Update on artificial intelligence against COVID-19: what we can learn for the next pandemic—a narrative review
Review Article

Update on artificial intelligence against COVID-19: what we can learn for the next pandemic—a narrative review

Shankargouda Patil1, Frank W. Licari1, Shilpa Bhandi1, Kamran H. Awan1, Rocco Franco2, Vincenzo Ronsivalle3, Marco Cicciù3, Giuseppe Minervini4,5

1College of Dental Medicine Roseman University of Health Sciences, South Jordan, UT, USA; 2Department of Biomedicine and Prevention, University of Rome “Tor Vergata”, Rome, Italy; 3Department of Biomedical and Surgical and Biomedical Sciences, Catania University, Catania, Italy; 4Saveetha Dental College and Hospitals, Saveetha Institute of Medical and Technical Sciences (SIMATS), Saveetha University, Chennai, Tamil Nadu, India; 5Multidisciplinary Department of Medical-Surgical and Dental Specialties, University of Campania Luigi Vanvitelli, Naples, Italy

Contributions: (I) Conception and design: S Patil, FW Licari, S Bhandi; (II) Administrative support: S Patil, FW Licari, S Bhandi; (III) Provision of study materials or patients: KH Awan, R Franco, V Ronsivalle; (IV) Collection and assembly of data: KH Awan, R Franco, V Ronsivalle; (V) Data analysis and interpretation: M Cicciù, G Minervini; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.

Correspondence to: Giuseppe Minervini, DMD, DS, PhD. Multidisciplinary Department of Medical-Surgical and Dental Specialties, University of Campania Luigi Vanvitelli, via Luigi de Crecchio 6, 80138 Naples, Italy; Saveetha Dental College and Hospitals, Saveetha Institute of Medical and Technical Sciences (SIMATS), Saveetha University, Chennai, Tamil Nadu, India. Email: Giuseppe.minervini@unicampania.it.

Background and Objective: The coronavirus disease or coronavirus diseases-19 (COVID-19) is an ongoing pandemic that has created a tremendous public health concern. Apart from the reverse transcription polymerize chain reaction (RT-PCR), imaging findings play a crucial role in confirming its diagnosis and also constraining virus transmission. Artificial intelligence (AI) and its subsets have come to the rescue during these challenging periods and have been largely applied in managing the current COVID-19 pandemic. Role of AI in combating COVID-19 at different levels witnessed an enormous growth as documented in scientific literature. The present narrative review aims to illuminate the current state of evidence by providing an update on the use of AI during the COVID-19 pandemic in the year 2023 by assimilating the literature in identifying roles essayed by AI at different strata of COVID-19 expedition.

Methods: English scientific articles were retrieved using Mesh terms—COVID-19, artificial intelligence with AND as Boolean operator in PubMed Database from January 2023 to December 2023 wherein AI essayed a role either in predicting, diagnosing, screening COVID-19 infection or any role essayed related to the condition. Abstracts, narrative or scoping reviews, systematic review, meta-analysis, comments, editorials were excluded. Data regarding the authors, the methodology with observations and inference as stated by the authors were retrieved and ponder upon to address the objectives of the study

Key Content and Findings: Out of 1,661 articles obtained on initial search, 17 relevant articles were selected on application of the selection criteria. Scientific literature reveals that AI has contributed significantly by exhibiting precise, safe, and efficient imaging potential. Studies have also proposed various deep learning algorithms for the detection and treatment of COVID-19, for follow-ups, to evaluate patient response to treatment and so on.

Conclusions: Thus, AI is swiftly evolving in the arena of healthcare. Further development of AI along with its subgroups can revolutionize public health care by reducing the work pressure on the front-line workers and be the backbone to improved management of a potential pandemic in the future.

Keywords: Artificial intelligence (AI); coronavirus diseases-19 (COVID-19); chest X-ray (CXR); diagnosis; prediction

Received: 30 October 2023; Accepted: 25 April 2024; Published online: 29 May 2024.

doi: 10.21037/jphe-23-139

Introduction

Background

Coronavirus diseases-19 (COVID-19), the ongoing pandemic that emerged in late December 2019 (1), had brought the whole world to a standstill in a short period of time. Its devastating presence through varied mutations led to numerous countries experience COVID-19 waves at different timelines (2). Earlier, though the population of all ages in different phases of life, suffered in discrete ways of their own, presently we are learning to live with it. One of the possible reasons that led to this transition could be attributed to science and technology. Research and advancement of technologies not only did it provide opportunities to understand the unprecedented situation but also helped to devise the response required. People centric technologies such as production of test swabs and protective masks through 3D printing technology helped to meet the demands of the global population. Also, robots were used to deliver food and medicine, disinfect the premises and treat patients thereby reducing the workload of the healthcare force and the risk of infection transmission. Likewise, digital contact tracing technology which were used to track individuals’ movement, notify COVID-19 hotspot regions; IoT-based smart disease surveillance systems used to monitor patients and ensure their compliance with the quarantine requirements and immunity certificate system through blockchain technologies comprised the system centric technologies. On the other hand, data centric technologies consisting of artificial intelligence (AI), big data analytics were employed for analysing the pandemic situation and trend, diagnosis, prognosis, target exploration, drug selection, new drug and vaccine development. The interrelatedness of these technologies enables the process of integrating them to develop even smarter strategies for combating the pandemic challenges (3,4).

AI growth during the 21st century is undeniable (5) and revolutionized different industries with its assorted applications (6-8). While abundant data is a requirement for AI, adaptability to several forms of inputs is an added advantage in healthcare (9). Machine learning, natural language processing, neural network, deep learning, machine vision/computer vision are the commonly utilized AI methods in healthcare applications and services (10). Ability of AI assisted clinical trials to handle large volume of data and produce accurate outcomes has eliminated excessive time consumption. Patient care at every level from detection to outcome prediction, from genome sequencing to curating tailored treatment, image diagnostics, nursing assistance, analyse and deliver inputs for quality of life enhancement, aid in medical error reduction are some of the applications of AI. Major pharmaceutical companies have either comparable collaborations or internal programmes for utilizing AI technology to speed up the drug discovery process either by target identification or drug repurposing (10,11). AI has been previously used in the forecast of influenza, Zika virus, and Ebola outbreaks too (12).

Rationale and knowledge gap

Though the scientific community witnessed the explosive burst in literature on application of AI for COVID-19, it is interesting to note that the earlier studies were mainly focussed on diagnosis especially utilizing the CT scans and chest X-rays (CXRs) as pulmonary presentations were significant (13,14). The apparent reason being the mysterious presentation of COVID-19 which shoved researchers for early detection for treatment initiation with significant fraction of the population being infected. This trend was followed by factors predicting the diagnosis (15), prognosis, mortality, morbidity, drug repurposing, screening, identification of biomarkers, and so on (16-18). Irrespective of COVID-19 declaration as pandemic on March 11 2020 (19), the different COVID-19 waves, emergence of new variants, exercise of COVID-19 measures at different levels of stringency globally are effective to date. Further, the research based empirical knowledge did open up different avenues for exploration and investigation which led to numerous trails and possibilities of application of AI for the prolonged active existence of COVID-19 infection. This emphasizes the need for update on the scientific literature on applications of AI in COVID-19, thereby enabling us to record the recent developments and learning for the unknown future.

Objectives

Hence, the present narrative review aimed to address the application of AI against COVID-19. In addition, the possible application of AI in dentistry too is explored. We present this article in accordance with the Narrative Review reporting checklist (available at https://jphe.amegroups.com/article/view/10.21037/jphe-23-139/rc).

Methods

To address the main objective of the present narrative review, being the application of AI against COVID-19, PubMed database was searched to retrieve articles using the MeSH terms—COVID-19, artificial intelligence with AND as Boolean operator. The search was restricted from January 2023 to December 2023. Only full text articles in English language were considered. Studies which utilized AI in either predicting, diagnosing, screening COVID-19 infection, evaluating biomarkers, drug repurposing and articles related to COVID-19 and its condition alone were considered. Abstracts, narrative or scoping reviews, systematic review, meta-analysis, comments, editorials were excluded. Any articles published before 2023 were not considered for the present review (Table 1). A total of 1,661 articles were obtained on initial search, of which only 1,292 articles presented with full text. On exclusion of reviews, meta-analysis, books, documents, clinical conference presentations, 30 articles were obtained of which 14 relevant articles were considered. Further, on evaluation of the references cited, additional 3 articles were retrieved. The titles and abstracts were screened according to the selection criteria and overall, 17 studies were selected and the major findings are compiled and presented in Table 2.

Table 1

The search strategy summary

Items Specification
Date of search 30 October 2023
Database searched PubMed
Search terms used COVID-19, artificial intelligence
Timeframe 1 January 2023 to 30 December 2023
Inclusion and exclusion criteria Inclusion criteria: studies which utilized AI in either predicting, diagnosing, screening COVID-19 infection, evaluating biomarkers, drug repurposing and articles related to COVID-19 and its condition alone were considered. Only full text articles in English language were considered
Exclusion criteria: abstracts, narrative or scoping reviews, systematic review, meta-analysis, comments, editorials were excluded. Any articles published before 2023 were not considered for the present review
Selection process The selection of studies was performed by two authors (S.P., S.B.) independently. Any disagreement between the two authors was resolved by a third author (K.H.A.)

COVID-19, coronavirus diseases-19.

Table 2

Studies on applications of artificial intelligence on COVID-19

Author Application of AI Data collection Data attributes Observation Inference
Chadaga et al. 2023, (20) Diagnosis of mild-moderate cases using haematological markers Developing a decision support system using AI to diagnose mild-moderate COVID-19 from other similar diseases using demographic and clinical markers Custom stacked ensemble model consisting of various heterogeneous algorithms were utilized Accuracy achieved—89%. Important diagnostic markers were eosinophil, albumin, T. bilirubin, ALP, ALT, AST, HbA1c and TWBC Can be deployed in real-time as a decision support system to validate the results obtained by the RT-PCR tests
Murphy et al. 2023, (21) Screening in low resource settings AI was used for COVID-19 detection using chest X-ray imaging and point-of-care blood tests and compared with antigen-based rapid diagnostic tests CAD4COVID-X ray AI software and COVID-LAB+ AI system was used to predict data from four low resource African settings AI for CXR images was a poor predictor—AUC =0.60, while for differential WBC or a combination with C-reactive protein—AUC for both was 0.74, 83% sensitivity and 63% specificity Reason for poor prediction through images was said to be that majority of positive cases had mild symptoms and no visible pneumonia in the lungs, while screening with point-of care blood tests was more feasible with higher sensitivity
Bercean et al. 2023, (22) Cognitive bias in the quantification of COVID‑19 Demonstration of psychophysical bias of radiologist’s subjective perceptions of CT analysis 40 radiologists were surveyed regarding nine intuitive and nine unintuitive, synthetically generated images. Retrospective analysis of the CT studies of 109 patients from two centers. AI-PROBE protocol used for effect of CAD lung involvement assessment Overestimation bias in the unintuitive cases; overestimation bias decreased with experience. Overall overestimated lung involvement by 15.829%±6.643%. Reduction of absolute overestimation error from 9.5%±6.6% to 1.0%±5.2% AI presented as a valuable complement in mitigating the radiologist’s subjectivity by reducing the overestimation
Zhou et al. 2023, (23) Evaluate COVID-19 vaccine hesitancy and belief Track temporal and spatial distribution of COVID-19 vaccine hesitancy and confidence expressed on Twitter 5,257,385 English-language tweets from 6 countries were classified as intent to accept or reject COVID-19 vaccination and whether it is effective or unsafe by Transformer-based deep learning models Similar trend across the countries within three months span was seen with prevalence of intent to accept vaccination decreased from 71.38% to 34.85%, while believing it to be unsafe continuously rose by 7.49 times Discrepancy across regions and vaccine manufacturers was noted. Importance of monitoring social media to detect emerging trends for timely interventions and the need to provide insight cannot be undermined
Han et al. 2023, (24) Predict COVID-19 from respiratory sound recordings Compared clinicians and ML model in predicting COVID-19 based on respiratory sound recordings 24 audio samples made by 36 clinicians were compared with an ML model trained on 1,162 samples. Also, investigated whether the combination could enhance the performance ML model outperformed the clinicians, with 0.75 sensitivity and 0.83 specificity. Combination achieved—0.83 sensitivity and 0.92 specificity Integration of clinicians and ML model could enhance the audio-based respiratory diagnosis in COVID-19 cases with higher confidence
Liang et al. 2023, (25) Detect amino acids with greatest impact on COVID-19 hospitalization rate Identification of critical SARS-CoV-2 amino acids associated with COVID-19 hospitalization rate using machine learning and statistical modelling 2,383,131 sequences of SARS-CoV-2 from GISAID database based on selection criteria were considered to detect critical amino acids that mostly affects the case hospitalization ratio by using four types of models Critical amino acids mostly affecting the case hospitalization ratio were T19, L24, P25, P26, A27, A67, H69, V70, T95, G142, V143, Y145, E156, F157, N211, L212, V213, R214, D215, G339, R346, S373, L452, S477, T478, E484, N501, A570, P681, and T716 Aid in monitoring high risk amino acids enabling to rapid immediate detection on future severe mutations
Genc et al. 2023, (26) Predict mortality To predict whether ICU patients admitted due to COVID-19 infection will result in mortality Of 589 ICU patients, 471 patients’ data was used to train the AI, and tested on 118 patients for ninety parameters Nine parameters were determined with highest mortality. AI estimated mortality with 83% sensitivity, 84% specificity Mortality among COVID-19 ICU patients can be predicted based on their first day laboratory parameters
Chaitanya et al. 2023, (27) Diagnosis using ECG images Diagnosing COVID-19 using ECG images and deep learning approaches and differentiate from other ECG findings Visual Geometry Group (VGG) and AlexNet architectures classified COVID-19, myocardial infarction, normal sinus rhythm, and other abnormal heart beats using Lead-II ECG image only AlexNet model accuracy—77.42%. VGG19 model accuracy—75% Able to classify COVID-19, myocardial infarction, normal sinus rhythm, and other abnormal heart beats effectively
Chadaga et al. 2023, (28) Diagnosis using clinical markers and five different AI techniques ML classifiers were used for COVID-19 diagnosis from routine blood markers Five different explainable artificial intelligence techniques were used to interpret the predictions Best performed algorithm was k-nearest neighbour with accuracy—84%. Clinical markers—eosinophils, lymphocytes, red blood cells and leukocytes were significant in differentiating COVID-19 More reliable and efficient method in diagnosing, can be utilized synchronously with the standard RT-PCR procedure
Zorzi et al. 2023, (29) Differentiating from other viral pneumonias on CT Developed automatic extraction of QM and RF from lung CT and develop AI models supporting differential diagnosis Different combinations of QM and RF were used to develop four multilayer-perceptron classifiers Classifiers accuracy ranged from 0.71 to 0.80 in differentiating COVID-19 cases from non-viral pneumonia cases Good diagnostic performances by the four AI models which could support clinical decisions
Asteris et al. 2023, (30) Prediction model for ICU hospitalization Designed a risk prediction model of COVID-19 outcome using ANN and minimum number of routine laboratory indices 248 records, each comprised of 25 laboratory indices and the outcome of hospitalization of each patient were split into training [166], validation [41] and testing [41] dataset Best ANN architecture model corresponded to only neutrophil-to-lymphocyte ratio, lactate dehydrogenase, fibrinogen, albumin, and D-dimer with 95.97% accuracy Based on 5 laboratory indices, ANN could predict ICU hospitalization accurately and early
Grodecki et al. 2023, (31) Prediction of adverse outcomes over visual scoring systems Utilized AI for quantification of pneumonia extent in COVID-19 cases from chest CT scans and determined its ability to predict clinical deterioration in comparison with semi-quantitative visual scoring systems Deep-learning algorithm was utilized to quantify the pneumonia burden and primary outcome of clinical deterioration, the composite end point including admission to the ICU and even in-hospital death was considered Prediction of primary outcome was significantly higher for AI-assisted quantitative pneumonia burden with lesser time (38±10 s) when compared with the visual lobar severity score (328±54 s) AI presented accurate prediction of clinical deterioration COVID-19 patients while requiring only a fraction of the analysis time
Zuo et al. 2023, (32) Quantitative analysis of lesion volume change AI was used to dynamically measure the lesion volume in CT images and evaluate the disease outcome in COVID-19 cases Distribution, location, and nature of lesions were analyzed and divided into groups ranging from without abnormal pulmonary imaging to dissipation group AI measured the total volume of pneumonia lesions with sensitivity of 92.10% and specificity of 100% AI aided in assessing the severity and development trend of the disease
Dimitsaki et al. 2023, (33) Predicting severity using plasma proteomics and clinical data Ensemble of ML algorithms that analyzed clinical and biological data of COVID-19 patients was designed for early COVID-19 patient triage Three ML “tasks” were defined and several algorithms were tested through a hyperparameter tuning method and multi-layer perceptron and support vector machines algorithms Discerned critical cases based on patient’s age and plasma proteins on B cell dysfunction, hyper-activation of inflammatory pathways, and hypo-activation of developmental and immune pathways Aid in increasing the possibilities of discovering novel biomarkers and “druggable” targets
Yagin et al. 2023, (34) Identification of COVID-19 gene biomarkers Application of explainable artificial intelligence methods based on machine learning techniques on COVID-19 mNGS samples 15,979 gene expressions of 141 negative and 93 positive cases data were used to determine COVID-19-associated biomarker candidate genes IFI27, LGR6, and FAM83A were the most important genes associated with COVID-19 while high level of IFI27 expression contributed to increasing the probability of positivity AI could explain the biomarker prediction for COVID-19 and provide clinicians with an intuitive understanding and impact interpretability of the risk factors
Arian et al. 2023, (35) Quantification of lung abnormalities and assessment of prognosis Quantification of the total and compromised lung parenchyma using volumetric image analysis software by experienced radiologists and compared against AI-aided quantification model Fraction of compromised lung parenchyma was calculated and the subjects were divided into critical and non-critical groups both by AI model and the radiologists In predicting critical outcomes, AI model achieved 83.9% accuracy, 79.1% sensitivity, 88.6% specificity AI-assisted measurements were similar to quantitative measurements obtained by radiologists

COVID-19, coronavirus diseases-19; AI, artificial intelligence; ECG, electrocardiogram; CT, computed tomography; ICU, intensive care unit; ML, machine learning; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; QM, quantitative metrics; RF, radiomic features; CAD, computer-aided detection; ALP, alkaline phosphatase; ALT, alanine transaminase; AST, aspartate transaminase; HbA1c, hemoglobin A1C; TWBC, total white blood cell; CXR, chest X-ray; AUC, accuracy; VGG, Visual Geometry Group; ANN, artificial neural network; RT-PCR, reverse transcription polymerase chain reaction; mNGS, metagenomic next-generation sequencing.

AI applications against COVID-19—an update

Though reverse transcription polymerize chain reaction (RT-PCR) is considered as a standard diagnostic method, it may be prone to erroneous and false negative results especially in mild to moderated COVID-19 positive cases. Hence, Chadaga et al. focused on using AI to develop a decision support system to diagnose such cases from similar presentation illness. Custom-stacked ensemble model of various heterogeneous algorithms, four deep learning algorithms, five explainers analysed demographic and clinical markers, of which eosinophil, albumin, Total Bilirubin, Alkaline phosphatase, Alanine transaminase, Aspartate transaminase, HBA1c and total white blood cells were prominent with 89% accuracy (20).

Many geographies around the world lack access to hospital laboratories and shortage of medical experts to interpret the basic diagnostic tests such as chest X-ray (CXR) and point-of-care (POC) blood testing for COVID-19 screening and diagnosis. Murphy et al. proposed the possible application of AI to bridge this gap wherein, they analysed data collected from four sites in low resource settings in Lesotho and South Africa. They demonstrated that POC blood tests is feasible with 83% sensitivity and 63% specificity, while the CXR images were a poor predictor owing to the mild symptoms presentation and absence of pneumonia (21).

Radiologists’ assessment of pulmonary involvement is correlated with clinical outcomes to formulate the treatment course in COVID-19 positive cases. Irrespective of the scoring type, the radiologist is first required to identify the affected area followed by estimation of lung damage. Human beings in general exhibit an acute lack of precision in visual geometric comparisons. Despite the wide adoption of lung involvement scores, the area judgement cognitive bias remains unaddressed in radiology. Therefore, Bercean et al. hypothesised that the geometric ratio assessment is prone to an overestimation bias in analysing CTs for COVID-19 and could be mitigated by AI clinical support system. Preliminary survey of 40 radiologists presented an overestimation bias for unintuitive cases, retrospective analysis of CT data from 109 patients from two different centres presented an overestimation of lung involvement by 10.23%±4.65% and 15.8%±6.6%, respectively, while AI decision support reduced the absolute overestimation error from 9.5%±6.6% to 1.0%±5.2% indicating a human perception bias in radiology has clinically meaningful impact on the CT quantitative analysis of COVID-19 (22).

Social media platform is popular for individuals to express their views and present their experiences on numerous topics. It became a supplement platform during COVID-19 pandemic as vast up-to-date longitudinal data from its onset to present are posted. COVID-19 vaccine hesitancy and confidence are a topic that needs to be addressed to predict public responses for new vaccine, to intercept and differentiate the positive and negative beliefs through effective implementation of vaccine policies and health communication strategies accordingly. Zhou et al. conducted deep learning analysis of Twitter data of six high income countries for trends, examined the disparities and identified the potential sociodemographic factors for COVID-19 vaccine hesitancy and confidence. Decline in the intent to accept the vaccination was prevalent while vaccine to be unsafe were dominant views among the public. Moderna and Pfizer vaccines had higher acceptance rates in the United States and Canada, while Astra Zeneca in UK and Johnson & Johnson vaccine in Ireland were most accepted (23).

Pneumothorax is a relatively atypical respiratory finding in COVID-19 with low clinical suspicion at the time of diagnosis, however, may pose a serious potential complication if present. Utilization of AI for pneumothorax detection has shown promise in accelerating its diagnosis. AI-enabled portable X-ray scanner can automatically detect its presence, generate an alert in the patient archiving and communication system, thereby enabling prioritization, rapid review by radiologist and immediate intervention. Hunter et al. presented three case series of COVID-19 positive patients of ages 23 to 76 years in ICU setting wherein prompt detection of pneumothorax detected through an AI-enabled portable radiography machine aided in the swift management of the cases (36).

COVID-19 is known to present with pulmonary manifestations such as cough, shortness of breath, sputum production, respiratory failure, and acute respiratory distress syndrome (ARDS) (37). Auscultation, a traditional diagnostic technique is effective in detecting and differentiating various respiratory conditions. However, mastering the technique remains a challenge. Clinicians are now assisted with machine learning (ML) models that not only analyse the internal sounds, but also breathing and voice sounds. Han et al. compared clinicians and ML model in predicting COVID-19 from respiratory audio recordings utilizing 24 audio samples of which 12 were tested positive, each sample represented one subject, consisting of voice, cough, and breathing sound recordings with length of 20 seconds wherein the ML model outperformed the clinicians (24).

Numerous SARS-CoV-2 mutations have emerged and circulated since the onset of COVID-19 pandemic, yet only few possess the ability to impact the virus phenotype and change the severity of the virus, thereby affecting the virus pathogenicity and hence the risk of hospitalization. Liang et al. utilized two machine learning models and two statistical models to evaluate the relationship between the prevalence of amino acid mutations and the case hospitalization ratio by analysing 2,383,131 SARS-CoV-2 sequences between 16 September 2020 and 22 March 2022 obtained from GISAID database. They identified critical variables comprising of critical amino acids and critical cofounders thereby creating awareness and the need to evaluate the methods by which these key amino acids affect the infection severity and to develop drugs resistant to these new mutations (25).

Studies have also considered AI application in diagnosing COVID-19 infection using electrocardiogram (ECG) images and clinical markers, differentiating from other viral pneumonia cases, predicting the need for ICU hospitalization, identification of biomarkers, severity using plasma proteomics and clinical data, disease outcome, assessment of prognosis and even predict mortality (26-35) (Table 1). These widespread applications have added to the literature, enhanced the understanding, enabled to formulate further research questions to deepen the knowledge, validate and authorize AI implementation methodologically.

AI application in dentistry

The synergy between AI and dentistry presents an exciting avenue for transforming patient care, especially in the context of the COVID-19 pandemic. By harnessing AI’s capabilities, dental practitioners can enhance patient outcomes, mitigate risks, and adapt to the evolving healthcare landscape (38). AI algorithms can assist in the early detection of dental conditions by analyzing radiographic images and identifying abnormalities that might be missed by human observers (39). The COVID-19 pandemic has accelerated the adoption of teledentistry, where AI-driven platforms enable dentists to conduct virtual consultations, assess patient conditions remotely, and provide guidance without the need for physical presence. AI-driven chatbots and virtual assistants can engage patients by providing information about oral hygiene, post-treatment care, and appointment scheduling. These tools enhance patient education and foster a stronger patient-dentist relationship. AI can streamline administrative tasks, such as appointment scheduling, billing, and managing electronic health records, allowing dentists to focus more on patient care. AI can aid in optimizing ventilation systems and predicting areas of high aerosol concentration, reducing the risk of virus transmission within dental clinics (40). AI-driven simulations can model the spread of aerosols during dental procedures, enabling dentists to adapt their techniques to minimize exposure risks. While the integration of AI in dentistry holds immense potential, several challenges must be addressed, including data privacy concerns, regulatory hurdles, and the need for ongoing professional training to effectively use AI tools. Continuous learning programs should be established to ensure dentists are proficient in using AI tools, interpreting AI-generated recommendations, and maintaining a balance between human expertise and AI assistance (41,42).

Lessons for next pandemic

Recorded literature evidence supports that historically from 430 BC, world had witnessed multiple distinct pandemics, epidemics, endemics and currently to date is in the midst of similar scenario (43,44). The adaptability nature of humans enables them to survive, accommodate, evolve and pioneer techniques to overcome such circumstances. Over the years, human presence became distinct and dominant on Earth unlike other species becoming extinct or endangered. Though we cannot predict what the future beholds, yet preparedness to intercept the challenges to come is what we can do. Learning from experience is what COVID-19 taught us all. Development of countermeasures in record time, testing vaccine technologies, sequencing technologies, enrollment of new technologies at large scale comprising of AI and others, online learning platforms and e commerce and many more (45,46). However, individuals’ responsibility towards one’s own physical, mental and overall health, society responsibility of working in unison, government responsibility to invest and support in science and technology and provide factual and verified information to the public (47) emphasis the need to work collectively for a prepared future.

Strengths and limitations

The present narrative review selected, complied and amalgamated data on the application of AI in COVID-19 at different levels of strata from the scientific literature for the year 2023 effectively by addressing many important and unique situations apart from the conventional methods followed or traditional applications such as prediction, diagnosis and follow up. However, by limiting to a specific year, language and database, possibility of unexplored data maybe the shortcoming of the present narrative review.

Conclusions

AI role during COVID-19 pandemic made things easier for navigating through the unknown. In addition to diagnosis, surveillance, detection and analysis, it also helped in understanding the basic structure, hypothesize the mechanism of infection to enable the therapeutic interventions and device of drugs and vaccinations. Global scientific community utilized AI science and technology to overcome the burden and ensure effective regulation of efforts to confine and decrease COVID-19 infection among the populace. It helped to reach the neglected and remote geographies, analyze the presentation among different age groups, identify the susceptible population, compare and contrast the varied findings thus providing the necessary data required for the frontline workers to manage the disease. Therefore, an update of recent findings on the role of AI in COVID-19 either opens new frontiers, re-evaluate or support previous data thereby consolidating the indispensable information.

Acknowledgments

Funding: The research leading to these results has received funding from the European Union-NextGenerationEU through the Italian Ministry of University and Research under PNRR-M4C2-I1.3 Project PE_00000019 "HEAL ITALIA" to Marco Cicciù CUP E63C22002080006. The views and opinions expressed are those of the authors only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.

Footnote

Reporting Checklist: The authors have completed the Narrative Review reporting checklist. Available at https://jphe.amegroups.com/article/view/10.21037/jphe-23-139/rc

Peer Review File: Available at https://jphe.amegroups.com/article/view/10.21037/jphe-23-139/prf

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jphe.amegroups.com/article/view/10.21037/jphe-23-139/coif). G.M. serves as an unpaid editorial member of Journal of Public Health and Emergency from September 2023 to August 2025. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

  1. Yuki K, Fujiogi M, Koutsogiannaki S. COVID-19 pathophysiology: A review. Clin Immunol 2020;215:108427. [Crossref] [PubMed]
  2. Updated working definitions and primary actions for SARS-CoV-2 variants. 4 October 2023. World Health organization. Available online: https://www.who.int/publications/m/item/updated-working-definitions-and-primary-actions-for--sars-cov-2-variants
  3. Minervini G, Franco R, Marrapodi MM, et al. Children oral health and parents education status: a cross sectional study. BMC Oral Health 2023;23:787. [Crossref] [PubMed]
  4. Minervini G, Franco R, Marrapodi MM, et al. Post-traumatic stress, prevalence of temporomandibular disorders in war veterans: Systematic review with meta-analysis. J Oral Rehabil 2023;50:1101-9. [Crossref] [PubMed]
  5. Liu J, Kong X, Xia F, et al. Artificial intelligence in the 21st century. IEEE Access 2018;6:34403-21.
  6. Minervini G, Franco R, Marrapodi MM, et al. The association between parent education level, oral health, and oral-related sleep disturbance. An observational crosssectional study. Eur J Paediatr Dent 2023;24:218-23. [PubMed]
  7. Bredt S. Artificial Intelligence (AI) in the Financial Sector-Potential and Public Strategies. Front Artif Intell 2019;2:16. [Crossref] [PubMed]
  8. Benos L, Tagarakis AC, Dolias G, et al. Machine Learning in Agriculture: A Comprehensive Updated Review. Sensors (Basel) 2021;21:3758. [Crossref] [PubMed]
  9. Patil S, Albogami S, Hosmani J, et al. Artificial Intelligence in the Diagnosis of Oral Diseases: Applications and Pitfalls. Diagnostics (Basel) 2022;12:1029. [Crossref] [PubMed]
  10. Väänänen A, Haataja K, Vehviläinen-Julkunen K, et al. AI in healthcare: A narrative review. F1000Research 2021;10:6. [Crossref]
  11. ShaheenMY. Applications of Artificial Intelligence (AI) in healthcare: A review. ScienceOpenPreprints 2021. doi: 10.14293/S2199-1006.1.SOR-.PPVRY8K.v1
  12. Venna SR, Tavanaei A, Gottumukkala RN, et al. A novel data-driven model for real-time influenza forecasting. IEEE Access 2018;7:7691-701.
  13. Jin C, Chen W, Cao Y, et al. Development and evaluation of an artificial intelligence system for COVID-19 diagnosis. Nat Commun 2020;11:5088. [Crossref] [PubMed]
  14. Mei X, Lee HC, Diao KY, et al. Artificial intelligence-enabled rapid diagnosis of patients with COVID-19. Nat Med 2020;26:1224-8. [Crossref] [PubMed]
  15. Jamshidi MB, Lalbakhsh A, Talla J, et al. Artificial Intelligence and COVID-19: Deep Learning Approaches for Diagnosis and Treatment. IEEE Access 2020;8:109581-95.
  16. Vaishya R, Javaid M, Khan IH, et al. Artificial Intelligence (AI) applications for COVID-19 pandemic. Diabetes Metab Syndr 2020;14:337-9. [Crossref] [PubMed]
  17. Khan M, Mehran MT, Haq ZU, et al. Applications of artificial intelligence in COVID-19 pandemic: A comprehensive review. Expert Syst Appl 2021;185:115695. [Crossref] [PubMed]
  18. Lalmuanawma S, Hussain J, Chhakchhuak L. Applications of machine learning and artificial intelligence for Covid-19 (SARS-CoV-2) pandemic: A review. Chaos Solitons Fractals 2020;139:110059. [Crossref] [PubMed]
  19. Cucinotta D, Vanelli M. WHO Declares COVID-19 a Pandemic. Acta Biomed 2020;91:157-60. [PubMed]
  20. Chadaga K, Prabhu S, Bhat V, et al. Artificial intelligence for diagnosis of mild-moderate COVID-19 using haematological markers. Ann Med 2023;55:2233541. [Crossref] [PubMed]
  21. Murphy K, Muhairwe J, Schalekamp S, et al. COVID-19 screening in low resource settings using artificial intelligence for chest radiographs and point-of-care blood tests. Sci Rep 2023;13:19692. [Crossref] [PubMed]
  22. Bercean BA, Birhala A, Ardelean PG, et al. Evidence of a cognitive bias in the quantification of COVID-19 with CT: an artificial intelligence randomised clinical trial. Sci Rep 2023;13:4887. [Crossref] [PubMed]
  23. Zhou X, Song S, Zhang Y, et al. Deep Learning Analysis of COVID-19 Vaccine Hesitancy and Confidence Expressed on Twitter in 6 High-Income Countries: Longitudinal Observational Study. J Med Internet Res 2023;25:e49753. [Crossref] [PubMed]
  24. Han J, Montagna M, Grammenos A, et al. Evaluating Listening Performance for COVID-19 Detection by Clinicians and Machine Learning: Comparative Study. J Med Internet Res 2023;25:e44804. [Crossref] [PubMed]
  25. Liang J, Ding Z, Liu K. Identification of critical SARS-CoV-2 amino acids associated with COVID-19 hospitalization rate using machine learning and statistical modeling: An observational study in the United States. Infect Genet Evol 2023;113:105480. [Crossref] [PubMed]
  26. Genc AC, Cekic D, Issever K, et al. Can artificial intelligence predict COVID-19 mortality? Eur Rev Med Pharmacol Sci 2023;27:9866-71. [PubMed]
  27. Chaitanya MK, Sharma LD, Rahul J, et al. Artificial intelligence based approach for categorization of COVID-19 ECG images in presence of other cardiovascular disorders. Biomed Phys Eng Express 2023; [Crossref] [PubMed]
  28. Chadaga K, Prabhu S, Bhat V, et al. COVID-19 diagnosis using clinical markers and multiple explainable artificial intelligence approaches: A case study from Ecuador. SLAS Technol 2023;28:393-410. [Crossref] [PubMed]
  29. Zorzi G, Berta L, Rizzetto F, et al. Artificial intelligence for differentiating COVID-19 from other viral pneumonias on CT: comparative analysis of different models based on quantitative and radiomic approaches. Eur Radiol Exp 2023;7:3. [Crossref] [PubMed]
  30. Asteris PG, Kokoris S, Gavriilaki E, et al. Early prediction of COVID-19 outcome using artificial intelligence techniques and only five laboratory indices. Clin Immunol 2023;246:109218. [Crossref] [PubMed]
  31. Grodecki K, Killekar A, Simon J, et al. Artificial intelligence-assisted quantification of COVID-19 pneumonia burden from computed tomography improves prediction of adverse outcomes over visual scoring systems. Br J Radiol 2023;96:20220180. [Crossref] [PubMed]
  32. Zuo YH, Chen Y, Chen LH, et al. Computed Tomography scanning in patients with COVID-19: artificial intelligence analysis of lesions volume and outcome. Eur Rev Med Pharmacol Sci 2023;27:5869-77. [PubMed]
  33. Dimitsaki S, Gavriilidis GI, Dimitriadis VK, et al. Benchmarking of Machine Learning classifiers on plasma proteomic for COVID-19 severity prediction through interpretable artificial intelligence. Artif Intell Med 2023;137:102490. [Crossref] [PubMed]
  34. Yagin FH, Cicek İB, Alkhateeb A, et al. Explainable artificial intelligence model for identifying COVID-19 gene biomarkers. Comput Biol Med 2023;154:106619. [Crossref] [PubMed]
  35. Arian A, Mehrabi Nejad MM, Zoorpaikar M, et al. Accuracy of artificial intelligence CT quantification in predicting COVID-19 subjects' prognosis. PLoS One 2023;18:e0294899. [Crossref] [PubMed]
  36. Hunter JG, Pierce JD, Gilkeson RC, et al. Clinical Implementation of an Artificial Intelligence Tool in the Detection and Management of Pneumothoraces in Patients With COVID-19. Cureus 2023;15:e42509. [Crossref] [PubMed]
  37. Johnson KD, Harris C, Cain JK, et al. Pulmonary and Extra-Pulmonary Clinical Manifestations of COVID-19. Front Med (Lausanne) 2020;7:526. [Crossref] [PubMed]
  38. Sakai T, Li H, Shimada T, et al. Development of artificial intelligence model for supporting implant drilling protocol decision making. J Prosthodont Res 2023;67:360-5. [Crossref] [PubMed]
  39. Jaiswal P, Bhirud S. An intelligent deep network for dental medical image processing system. Biomed Signal Process Control 2023;84:104708. [Crossref]
  40. Singh N, Pandey A, Tikku AP, et al. Attitude, perception and barriers of dental professionals towards artificial intelligence. J Oral Biol Craniofac Res 2023;13:584-8. [Crossref] [PubMed]
  41. Shafi I, Fatima A, Afzal H, et al. A Comprehensive Review of Recent Advances in Artificial Intelligence for Dentistry E-Health. Diagnostics (Basel) 2023;13:2196. [Crossref] [PubMed]
  42. Vila-Blanco N, Varas-Quintana P, Tomás I, et al. A systematic overview of dental methods for age assessment in living individuals: from traditional to artificial intelligence-based approaches. Int J Legal Med 2023;137:1117-46. [Crossref] [PubMed]
  43. Piret J, Boivin G. Pandemics Throughout History. Front Microbiol 2020;11:631736. [Crossref] [PubMed]
  44. Minervini G, Franco R, Marrapodi MM, et al. Conservative treatment of temporomandibular joint condylar fractures: A systematic review conducted according to PRISMA guidelines and the Cochrane Handbook for Systematic Reviews of Interventions. J Oral Rehabil 2023;50:886-93. [Crossref] [PubMed]
  45. Krammer F. The role of vaccines in the COVID-19 pandemic: what have we learned? Semin Immunopathol 2023; Epub ahead of print. [Crossref] [PubMed]
  46. Adedoyin OB, Soykan E. Covid-19 pandemic and online learning: the challenges and opportunities. Interactive Learning Environments 2023;31:863-75. [Crossref]
  47. Faghy M, Arena R, Hills AP, et al. The response to the COVID-19 pandemic: With hindsight what lessons can we learn? Prog Cardiovasc Dis 2023;76:76-83. [Crossref] [PubMed]
doi: 10.21037/jphe-23-139
Cite this article as: Patil S, Licari FW, Bhandi S, Awan KH, Franco R, Ronsivalle V, Cicciù M, Minervini G. Update on artificial intelligence against COVID-19: what we can learn for the next pandemic—a narrative review. J Public Health Emerg 2024;8:16.

Article Options

Download Citation

Share