The role of artificial intelligence (AI) in the elimination of malaria: a systematic literature review

Abstract

The rising incidence of malaria on a global scale poses a substantial public health concern, necessitating effective and innovative ways for tackling it. The review paper explores the capability of Artificial Intelligence (AI) in aiding India’s attempt to eliminate malaria by 2030. Totally, 62 research articles published between 2014 and 2025 are systematically reviewed, covering AI applications in diagnosis (e.g., image-based parasite detection), treatment optimization (predictive analytics), and outbreak prediction. The main findings from research articles indicate that AI models have attained better diagnostic accuracies and significantly decreased diagnostic time as well as human error. Even though there are developments, poor data interoperability, limited rural infrastructure, and gaps in healthcare worker training are some of the major limitations. The review also analyses various AI techniques currently employed globally, which can be adapted for use in India. The paper analyses various AI techniques currently employed globally, which can be adapted for use in India. It also highlights the associated limitations that need to be addressed and suggests areas for future research. This review follows a structured and transparent approach to select, evaluate, and present the role of AI in malaria elimination.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12936-025-05661-7.

Background

Worldwide, the cases of malaria increased to 263 million in 2023, with the incidence rate rising from 60.4 (per 1000 people) to 58.6 in 2022 [1]. This increase was alarming, especially considering that malaria cases had been declining steadily from 2000 to 2019, from 243 to 233 million [1]. Malaria is a dangerous infectious disease caused by the parasite Plasmodium and transmitted via infested female Anopheles mosquitoes. It is an important cause of febrile illness in endemic countries, requiring rapid therapeutic management [2]. In 2017, the World Health Organization (WHO) launched the E-2020 initiative, a therapeutic management strategy aimed at achieving zero indigenous malaria cases. India’s malaria control efforts are supported by the Global Malaria Action Plan, assisting endemic countries in their attempts to minimize the malaria burden using advanced technologies, such as AI [3]. Several new AI systems using Deep Learning (DL) have been recently developed, with the potential to significantly improve how malaria is diagnosed and treated [4].

The use of AI-powered tools has increased in the area of malaria control across various parts of the world. For example, AI techniques that compare the Complete Blood Count (CBC) of patients with and without malaria have been built in a comprehensive manner utilizing algorithmic models, which are constructed using big datasets. These tools, especially those leveraging DL algorithms like Convolutional Neural Networks (CNNs), offer significant advantages over traditional microscopy methods. CNNs have the capability to process images in a grid-like structure, which can accurately identify malaria parasites in blood samples by detecting specific features and/or patterns [5]. They minimize the chances of human error and have the added advantage of faster diagnosis, which can be extremely beneficial in remote or endemic regions with limited healthcare resources (both trained personnel and facilities) [6]. Additionally, AI can also help to analyse medical records and treatment histories of patients to recommend personalized treatment plans based on the disease severity, individual’s genetic factors, ongoing medication, and drug resistance patterns, ensuring that each individual patients receive appropriate and optimal treatment.

The emergence of anti-malarial drug resistance is a growing concern that needs to be tackled. Incorporation of AI technology in this field (i.e., developing DL models based on analysis of vast datasets of chemical compounds using structure-based or ligand-based approaches) is a possible way to identify potential anti-plasmodial candidates. Such AI-assisted approaches can both accelerate the drug detection process and reduce associated costs. The role of AI in malaria treatment extends from early detection and drug development to understanding the parasitic life cycle, which is crucial for developing efficient control strategies [7].

Establishment of a national level database or multiple inter-linked regional level databases of malaria cases, updated at least on a weekly basis and incorporating AI technology with predictive modelling capabilities, can possibly aid in a) early case detection, b) controlling the spread of malaria in a particular region, c) preventing or at-least warn of outbreaks, and d) stave off events of reintroduction. Researchers in South Africa have developed systems capable of forecasting outbreaks up to 9 months in advance with approximately 80% accuracy. Such Machine Learning (ML) models, taking into account factors like sea surface temperatures and historical malaria data, are programmed to assess a combination of overarching environmental, meteorological, and social factors. This can be an invaluable measure to supplement the health system preparedness across the country, enabling authorities to plan for and implement malaria control/prevention measures, such as the appropriate insecticide-treated mosquito net distribution and the strategic deployment of medical personnel.

Effective integration of AI in malaria control strategies and elimination programmes can bring about a paradigm shift. However, its full potential can only be realized by addressing challenges related to data quality and accessibility of the technology. AI algorithms require big data (large datasets) without error (utmost accuracy) to provide precise predictions and analysis. Inaccurate, incomplete, or insufficient data can produce misleading results, which in turn can affect diagnosis, treatment decisions, and intervention effectiveness. This becomes a critical hurdle in a country like India, owing to factors such as limited resources, inadequate training of healthcare personnel, and a lack of standardized data collection protocols, especially in remote/endemic regions. Moreover, the lack of infrastructure for data sharing and inadequate resources for data management limit the accessibility to data and can hinder the setting up of any large-scale databases. This review paper attempts to expand upon the various possibilities and benefits of AI integration in malaria control strategies, drawing on papers published over the past eight years. Moreover, this review incorporates information gathered from a careful review of earlier feedback. The paper includes more recent studies, clearer explanations of methods, and a better arrangement of sections to help readers understand how AI is being used in malaria control (Fig 1).

Fig. 1.

Role of AI in the elimination of malaria

Research questions and article selection strategy

The research questions based on the review paper’s objectives are given in Fig. 2.

Fig. 2.

Categorisation of Research Questions for AI-Driven Malaria Elimination

Research questions

The AI-based research approaches to malaria elimination can be organized into four functional domains: optimization of vector control approaches, population and diagnostic support, health system strengthening approaches, and predictive modelling, such as predicting malaria outbreaks or transmission patterns. These domains are defined through an analysis of 62 studies from 2014 to 2025 and are used to categorize the use of AI in malaria-endemic and epidemic regions. The quantitative findings of the literature show that diagnostic AI tools have better accuracy, predictive models decrease outbreak response time, AI in vector surveillance allocates resources better, and the majority of studies use AI in diagnostics and predictive situations, offering the least amount of commonality in health system changes. All four domains serve to support a more systematized understanding of AI use in the allocation of malaria resources, efficacy of surveillance, and health intervention progressions, which is fundamental to the goal of eliminating malaria in India by 2030.

Article selection strategy

This review was conducted in a systematic manner with the use of a PRISMA approach, providing structure to the identification and selection of studies relevant to the application of AI for ME. The review was conducted in a methodical manner with clearly stated inclusion/exclusion criteria, database search strategies, and protocols for quality assessments to make the review transparent and replicable.

Search strategy

Although the study included research papers published between 2014 and 2025 that focused on the role of AI in malaria elimination and utilized various sources, such as IEEE Xplore, PubMed, Google Scholar, Elsevier, Springer, Scopus, SCIE, and Web of Science, it failed to specify how these databases were systematically searched. A combination of Boolean operators (AND and OR) and keywords was used to create detailed search strings, such as:

("Artificial Intelligence" OR "AI") AND ("Malaria Elimination" OR "Malaria Control") AND ("Larval Source Management" OR "Vector Control") AND ("Zzapp system" OR "Predictive Modeling").

Each search was customized for database-specific syntax. The full search strings used for each database were provided in Appendix A for reproducibility. Table 1 renders parameters for inclusion. Moreover, in the exclusion criteria, the articles that did not focus on the role of AI in malaria elimination were excluded. The articles that were in other languages were not included in the review. The articles that were published before 2014 were excluded. In the exclusion criteria, studies not directly addressing the role of AI in malaria elimination were excluded. Duplicate entries and inaccessible full texts were excluded.

Table 1.

Inclusion criteria

Inclusion criteria
In the inclusion criteria, the research articles published in English between 2014 and 2025 were included
Real-world applications of AI focused only on malaria-related domains: diagnosis, prediction, treatment, vector surveillance, health systems, or public health
The research studies that were published betwixt 2014 to 2025 were included

Sources

The articles were selected by using online databases, including IEEE Xplore, PubMed, Scopus, Springer, Elsevier, and Google Scholar. The search was restricted to articles published in English betwixt 2014 and 2025. The keywords used included combinations of: “AI,” “malaria elimination,” “larval source management,” and “Zzapp system.” These were used in various formats based on "AND" and "OR" to identify a range of papers.

Papers that didn’t focus on the use of AI in malaria control were excluded. Articles that were published before 2014 or were not written in English were also removed. Duplicate entries and papers without full access were filtered out during the screening process. Only full research articles and review papers that discussed real-world AI use in diagnosis, prediction, treatment, vector control, or public health support were kept.

Paper selection

A total of 62 papers were chosen grounded on the criteria given above. The details of the search and the search outcome of this literature survey are depicted in Fig. 3. This review followed a structured method to search and select relevant studies. Research articles were collected from several databases, encompassing IEEE Xplore, PubMed, Scopus, Springer, Elsevier, along with Google Scholar. Only papers published in English between 2014 and 2025 were included. The main keywords used were “AI,” “Malaria Elimination,” “Larval Source Management,” and “Zzapp software system.” Boolean operators, such as AND and OR, were applied during database searches to find matching articles. Duplicate records were removed. Initially, articles were screened by reading the title and abstract. Full-text reading was done for selected records to decide final inclusion. Studies that did not directly focus on AI use in malaria control were removed. Papers that were not accessible or not written in English were also excluded. The entire process was displayed using a PRISMA flow diagram. Figure 4 depicts the PRISMA flowchart.

Fig. 3.

Clustered Bar Graph showing breakdown of the result of the search and number of indexes from different sources and databases

Fig. 4.

PRISMA Flowchart

The study selection process was visualized using the PRISMA 2020 flow diagram (Fig. 4). It outlined the number of records identified, screened, assessed for eligibility, and finally included in the review. All identified articles were imported into reference management software. In the stage 1 process (Title & Abstract Screening), irrelevant studies were excluded based on scope. In the stage 2 process (Full-Text Screening), the remaining articles were assessed against the inclusion criteria. In the final selection, 62 articles were included in the review. To ensure that the findings were reliable, each of the included studies was critically appraised using the Joanna Briggs Institute (JBI) critical appraisal checklists according to study type (diagnostic, modelling, or intervention). Each article was rated as having a Low Risk of Bias, Moderate Risk of Bias, and High Risk of Bias. Low risk of bias studies clearly outlined methodology, used validated models, and provided reproducible results. Moderate Risk of Bias included exploratory studies in which minor methodological weaknesses were discussed. Finally, high-risk of bias studies that contained incomplete methods did not provide validation or resulted in unclear outcomes (Fig 5)

Fig. 5.

AI-based disease Diagnosis and Treatment [8]

Each of the 62 included studies was critically appraised using the Joanna Briggs Institute (JBI) checklists, classified into low, moderate, or high risk of bias. Low-risk studies clearly outlined methodology and validation, moderate-risk studies contained minor design limitations, and high-risk studies lacked clarity in outcomes. A summary table of these appraisals was included in Appendix B, strengthening methodological transparency and reliability of findings.

Research design

In this systematic review, a structured approach on the basis of PRISMA was followed to confirm the consistency and clarity. Investigating how different AI approaches were used to assist efforts to eradicate malaria in the areas of diagnosis, surveillance, prediction, and resource optimization was the systematic review’s key objective. The research also followed a clear theoretical framework that focused on population, interventions, and findings attained when using AI in the elimination of malaria. The following keywords were used in a systematic search of several databases: IEEE Xplore, PubMed, Scopus, Web of Science, and Google Scholar. These keywords included "AI," "ML," "DL," "malaria diagnosis," "malaria prediction," and "vector surveillance”. Inclusion criteria encompassed peer-reviewed articles published betwixt 2014 and 2025, focused on malaria and AI applications. Studies were screened in multiple phases (i.e., title, abstract, and full-text) using predefined eligibility criteria. Also, a focus was given to incorporating evidence from malaria-endemic regions, especially sub-Saharan Africa, Southeast Asia, and South America, to capture the pragmatic challenges and solutions applicable in the real-world context. Besides English-language journals, there was an attempt made to include grey literature and institutional reports of known global health organizations like WHO, PATH, and MESA that served an important role in the development and implementation of AI technologies in malaria management. The search scope included interdisciplinary research that integrated concepts from ML, epidemiology, health informatics, and clinical diagnostics. The review covered both the technological innovation as well as contextual problems that shaped the utilization of AI in malaria elimination programmes by expanding the scope. This broader focus allowed a complete investigation of the uses of AI, its limitations, and its capacity for future use across different health system contexts.

Literature survey: role of AI in the elimination of malaria

An increasing amount of research is being conducted on leveraging AI's potential to enhance treatment strategies, improve data quality and accessibility, and enable early diagnosis of diseases. Therefore, it becomes necessary to identify the AI’s role in malaria control and, ultimately, removal. This section delves into the applications of AI technology to develop advanced tools that can aid in surveillance, diagnosis, treatment, and prevention efforts.

AI-based disease diagnosis and therapy uses natural language processing, DL, and ML algorithms to evaluate complex medical data, including genomic profiles, patient records, and photos, in order to accurately detect diseases. Through image-based parasite detection in blood smears, AI can improve diagnostic accuracy in the setting of malaria. AI can also provide real-time decision support for treatment recommendations based on patient history and regional drug resistance patterns. AI can optimize clinical outcomes, lower diagnostic errors, and facilitate more effective healthcare delivery in both resource-rich and low-resource environments by automating repetitive diagnostic activities and streamlining treatment regimens.

Uses of AI in the elimination of malaria

Malaria detection: Studies deploying AI have shown that the image quality of both thick and thin blood smears can be enhanced using AI [9]. They also depict that AI can be used to precisely classify malaria based on its severity and type [10]. Even though lower parasite-density infections challenge diagnostic accuracy, AI-centric systems can be an appropriate tool to diagnose malaria in non-endemic settings. The employment of this kind of technology in real clinical diagnostic circumstances should be supervised by microscopists. Moreover, the results are understood by perceiving the digital images depicted by the system [55].

By integrating AI-powered technologies into blood smear analysis, healthcare providers can improve patient outcomes, streamline laboratory workflows, and accelerate disease diagnosis and treatment [11].

Disease surveillance and prediction: AI can analyse large datasets from environmental, meteorological, and social factors, namely humidity, temperature, and human mobility, for predicting malaria outbreaks by anticipating where and when outbreaks are likely to occur [12]. Health consultants can take proactive measures for preventing the spread of the disease. The AI application extends beyond infectious disease surveillance to real-time monitoring of several health situations. For example, AI-powered tools are utilized in influenza surveillance, incorporating Google search data into historical illness data to enhance prediction accuracy [56].

Improved diagnosis: AI algorithms can analyse blood smears for detecting the existence of malaria parasites, offering an accurate and faster alternative to manual microscopic examination [13]. This can help to overcome shortages of skilled technicians and reduce human error.

Mobile and point-of-care diagnostics: Smartphone-based AI apps and devices can provide low-cost, rapid diagnostic testing in remote areas with restricted access to healthcare [14]. They also assess Rapid Diagnostic Tests (RDTs) and provide instant results.

Personalized treatment plans & monitoring of drug resistance: AI can be used to analyse patient data and recommend personalised treatment regimens based on the severity of infection, genetic factors, and drug resistance patterns. This ensures that patients receive the most effective therapies. AI can also identify patterns in genetic mutations linked to drug resistance in malaria parasites, enabling early detection of resistance and guiding the selection of appropriate medications. This acceleration in the diagnostic process is of paramount importance in scenarios in which timely treatment can be lifesaving in severe malaria cases. AI-driven approaches have been utilized to predict drug resistance markers in parasites [61].

AI in mosquito population control: AI can optimize the release of genetically modified or sterile mosquitoes to reduce vector populations. ML techniques can also predict the effectiveness of various mosquito control strategies, including insecticide spraying, and suggest the most effective intervention based on local conditions [15]. Also, they can accurately evaluate and forecast mosquito abundance across wide socioeconomic gradients and heterogeneous landscapes, particularly in an urban region. They render an incorporated indication for public awareness of Vector Borne Disease (VBD) risk and decision-making around VBD control efforts [57].

Automated drone-based spraying: AI-powered drones can be used for targeted mosquito control by detecting and spraying breeding sites with larvicides or insecticides, enhancing the effectiveness of vector control efforts [16].

Supply chain optimization: AI can help to ensure the timely and efficient delivery of essential supplies, such as diagnostic tools, medications, and mosquito nets, by optimizing the supply chain management, predicting demand, and reducing waste [17].

Telemedicine and AI-driven decision support: In areas with restricted healthcare access, AI assists healthcare workers in diagnosing malaria, providing treatment guidance, and ensuring adherence to best practices [18]. This also helps to reduce the strain on healthcare systems.

Monitoring and evaluation of interventions: AI can track the effectiveness of various malaria control strategies, allowing for real-time adjustments and improvements to ongoing programmes [19, 20]. An AI approach can serve as the foundation for creating real-time malaria monitoring and control systems, evolving from systems based on periodic reports [60]. For India, AI-driven microscopy can supplement the shortage of skilled microscopists in remote districts, particularly in Odisha, Chhattisgarh, and the Northeast, where malaria remains concentrated.

Comparative overview of AI techniques for malaria elimination

Effective introduction of AI technology into malaria programmes can optimize resource allocation, thereby aiding in accelerating the efforts towards the elimination of the disease [21]. A technical framework that employs numerous learning algorithms for eliminating malaria is termed ensemble learning [22]. For instance, there is a lacuna in the detection of malaria in areas with minimal burden, and the prognosis often doesn’t go beyond a fever instance [23]. Some studies highlighting the diverse applications of AI techniques in malaria elimination, ranging from improving diagnostic accuracy and speed to predicting outbreaks and optimizing intervention strategies, are given in Table 2.

Table 2.

Studies of a comparative overview of AI techniques for ME

Authors’ Name	Aims	AI Techniques	Findings	Advantages
Peter & Clement [24]	To maximise malaria detection accuracy and energy consumption,	Basic Conventional Neural Network (BCNN)	99% recall	High sensitivity with resource efficiency
Oliver, et al. [25]	To explore the malaria endemic in Africa, as well as its simultaneous socioeconomic limitations	Conventional Neural Network (CNN)	96.5% accuracy	Robust performance across diverse datasets
Matheus, et al. [26]	To study the probability of malaria cases and to group cities centered on malaria incidence similarity	Long Short-Term Memory (LSTM)	Improved performance in clusters with less variability	Adaptive to spatial–temporal variability
Gede, et al. [27]	To introduce the data-driven computer-aided diagnostic approach and eliminate malaria	Deep Conventional Neural Network (DCNN)	95.2% accuracy on field samples	Scalable for real-world deployment
Sammy, et al.[28]	To analyse the images by learning image patterns and recognising malaria	CNN	96% accuracy	Fast, non-expert dependent analysis
Kristofer, et al. [29]	To focus on an intelligent system for discovering malaria parasites	CNN	95.2% sensitivity	Suitable for mobile/remote diagnostics
Thomas, et al. [30]	To explore building a cross-border early warning system	LSTM	95% Confidence Interval	Facilitated regional early-warning systems
Muthoni Masinde [31]	To identify the simulator function in Rapid Miner	Artificial Neural Network (ANN)	High performance via Gradient Boosted Trees	Tool-agnostic deployment capabilities
Hilary, et al. [32]	To analyse the different data mining models and diagnose malaria	Decision Support System (DSS)	Specificity of 0.982	Decision-support in clinical workflow
Ajeet, et al. [33]	To develop stability, which was necessary in the robust model	Neural Network	Reliable model performance	Robust under variable conditions
Eric & Jijun [34]	To explore climatic factors’ impact on malaria re-emergence	LSTM	87.3% prediction accuracy	Integrated environmental variability

This comparative synthesis showed that while CNNs dominated diagnostic applications, lightweight models, such as MobileNet, hold more promise for rural India. LSTM models were relevant for integration with India’s surveillance infrastructure, whereas ensemble methods required higher resources and were better suited for research hubs. This sharper analysis avoided redundancy and directly addressed applicability within India’s elimination roadmap.

Advanced tools for malaria elimination programmes

High transmission rates in malaria-endemic areas necessitate innovative tools and strategies to complement existing interventions. One such recent tool is supplementing Larval Source Management (LSM) with the Zzapp software system [35]. LSM, which targets mosquito larval habitats, aims at controlling the malaria vector population [36] [37]. It is regarded as an effective method for malaria vector control [38]. The majority of the malaria-endemic nations mention LSM, mainly larviciding, in their strategic plans. The LSM’s status within the WHO guidelines might be limiting since it diminishes the strategy to a low priority compared to other interventions, meaning that there is potential for countries and donors to consider it a low priority, in which funds are limited [58]. The Zzapp system is a digitally managed intervention tool that is incorporated into this system and has shown potential as an economical approach for urban malaria control [39]. It incorporates operational quality assurance methods for effective field management. The larval positivity data can be simulated by the Zzapp system for creating a spatial model of malaria transmission in a given village/town [39].

Wolfgang et al. described the use of drones for area-wide LSM of malaria mosquitoes [40]. These drones equipped with larvicides demonstrated significant reductions in larval counts. Drones were deployed with an Aquatain Mosquito Formulation (AMF) to identify and curtail African malaria mosquitoes, Anopheles arabiensis, in an irrigated rice agro-ecosystem in Unguja Island, Zanzibar, Tanzania, leading to a highly substantial (p < 0.001) reduction in larvae count [40]. Studies combining drones and smartphones have shown the capability to simultaneously reduce costs and improve the operational quality [41]. The research employs a Spatial Intelligence System (SIS) by weighing key mapping accuracy indicators of an LSM programme for four months. Larvicidal activities employing SIS have been proven capable of attaining about 90% credible intervals for cost per km [41].

Application of AI techniques in malaria elimination

The period between 2000 and 2018 globally witnessed a substantial reduction in malaria mortality. The aim now must be to monitor receptivity, vulnerability, and health system capacity in order to sustain this decline in malaria burden and to prevent reintroduction [42]. Climatic factors influence the incidence as well as the spread of malaria. AI-powered predictive models incorporating environmental and climatic factors, namely wind, humidity, location, temperature, and floods, can help to identify hidden ecological factors contributing to malaria outbreaks [43]. Studies using AI-based systems that analyse malaria surveillance data and climate data have been able to detect and forecast outbreaks [44]. This early warning capability allows for the timely deployment of resources and interventions. Over the past few years, AI-powered image analysis has evolved as a transformative tool for malaria diagnosis, providing clear-cut advantages over traditional microscopy. DL models, particularly CNNs, are increasingly employed for accurate and efficient parasite detection in microscopic images [13]. Such AI-based automated image analysis provides several benefits, namely increased accuracy and speed, reduced reliance on skilled technicians, and improved chances for early-stage detection. In one of the studies [59], the CNN approach exhibits better classification performance subsequent to training with more than 27,000 images. Therefore, DL significantly improves the working efficacy and accuracy of malaria diagnosis and other health-correlated applications [59].

MobileNet V2, an example of this deep ML model, demonstrates high prediction and detection accuracy for malaria [24]. Additionally, its low storage memory requirement and fast response time make it suitable for field-level applications.

AI is also being used to speed up the detection of new anti-malarial drugs. Traditional drug discovery methods, such as High-Throughput Screening (HTS) of larger compound libraries, are laborious and resource-intensive [4]. AI offers a faster and more efficient approach to drug discovery by employing DL models for virtual screening. Tools like Graph Convolutional Neural Networks (GCNNs) enable virtual in-silico analysis of target compounds by using ligand-based or structure-based approaches, which significantly accelerates the identification of potential drug candidates [45]. AI can also be used to predict the anti-Plasmodium activity of potential components by analysing their Simplified Molecular Input Line Entry System (SMILES) [4]. A few studies that applied AI techniques in the area of malaria control as well as elimination are outlined in Table 3. Certain studies also highlighted the application of AI in the realm of surveillance systems to work towards malaria elimination. It was found that the ideal systems envisaged through AI demonstrated better distribution of values, indicating their effectiveness [51]. However, the study also highlighted that AI faced difficulties in case identification due to data classification limitations. Another AI-based study, which conducted an analysis of vector surveillance programmes employed by various national malaria control initiatives, noted significant variations in monitoring practices and their impact on malaria elimination efforts. However, the analysis of metadata on vector control activities of individual programmes found that the data management tools used for recording and presenting data were inadequate to conduct large-scale AI-centred studies [52]. Data quality and availability were also identified as critical aspects for the effective implementation of AI technology. A study conducted in Nigeria assessing these factors found that even though the local District Health Information System (DHIS) performed at approximately 80%, gaps were identified in data availability [53]. Studies employing AI techniques had also been conducted to arrive at a decision as to which malaria control intervention should be adopted based on the financial resources available [54]. It involved the development of a mathematical approach to malaria transmission to evaluate the cost-effectiveness of various interventions. Overall, AI had the potential to enhance surveillance and treatment optimization, but effectiveness always depended on the quality of the underlying data, the infrastructure, and how it was integrated with current systems. Many limitations were a result of incomplete data, misclassification, and limited system integration; thus, effective data governance and investment in digital health infrastructure must be developed. Addressing these data-related challenges was critical to realize the capability of AI in achieving a malaria-free world. Outbreak forecasting tools were especially relevant to India’s highly seasonal malaria transmission, where climatic variability across states could be integrated into predictive surveillance.

Table 3.

Applications of AI techniques in the elimination of malaria

Authors’ Name	Aims	Applications	Findings	Limitations
Ousman, et al. [46]	Addressing the challenges by establishing a predictive approach for malaria outbreaks was the goal	Prediction of malaria outbreak	Average accuracy:79.1%	Data imbalance
Colborn, et al. [47]	The aim was to predict the advanced areas with elevated risk for malaria transmission as well as measure outbreaks	Early epidemic detection of malaria	Correlation time performance was good	Negative estimates of incidence in the event
Gerardin, et al. [48]	The aim was to set individual infection trajectories to construct the household in each catchment area	Optimal Treatment Recommendations	The mean of 100 stochastic realisations per coverage level was good	Operational limitations in infection strategy
Xiangli, et al. [49]	The aim was to perform the individual cases from a web-centric reporting system and to discover malaria endemic characteristics	Surveillance Systems	The epidemiology study of the infection sources of every single case was clearly clarified	Persuasive strategies: insufficient
Mercado, et al. [50]	The aim was to concentrate on the regional effort to support populations	Surveillance Systems	National surveillance rate: 64%	Limited dataset

Challenges and limitations of AI in malaria elimination

• 1. Data quality and availability: AI algorithms require high-quality, relevant data to produce accurate results. However, data collection and management in rural areas can be challenging.

• 2. Infrastructure and connectivity: AI-powered systems require reliable internet connectivity, electricity, and hardware, which can be scarce in rural areas.

• 3. Cybersecurity concerns: AI systems can be vulnerable to cyberattacks, compromising sensitive patient data and disrupting healthcare services.

• 4. Dependence on technology: Over-reliance on AI can lead to a decline in traditional skills, such as manual diagnosis, and cause a shortage of skilled healthcare professionals.

• 5. Ethical concerns: AI decision-making can raise ethical concerns, such as bias in algorithms, unequal access to healthcare, and potential misuse of patient data.

By acknowledging both the benefits and challenges of AI in malaria elimination, the review addresses limitations and ensures responsible and effective use.

Comparative view of AI techniques in malaria elimination

There are different AI methods that vary in performance and suitability depending on their application. CNN models work well for image classification in malaria diagnosis, particularly for blood smear analysis. They are best suited to lab environments but may not work well with low-quality field data. LSTM models are useful in outbreak prediction because they effectively handle time-based data. They are valuable in regions with weekly or monthly surveillance data. MobileNet models are lightweight and suitable for mobile-based tools in remote areas. Ensemble methods can combine multiple models to achieve better accuracy. But, they require more computing power. Lightweight mobile-based models should be prioritized in India’s resource-constrained districts, ensuring that AI complements national diagnostic and surveillance priorities.

Framework for AI-based malaria elimination

The reviewed literature highlights diverse AI applications across diagnosis, prediction, surveillance, vector control, logistics, and drug discovery. However, for the framework to be operationally relevant within India’s malaria elimination roadmap (aligned with the National Vector Borne Disease Control Programme [NVBDCP] and WHO’s E-2030 goals), AI must be positioned as a translational tool that links research outcomes to public health practice.

Diagnosis and case management Convolutional Neural Networks (CNNs) and lightweight models, such as MobileNet, can support automated blood smear diagnosis at district laboratories and community health centres. Pilot work in India (e.g., AI-based microscopy tools [19]) demonstrates feasibility. The framework proposes integration of these tools into India’s tiered diagnostic network under NVBDCP, reducing dependence on limited human microscopists.

Prediction and surveillance Long Short-Term Memory (LSTM) and hybrid models have been shown to forecast outbreaks with accuracies of 80–90% in African settings. Within India, these can be embedded in the Integrated Disease Surveillance Programme (IDSP) to generate state-level early warnings, particularly for tribal and forested districts, where malaria resurgence remains a risk.

Vector control AI-driven drones and larval mapping systems [40] can be integrated with India’s existing Larval Source Management (LSM) guidelines. State malaria units can adopt AI-assisted geospatial models to prioritize high-risk habitats and improve resource allocation for vector control.

Supply chain and logistics Predictive AI models can optimize the distribution of diagnostic kits, antimalarial drugs, and insecticide-treated nets across India’s endemic states. These systems can complement NVBDCP’s logistics channels by reducing wastage, improving timely replenishment, and ensuring that stock-outs in rural PHCs are minimized.

Feedback and policy loop The framework introduces a continuous feedback cycle in which diagnostic, surveillance, and vector data are linked through digital dashboards accessible to policymakers. This ensures that AI outputs directly inform operational decisions at district, state, and national levels. By embedding AI functions into existing national structures rather than as stand-alone tools, the revised framework shifts from a schematic model to a contextual roadmap for India. It demonstrates how global evidence can be realistically localized, ensuring alignment with India’s malaria elimination strategy for 2030.

Conclusion

This comprehensive review explored the shifting role of artificial intelligence in the push towards malaria eradication, particularly in endemic and resource-poor contexts. The final findings of this research pointed out the ways to improve the operational efficiency of public health interventions. It was also found that AI indicated an effective health performance beyond any purely technological attributes through emerging real-time cost-effective surveillance, early detection, and improved vector control methods. From a practice perspective, the review also identified that there were multiple ways for the integration of AI into health structure workflows (e.g., mobile diagnostics and decision support systems). This could further provide useful innovations to overcome concerns relating to health personnel shortages and patient flow accessibility. In relation to public health policy, research showed how AI forecast tools and spatial intelligence systems could lead to targeted interventions, including appropriate resourcing for intervention deployment, rendering alignment for national malaria control plans with pressing epidemiological contexts. Without enabling these policy factors, the issues facing scalability and long-term sustainability for AI solutions remained elusive. Finally, future research should design lightweight and interpretable AI algorithms for persistent health challenges with low resource settings, investigate more hybrid approaches that combine AI with traditional local knowledge or approaches, and design participatory frameworks that involve local stakeholder input to provide a responsible AI design and deployment solution.

Author contributions

M: Original manuscript writing, concept development, literature search, editing N: Manuscript editing, quality monitoring, proof reading Ahmed: Manuscript editing, literature search Negi: Manuscript editing, literature search SB: Manuscript editing, literature search Tomer: Manuscript development, formatting, literature search Bankoti: Manuscript development, formatting, literature search HC: Overall monitoring, proof reading BM: Overall monitoring, proof reading Warkade: Literature search.

Funding

Not applicable: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.