort clinical study was conducted in a hospital setting with approximately 200 consenting patients and healthy controls. Images of the subjects' faces and hands were captured with a high-speed camera, and blood pressure measurements were taken simultaneously for correlation. A machine learning algorithm was developed to analyze skin blood flow and spectral features from these images to identify potential diseases. The study specifically focused on early detection of hypertension and diabetes. Hypertension was categorized for individuals with systolic blood pressure ≥115 mmHg or diastolic blood pressure ≥75 mmHg. Diabetes was identified in subjects with HbA1c levels ≥6.5 or those previously diagnosed. The algorithm's performance in detecting hypertension showed a 94.2% in alignment with the AHA guidelines for stage1 hypertension, based on pulse wave features. For a subset including 40% early hypertension cases, accuracy rates were 86.2% for 30-second data segments and 80.9% for 5-second segments. For diabetes detection, the application of the algorithm to video data achieved a 75.3%, utilizing blood flow patterns as markers. Hypertension and diabetes were analyzed using different features. Conclusions: The integration of AI algorithms with video data analysis of skin and blood vessels demonstrates a promising avenue for the non-contact, early detection of hypertension and diabetes. This approach not only offers a viable non-contact monitoring technology, but also represents a significant leap towards inclusive, accessible health care prevention and management strategies.

Biometrically measured sleep in medical students as a predictor of psychological health and academic experiences in the preclinical years

ABSTRACT Background Student wellness is of increasing concern in medical education. Increased rates of burnout, sleep disturbances, and psychological concerns in medical students are well documented. These concerns lead to impacts on current educational goals and may set students on a path for long-term health consequences. Methods Undergraduate medical students were recruited to participate in a novel longitudinal wellness tracking project. This project utilized validated wellness surveys to assess emotional health, sleep health, and burnout at multiple timepoints. Biometric information was collected from participant Fitbit devices that tracked longitudinal sleep patterns. Results Eighty-one students from three cohorts were assessed during the first semester of their M1 preclinical curriculum. Biometric data showed that nearly 30% of the students had frequent short sleep episodes (<6 hours of sleep for at least 30% of recorded days), and nearly 68% of students had at least one episode of three or more consecutive days of short sleep. Students that had consecutive short sleep episodes had higher rates of stress (8.3%) and depression (5.4%) symptoms and decreased academic efficiency (1.72%). Conclusions Biometric data were shown to significantly predict psychological health and academic experiences in medical students. Biometrically assessed sleep is poor in medical students, and consecutive days of short sleep duration are particularly impactful as it relates to other measures of wellness. Longitudinal, biometric data tracking is feasible and can provide students the ability to self-monitor health behaviors and allow for low-intensity health interventions.

Provably Quantum Secure Three-Party Mutual Authentication and Key Exchange Protocol Based on Modular Learning with Error

With the rapid development of quantum computers, post-quantum cryptography (PQC) has become critical technology in the security field. PQC includes cryptographic techniques that are secure against quantum-computer-based attacks, utilizing methods such as code-based, isogeny-based, and lattice-based approaches. Among these, lattice-based cryptography is the most extensively studied due to its ease of implementation and efficiency. As quantum computing advances, the need for secure communication protocols that can withstand quantum computer-based threats becomes increasingly important. Traditional two-party AKE protocols have a significant limitation: the security of the entire system can be compromised if either of the communicating parties behaves maliciously. To overcome this limitation, researchers have proposed three-party AKE protocols, where a third party acts as an arbiter or verifier. However, we found that a recently proposed three-party AKE protocol is vulnerable to quantum-computer-based attacks. To address this issue, we propose a provably quantum secure three-party AKE protocol based on MLWE. The proposed scheme leverages the user’s biometric information and the server’s master key to prevent the exposure of critical parameters. We analyzed the security of the protocol using simulation tools such as the Burrows–Abadi–Needham (BAN) logic, Real-or-Random (RoR) model, and Automated Validation of Internet Security Protocols and Applications (AVISPA). Furthermore, comparative analysis with similar protocols demonstrates that our protocol is efficient and suitable.

Privacy-Preserving Computing Services for Encrypted Personal Data Through Streams Over Distributed Ledgers

The growing adoption of wearables is driving the demand for personalized services that leverage unprocessed data, such as biometric and health information, to enhance user experiences and support through software applications. However, several existing use cases involving this information still prioritize traditional schemes, neglecting user privacy. Consequently, the transparency of data transmission paths and the potential for tampering remain ambiguous when users share data with service providers. In this paper, we propose the application of an Internet of Things device-focused distributed ledger as an underlying layer for the transmission of encrypted data using streams. Moreover, our proposal enables data recording for future events and the implementation of multi-subscriber models, allowing client information to be shared securely with different service providers. Through simulation experiments conducted on constrained devices, we demonstrate that our proposed framework efficiently transmits large ciphertexts through streams on a distributed ledger, overcoming the inherent limitations of such networks when dealing with substantial data volumes. Ultimately, the performance metrics presented prove that the proposed model is suitable for real-world applications requiring continuous data collection by wearables and subsequent transmission to service providers.

A Mathematical Modelling for Three-Factor Quantum Biometric Authentication Using Machine Learning, Double-Layer Encryption, Cryptographic Algorithms

Traditional authentication mechanisms are vulnerable to sophisticated attacks in the ever-changing cybersecurity paradigms. In this paper, a new three-factor quantum biometric system is proposed that were analysed to increase security level and reduce threats. This model employs multiple strategies, like quantum key distribution, a biometric identification system and machine learning which encrypts data based on their unique properties in an innovative way to provide not only one but two layers of encryption from unauthorized access. The system in its essence combines three critical elements: quantum generated cryptographic keys, biometric data (fingerprint or retinal scans) and dynamic behaviour-based identification using machine learning algorithms. By adding QKD to the hardware, an encryption key that should be practically impossible for anyone else (except sender and receiver) to intercept is securely generated and distributed through quantum mechanics principles of detection preventing possible eavesdropper. Secondly, the biometric data means that an imposter would have to not only steal your smartphone but also be able to authenticate themselves using unique physiology which takes even more of the likelihood out. Machine learning models study user behaviour patterns and work to adapt against evolving threats, gradually improving the accuracy of authentication with time. For added security, the system uses a double-layer encryption routine. The newly proposed system has a two-level encryption, and in the first layer, quantum keys encrypted bio-metric data while classical algorithms are applied to encrypt overall communication as well storage process. Besides ensuring the safety of private biometric information through dual encryption, this added redundancy also allows them to function as extra fail safe so that if one layer is breached, the entire system will still stand uncompromised. We evaluate the implementation of this new authentication scheme to a series-simulations and real-world test, showing the performance in different scenarios including high threat level environments. The results show major security enhancement compared to the traditional implementation of authentication methods with an observed amount decrease in data breaches and unsanctioned access attempts.

BIAS: A novel secure and efficient biometric-based anonymous authentication scheme

Currently, biometric-based authentication schemes are widely adopted in the field of online payments. Consequently, this has led to an increasing number of people becoming concerned about the privacy-preservation of their biometric data. Gunasinghe et al. presented PrivBioMTAuth, a solution for mobile phone biometric-based authentication designed to protect users’ privacy. However, the solution has drawbacks, such as its impact on the execution efficiency of the authentication protocol or its vulnerability to man-in-the-middle attacks during the authentication phase. Moreover, the user’s biometric image and the password must be revealed to the identity provider, which may raise security concerns. In this work, we present a novel secure and efficient biometric-based anonymous authentication solution with fully succinct verification and significantly lower storage and communication overhead. Different from PrivBioMTAuth, we rely on the NIZK argument given in Groth’s work to reduce the size of the anonymous identity and simplify the verification complexity. In addition, we design a high-performance protocol for conducting large-scale verification of the user’s anonymous identities. We propose an optimized multi-exponentiation argument based on Bayer et al.’s work and utilize it to ensure that a semi-trusted identity provider who seeks to access the users’ sensitive biometric information can faithfully execute the users’ identity registration protocol. The experiment results show that our proposed scheme is efficient and has privacy-preserving capabilities, and it can be applied in the resource-constrained devices.

The Use of Fuzzy Vault Technology to Find Out Fingerprint Differences

Fuzzy vault technology is different from biometric security and information recovery. In this article, we explore the utilization of fuzzy vaults in healthcare, because we can acquire crucial information like the vital signs of a person and the specific details of their fingerprint, even when there is limited information about the individual. We discuss both the conceptualized ideas that serve as the foundation for fuzzy vaults, as well as the practical application of these ideas to secure biological templates via biometric means. Our research involves both computer-simulated and actual case history data. Through this, we attempt to balance between preserving biometric privacy while still achieving a degree of identificable physiological variation. In our presentation, we demonstrate a hybrid of tabular data alongside visual illustrations that illustrate the association between different parameters regarding accuracy and the degree of difficulty associated with information recovery via fuzzy vaults. Let's consider the benefits and drawbacks of our options- elements like strength in security, precision, or the potential for application in limited spaces. To conclude this discussion without a definitive conclusion, but by pointing out what additional investigations should be conducted regarding the subject; likely medical advances caused by the adoption of new methods in medical settings should be noted.

GazeForensics: DeepFake detection via gaze-guided spatial inconsistency learning

DeepFake detection is pivotal in personal privacy and public safety. With the iterative advancement of DeepFake techniques, high-quality forged videos and images are becoming increasingly deceptive. Prior research has seen numerous attempts by scholars to incorporate biometric features into the field of DeepFake detection. However, traditional biometric-based approaches tend to segregate biometric features from general ones and freeze the biometric feature extractor. These approaches resulted in the exclusion of valuable general features, potentially leading to a performance decline and, consequently, a failure to fully exploit the potential of biometric information in assisting DeepFake detection. Moreover, insufficient attention has been dedicated to scrutinizing gaze authenticity within the realm of DeepFake detection in recent years. In this paper, we introduce GazeForensics, an innovative DeepFake detection method that utilizes gaze representation obtained from a 3D gaze estimation model to regularize the corresponding representation within our DeepFake detection model, while concurrently integrating general features to further enhance the performance of our model. Experimental results demonstrate that our proposed GazeForensics method performs admirably in terms of performance and exhibits excellent interpretability.

Solid-State Electrodes Towards a Real-Time Sensor Array for Health Monitoring

The healthcare industry has been transformed by advances in biotechnology, resulting in improved health outcomes. Central to this has been advances in medical diagnostics that can both monitor patients as well as provide actionable data at the point-of-care. However, further improvements are needed to maximize the utility of diagnostics at times where acute care may be required, such as when individuals are unstable after experiencing physical trauma. When medical resources are limited, the ability to rapidly and effectively triage patients that have experienced trauma is critical to maximizing health outcomes. Notably, individuals that have experienced significant trauma can degrade quickly. Therefore, continuous monitoring approaches are needed that can provide critical biometric and biomarker information that will enable updating of the triage in real-time.At present time, patients are assessed continuously using biometric monitoring. However, biometrics are often a lagging indicator to biochemical markers, and thus while easier to measure, less informative. Currently, the ISTAT-8 is a point-of-care unit to perform an automated, rapid biochemical panel that can assess the patient’s ability to maintain homeostasis by monitoring eight biomarkers. However, the ISTAT-8 captures biochemical data only at a point in time instead of continuously; this is a drawback, as the health of patients who have undergone trauma can improve or deteriorate rapidly. Therefore, a continuous biochemical monitoring solution is needed to allow for triage to happen in real-time.Advances in biotechnology have for the first time made real-time biochemical sensing a reality. Unfortunately, commercial continuous biochemical sensors to-date are limited to subcutaneous continuous glucose monitors (CGMs). CGMs monitor glucose concentrations in the interstitial fluid (ISF), which strongly correlates to blood values. While the CGM is now commonly used, the current devices are not easily multiplexed. Therefore, a scalable, easy to insert system is needed to perform multiplexed measurements; microneedle-based platforms have the promise to enable sensing in the ISF leveraging five or more distinct sensors.Here we present a solid-state platform that can enable multiplexed sensing leveraging hollow, wicking microneedles in fluidic contact with solid-state electrodes. Our approach leverages optimized sensor formulations and electrode materials fabricated onto a printed circuit board (PCB). The putative sensor array can be produced in a modular fashion leveraging scalable technologies and processes. Notably, this talk focuses on methods and approaches to fabricate an array of solid-state sensors that are capable of detecting ions and metabolites over extended periods (> 1 week) and can be readily integrated with microneedles for sensing interstitial fluid.

Impact of a Hospital-Based Food Pharmacy Program on Health Outcomes of Vulnerable Patients.

Access to healthy foods, especially for those who are living with diabetes and hypertension, is crucial in managing these chronic diseases. This study evaluates the implementation of a food pharmacy and food prescription program at a safety-net hospital that serves vulnerable populations. Patients who screen as food insecure using the USDA adult food security survey receive a referral from the physician to the food pharmacy program where a dietician reviews their dietary requirements based on their chronic disease and develops a diet plan. Patients then receive fresh produce, meats and other products every 2weeks that meets their nutritional needs from the food pharmacy. Biometric data from the patient's most recent clinic visit at the time of enrollment was collected as the baseline measures including blood pressure, weight, and HbA1c (if diabetic). Additionally, biometric information was collected from the patient's medical records from regularly scheduled clinic visits at 6month intervals. A total of 266 patients were enrolled in the program during the 13-month period that was studied (121 patients with 6-month data and 68 patients with 12-month data). The statistical analysis showed a significant improvement in diastolic blood pressure at 12months and in weight at both the 6months and 12months timeframes when comparing to baseline biometrics.

A Lateral Flow Biosensor Based on Isothermal Amplification for Visual Identification of Species and Sex from Bloodstain.

The prompt species identification from biological samples at a crime scene can rapidly filter out truly valuable biometric information for subsequent personal identification. Meanwhile, early sex determination can assist in narrowing the pool of suspects. However, the current methods for forensic DNA analysis, particularly in point-of-care scenarios, are often limited by the intricate equipment for signal generation and the laborious procedure for DNA purification. The present study introduces a novel portable lateral flow biosensor that possesses extraction-free and anti-aerosol characteristics for on-site determination of species and sex. The bloodstain can be directly submitted to loop-mediated isothermal amplification (LAMP) for the analysis of both mitochondrial and nuclear DNA. The incorporation of a lateral flow device with gold magnetic nanoparticle probes allows for visual interpretation of results through colorimetric signals while also preventing interference on result judgment from pigments such as hemoglobin. Carryover contamination, which is a disharmonious factor in LAMP, especially as the inherent contradiction derived from uncapping in the lateral flow strategy, has been effectively addressed through the integration of uracil DNA glycosylase without compromising the isothermy throughout the process. As a proof-of-concept experiment, species and sex can be accurately identified within 40 min from trace bloodstains amidst significant background interference by targeting cytochrome b and Y-chromosomal amelogenin. Furthermore, the single-blind study revealed a concordance rate of up to 100% in both simulative degraded and true dated bloodstains. This suggests that this biosensor has the potential to be utilized in forensic DNA analysis at crime scenes.

From Russian filtration operations to the lex ferenda of belligerent occupation: how to protect biometrics and privacy in occupied territories?

This article begins by analysing Russian filtration operations and the collection and use of Ukrainian inhabitants’ biometric information. It finds that whereas the Russian filtration operations constitute a clear violation of the law of belligerent occupation, the collection of personal and biometric information is unlikely to constitute a separate and independent violation. This article argues that the current law of belligerent occupation is unsatisfactory and fails to provide sufficient protection for inhabitants. It compares six different ways to further develop the laws of belligerent occupation to enhance the protection of inhabitants’ biometric data and privacy interests. This article concludes that the best approach is to recognize biometric data as private property under the lex ferenda of belligerent occupation.

Biometrics and new and emerging technologies – future challenges in the use of biometric data in the conduct of military operations

The collection of biometric information is subject to domestic, as well as international legal regulation. These regimes can pose complex challenges to military forces seeking to use biometric data during military operations. Using Australia as a case study, this article identifies how the Australian domestic legal regime applies to the collection, disclosure and use of biometric data by the Australian Defence Force (ADF) when conducting military operations outside Australia. It analyses the impact of this legal regime on the ADF’s participation in coalition operations, deriving a range of principles relevant to the exchange of biometric data that can aid interoperability during coalition operations. It highlights the value of conscious planning for the use of biometric data, detailed knowledge of coalition partner domestic law requirements and information sharing agreements to the conduct of legally robust coalition military operations.

IEEE Transactions on Biometrics, Behavior, and Identity Science Information for Authors

IEEE Transactions on Biometrics, Behavior, and Identity Science Information for Authors

Fairness &amp; Privacy in an Age of Generative AI

Generative AI technologies have made tremendous strides recently and have captured the public’s imagination with their ability to mimic what was previously thought to be a fundamentally human capability: creativity. While such technologies hold great promise to augment human creativity and automate tedious processes, they also carry risks that stem from their development process. In particular, the reliance of foundation models on vast amounts of typically uncurated, often web-scraped training data has led to concerns around fairness and privacy. Algorithmic fairness in this context encompasses concerns around potential biases that can be learned by models due to skews in their training data and then reflected in their generated outputs. For example, without intervention, image generation models are more likely to generate images of lighter skin tone male individuals for professional occupations and images of darker skin tone female individuals for working class occupations. This further raises questions around whether there should be legal protections from such pernicious stereotypical representations. Privacy is also a concern as generative AI models can ingest large amounts of personal and biometric information in the training process, including face and body biometrics for image generation and voice biometrics for speech generation. This Essay will discuss the types of fairness and privacy concerns that generative AI raises and the existing landscape of legal protections under anti-discrimination law and privacy law to address these concerns. This Essay argues that the proliferation of generative AI raises challenging and novel questions around (i) what protections should be offered around the training data used to develop such systems and (ii) whether representational harms should be protected against in an age of AI-generated content.

Jurisprudential Discernment and Protection Path of Personal Biological Information in the Context of Metaverse

元宇宙是借助数字技术打造的沉浸式全息虚拟社会，其实现需要依靠增强现实（AR）、虚拟现实（VR）等沉浸式技术设备获得用户的动态，在此过程中需要大量收集用户的身体特征数据、手势动作、眼球位置信息等广义的“个人生物信息”。当前，我国《个人信息保护法》对于广义上的“个人生物信息”并无明确定义，仅将“个人生物识别信息”列举为“特殊个人敏感信息”的七种情形之一。结合个人生物信息在元宇宙中的技术中立、平台管理、跨国利用、去中心化追责风险，应拓宽“个人生物识别信息”的法律概念，将“可识别”纳入其概念范畴内，以回应个人生物信息的广义内涵。此外，还需要建立去中心化身份认证系统，完善 “My Data” 数据模式，并强化代码与法律的互补，以构建更加全面和有效的个人生物信息保护机制，在技术发展的同时确保个人隐私不受侵害，构建安全、公正、有序的元宇宙环境。 Metaverse is an immersive holographic virtual society created with the help of digital technology, the realisation of which needs to rely on immersive technological devices such as Augmented Reality (AR) and Virtual Reality (VR) to obtain the dynamics of the user, and in the process needs to collect a large number of user’s physical characteristics data, gestures, eye position information and other broadly-defined “personal bio-information”. Currently, China’s Personal Information Protection Law does not have a clear definition of “personal biological information” in the broad sense, and only lists “personal biometric information” as one of the seven types of “special personal sensitive information”. It only lists “personal biometric information” as one of the seven cases of “special personal sensitive information”. Taking into account the risks of technological neutrality, platform management, transnational use and decentralisation of personal biometric information in the meta-universe, the legal concept of “personal biometric information” should be broadened to include “identifiable” in its conceptual scope, in order to respond to the broad connotation of personal biometric information. The legal concept of “personal biometric information” should be broadened to include “identifiable” in order to respond to the broader meaning of personal biometric information. In addition, it is also necessary to establish a decentralised authentication system, improve the “My Data” data model, and strengthen the complementarity between code and law, in order to build a more comprehensive and effective protection mechanism for personal biometric information, to ensure that personal privacy is not infringed upon while the technology develops, and to build a safe, just and orderly meta-universe environment.

13‐1: Invited Paper: New Frontier in Display Technology: OPD Sensor in OLEDs for Healthcare Application

People's interest in health care has increased these days. As various health care platforms are emerging, the corresponding market is also expanding. In order to monitor health conditions in everyday life, no physical damage should be made. One of the non‐invasive method, using light for sensing, is preferred due to this effect. In general, as the wavelength increases, light can penetrate deeper into the skin, leading to different bio‐signals. For example, All‐in‐One Sensor Display, which was announced by Samsung Display at SID 2023, used green light to selectively sense fingerprints on the skin surface by reducing the depth of skin penetration. To apply this progress to health care, red light can also be used for OPDs in the All‐in‐One Sensor Display. In this case, red light from red pixel is used as the light source. The light can reach blood vessels under the epidermis because the skin penetration depth is deeper than green light. The light reaching the blood vessel becomes a signal to gather biometric information. This biometric information is then collected at OPD as the signal reaches the OPD in the panel again. For these series of processes, an OPD capable of effective detecting of light in the 625 nm region emitted by red pixels is required. In this study, OPD was fabricated using RD‐01, a high absorption donor material suitable for red light absorption, and RA‐01, a non‐fullerene acceptor for mass production, to respond to red light near 625 nm. The EQE of the OPD was 25% at wavelength of 625 nm. In addition, absorption spectrum of the OPD covers the Red emission effectively. Our results represent that OLED panel can be used for not only fingerprint recognition but also health care platform.

AI Based Smart Voting using Face Recognition

As society changes and technology plays a bigger role in shaping daily life, the security, effectiveness, and accessibility of traditional voting systems are being questioned. This research presents a novel approach to create an online smart voting system using facial recognition technology. The technology promises to improve the convenience and integrity of electoral processes by integrating cutting-edge facial recognition algorithms with secure online voting platforms. It also addresses common problems such as voter fraud, long lines, and geographical restrictions. The system verifies each vote's legitimacy by registering voters' facial biometric information, securely storing it in a centralized database, and providing live facial authentication while voting. Strong security protocols, such as multi-factor authentication and encryption methods, are notably used to protect voter privacy and prevent election results manipulation. Additionally, the system's removal of the requirement for voters to physically be present at polling places improves voter accessibility, especially for underprivileged or geographically separated groups, which encourages higher levels of political engagement. Through extensive testing and validation that includes both simulated and real-world voting scenarios, the efficacy of this novel approach will be closely examined. Consent, transparency, and data privacy are ethical issues that need to be properly addressed in order to ensure compliance with legal and regulatory requirements. In conclusion, the face recognition online smart voting system is a ground-breaking advancement in electoral technology that might fundamentally alter the way elections are held by enhancing their integrity, efficiency, and inclusivity

Injectable 2D Material-Based Sensor Array for Minimally Invasive Neural Implants.

Intracranial implants for diagnosis and treatment of brain diseases have been developed over the past few decades. However, the platform of conventional implantable devices still relies on invasive probes and bulky sensors in conjunction with large-area craniotomy and provides only limited biometric information. Here, an implantable multi-modal sensor array that can be injected through a small hole in the skull and inherently spread out for conformal contact with the cortical surface is reported. The injectable sensor array, composed of graphene multi-channel electrodes for neural recording and electrical stimulation and MoS2-based sensors for monitoring intracranial temperature and pressure, is designed based on a mesh structure whose elasticrestoringforce enables the contracted device to spread out. It is demonstrated that the sensor array injected into a rabbit's head can detect epileptic discharges on the surface of the cortex and mitigate it by electrical stimulation while monitoring both intracranial temperature and pressure. This method providesgoodpotentialfor implanting a variety of functional devices via minimally invasive surgery.

Multi-risk factors joint prediction model for risk prediction of retinopathy of prematurity

PurposeRetinopathy of prematurity (ROP) is a retinal vascular proliferative disease common in low birth weight and premature infants and is one of the main causes of blindness in children.In the context of predictive, preventive and personalized medicine (PPPM/3PM), early screening, identification and treatment of ROP will directly contribute to improve patients’ long-term visual prognosis and reduce the risk of blindness. Thus, our objective is to establish an artificial intelligence (AI) algorithm combined with clinical demographics to create a risk model for ROP including treatment-requiring retinopathy of prematurity (TR-ROP) infants.MethodsA total of 22,569 infants who underwent routine ROP screening in Shenzhen Eye Hospital from March 2003 to September 2023 were collected, including 3335 infants with ROP and 1234 infants with TR-ROP among ROP infants. Two machine learning methods of logistic regression and decision tree and a deep learning method of multi-layer perceptron were trained by using the relevant combination of risk factors such as birth weight (BW), gestational age (GA), gender, whether multiple births (MB) and mode of delivery (MD) to achieve the risk prediction of ROP and TR-ROP. We used five evaluation metrics to evaluate the performance of the risk prediction model. The area under the receiver operating characteristic curve (AUC) and the area under the precision-recall curve (AUCPR) were the main measurement metrics.ResultsIn the risk prediction for ROP, the BW + GA demonstrated the optimal performance (mean ± SD, AUCPR: 0.4849 ± 0.0175, AUC: 0.8124 ± 0.0033). In the risk prediction of TR-ROP, reasonable performance can be achieved by using GA + BW + Gender + MD + MB (AUCPR: 0.2713 ± 0.0214, AUC: 0.8328 ± 0.0088).ConclusionsCombining risk factors with AI in screening programs for ROP could achieve risk prediction of ROP and TR-ROP, detect TR-ROP earlier and reduce the number of ROP examinations and unnecessary physiological stress in low-risk infants. Therefore, combining ROP-related biometric information with AI is a cost-effective strategy for predictive diagnostic, targeted prevention, and personalization of medical services in early screening and treatment of ROP.