Endometrial cancer (EC) is the most common malignancy of the female reproductive system, originating in the lining of the uterus. It accounts for roughly 3% of all cancer diagnoses among women globally, with an estimated 417,000 new cases and more than 97,000 deaths reported annually since 2020. The disease is divided into two main categories. Type I ECs make up about 80% of all cases, are typically estrogen-dependent low-grade tumors, and carry a favorable prognosis with over 80% five-year survival rates. Type II ECs, by contrast, are high-grade tumors that progress rapidly, resist conventional therapies, and account for the majority of EC-related deaths.
The treatment gap: Despite their far more aggressive behavior, Type II ECs are still being treated with essentially the same protocols used for Type I ECs. Currently, no specific targeted therapies exist for Type II ECs. Type II ECs include three main histological subtypes: serous adenocarcinomas (10-20% of cases), clear cell adenocarcinomas (less than 5%), and carcinosarcomas (less than 5%). They are associated with molecular markers such as TP53 mutations, HER2/neu amplification, and PI3K/AKT/mTOR pathway alterations, all of which drive aggressive tumor behavior and limited hormonal responsiveness.
Scope of this review: Published in Frontiers in Oncology in 2025 by researchers at the University of KwaZulu-Natal and the University of Pretoria in South Africa, this review examines how artificial intelligence can transform the care of Type II ECs. The authors explore AI-driven diagnostic tools, tailored therapeutic options, and innovative projects that aim to address healthcare disparities, particularly across Africa. The paper uniquely integrates a global perspective on AI in oncology with a focused analysis of African healthcare contexts.
The AI opportunity: AI technologies, including machine learning, deep learning, and convolutional neural networks (CNNs), can process vast amounts of clinical, imaging, and genomic data. CNNs have been used to enhance imaging analysis of MRI, CT, and PET scans, detecting abnormalities with precision comparable to or exceeding that of experienced radiologists. AI also accelerates drug discovery by identifying promising candidates and predicting molecular-level interactions, making cancer treatment research more efficient.
The epidemiology of Type II EC in Africa is shaped by a complex interplay of demographic, socioeconomic, genetic, and healthcare system factors. Although EC is generally less common in sub-Saharan Africa than in more developed regions, the incidence of Type II EC is rising globally, driven by longer lifespans and, indirectly, the obesity epidemic. The mortality rates from Type II EC are disproportionately high relative to its incidence, owing to its aggressive phenotype with rapid progression and high propensity to metastasize.
Underestimation and underreporting: The true incidence of Type II EC in Africa is almost certainly underestimated. Many African countries lack formal cancer registries, and diagnostic capacity for malignancies remains low. The disease is often diagnosed at advanced stages when treatment options are far less effective. Routine screening for EC is not currently recommended for asymptomatic women, which means Type II ECs are frequently discovered incidentally or when symptoms of late-stage disease appear.
Genetic predisposition in women of African descent: Research by Sponholtz and coworkers has shown that women of African descent are at higher risk of developing Type II EC compared to white women. This elevated risk is thought to have a genetic basis, as alterations in genes like TP53 and PTEN are commonly found in Type II EC. Genetic syndromes such as Lynch Syndrome, characterized by mutations in mismatch repair genes, also increase EC risk. Understanding these predispositions is critical for improving early detection and treatment outcomes for African women.
Treatment barriers: Type II EC is notorious for its resistance to standard chemotherapy, which is often the only treatment option available since many patients present with metastatic disease at diagnosis, making surgery and radiation non-viable. Targeted therapies under global development, such as HER2 inhibitors and PI3K/AKT/mTOR pathway inhibitors, show promise but would likely be unaffordable in many African countries. The standard of care in African regions still heavily relies on conventional chemotherapy with limited use of personalized medicine or companion molecular diagnostics. When Type II EC is diagnosed and treated at an early stage, the 5-year survival rate is significantly higher, underscoring the urgent need for improved detection.
Artificial intelligence encompasses a range of technologies that enable computers to learn from data, identify patterns, and make decisions with minimal human intervention. In oncology, the three primary AI approaches are machine learning (ML), which trains algorithms on large datasets to identify patterns and make predictions; deep learning (DL), a subset of ML that uses artificial neural networks to model complex relationships, making it particularly effective for interpreting medical images and genomic data; and natural language processing (NLP), which extracts valuable information from unstructured data sources such as electronic health records and scientific literature.
AI in diagnosis: Deep learning algorithms can analyze mammograms, CT scans, and MRIs to detect abnormalities with precision comparable to or exceeding that of experienced radiologists. In pathology, AI aids in the analysis of histopathological images to identify cancerous tissues, reducing human error and improving diagnostic consistency. For Type II EC specifically, CNNs have shown strong performance in analyzing endometrial biopsies, differentiating between Type I and Type II subtypes by recognizing subtle histological patterns that human pathologists may miss.
AI in treatment planning and prognosis: AI tools assist oncologists in personalizing therapy by integrating and analyzing complex datasets including genomic profiles, patient histories, and clinical trial data. AI models can predict which therapies will be most effective for individual patients based on their tumors' molecular characteristics. AI algorithms can also optimize radiation therapy planning by accurately segmenting tumors and critical structures, minimizing exposure to healthy tissues. For prognosis, AI identifies patterns across large-scale datasets that correlate with specific outcomes, aiding risk stratification and treatment decisions.
AI in drug discovery: AI hastens drug discovery by identifying promising drug candidates and predicting their interactions at a molecular level. This is especially relevant for Type II EC, where no specific targeted therapies currently exist. By analyzing large datasets of genomic, transcriptomic, and proteomic information, AI algorithms can discover biomarker patterns indicating early-stage cancer, potentially enabling non-invasive screening tests that lead to earlier diagnosis.
Performance evaluation: AI model reliability is typically measured using sensitivity (ability to correctly identify cancer), specificity (ability to correctly rule out cancer), overall accuracy, and the Area Under the Receiver Operating Characteristic Curve (AUC). These metrics collectively reveal how well a model performs in real clinical scenarios, where both false positives and false negatives carry significant consequences for patients.
The deployment and effectiveness of AI-based oncological prediction models in Africa vary significantly across regions, reflecting the continent's diverse socio-economic, infrastructural, and healthcare contexts. The authors present a timeline of oncological AI tools applied in Africa, spanning from 2018 to 2022. These include machine learning for breast cancer staging (Hamouda et al., 2018), deep learning for contouring clinical treatment volumes in cervical cancer radiotherapy (Rhee et al., 2020), NLP for identifying malignant cases in electronic records (Achilonu et al., 2021), and machine learning for cell-free DNA-based esophageal cancer diagnosis (Kandimalla et al., 2021), among others.
Concentration of efforts: AI development for oncology in Africa is concentrated in nations like South Africa, Egypt, Nigeria, and Kenya, where technological capabilities and resources are more advanced. This leaves approximately 90% of African countries, particularly in Central Africa and other low-income or lower-middle-income regions, with limited AI deployment for cancer care. The AI tools that do exist have focused predominantly on improving outcomes for breast, cervical, and colorectal cancers, aligning with the continent's most prevalent cancers.
The Type II EC gap: Despite several AI tools being used in the global north that continue to make oncology more precise and personalized, there are currently limited AI-driven platforms under development or in use for managing Type II EC in Africa. While EC models have been suggested primarily to support incidence reporting through cancer registry data, the authors argue that Type II EC models would have the greatest impact on diagnostic or treatment processes. Type II EC has higher age-standardized incidence rates in Africa than in other regions, making it a particularly strong candidate for AI-powered predictive tools.
Barriers to adoption: Compared to the global north, Africa faces underdeveloped technological and digital infrastructure, unstandardized data collection and storage protocols, limited healthcare resources, unregulated frameworks governing AI use in healthcare, and a shortage of healthcare professionals with AI expertise. These factors collectively limit the adoption and effective use of AI technologies across the continent. The authors note that most AI models are trained on data from high-income countries and may not reflect the genetic, environmental, and socioeconomic factors prevalent in African populations.
AI is revolutionizing histopathology by enabling more accurate and efficient analysis of tissue samples for Type II EC. Convolutional neural networks (CNNs) are specifically designed to detect spatial features in visual data, enabling accurate identification and classification of cancerous cells. These AI tools can differentiate between Type I and Type II ECs by recognizing subtle histological patterns that may be missed by human pathologists. Studies have demonstrated that AI-based image analysis can significantly improve diagnostic accuracy, reduce interobserver variability, and speed up the diagnostic process.
Integration with imaging modalities: AI is being integrated with MRI and CT scans to improve Type II EC diagnosis. AI algorithms can analyze complex imaging data to identify tumors, assess their size and extent, and differentiate between benign and malignant lesions. This integration helps in better staging of the cancer, which is crucial for treatment planning. Additionally, AI-powered imaging enhances the detection of metastasis, improving the overall diagnostic workflow. These advancements are especially critical in resource-limited settings where access to highly trained specialists may be scarce.
Personalized treatment planning: AI plays a pivotal role in developing personalized treatment plans for Type II EC by analyzing patient data including genetic profiles, tumor characteristics, and response to previous treatments. AI models consider factors such as the likelihood of treatment success, potential side effects, and patient preferences to optimize outcomes. Machine learning models can analyze data from clinical trials and real-world patient outcomes to predict the effectiveness of chemotherapy, radiation therapy, and immunotherapy, helping oncologists choose the most appropriate treatment for each patient.
Risk stratification and prognosis: AI tools are proving invaluable for risk stratification in Type II EC management. AI models analyze a wide range of patient data, including genetic information, tumor characteristics, and lifestyle factors, to classify patients into different risk categories. This stratification helps in tailoring treatment strategies and identifying patients who may benefit from more aggressive treatment or closer monitoring. AI-based prognostic models that combine clinical, pathological, and molecular data can generate individualized predictions for cancer recurrence and overall survival, guiding post-treatment surveillance and adjuvant therapy decisions.
Infrastructure gaps: The technological infrastructure in Africa for supporting AI in healthcare is still in its early stages, with significant disparities across the continent. While some urban centers in South Africa, Kenya, and Nigeria have started integrating AI into healthcare, many regions lack high-speed internet, advanced computing power, and robust data storage facilities. In low-resource settings, additional barriers include limited access to advanced diagnostic tools, inadequate healthcare funding, and a shortage of trained professionals who can operate AI systems. The high cost of AI technology and its maintenance creates a digital divide where only a few well-funded institutions can benefit.
Data scarcity and quality: A significant challenge is the scarcity of high-quality, large-scale datasets. Many African healthcare facilities lack the digital infrastructure to collect, store, and manage vast amounts of patient data. Existing data are often fragmented, incomplete, or inconsistent, making it difficult to train robust AI models. The lack of standardized data collection protocols further compounds this problem. Most AI models are trained on data from high-income countries that may not reflect the genetic, environmental, and socioeconomic factors prevalent in African populations.
Algorithmic bias: AI models trained on non-representative datasets can perpetuate biases, leading to disparities in healthcare outcomes. Systems developed using Western population data may not accurately predict disease risk or treatment responses in African populations, potentially resulting in misdiagnoses or inappropriate treatment plans. This can exacerbate existing healthcare inequalities rather than reducing them. Ensuring that AI models are trained on diverse, representative datasets that include African genetic and epidemiological data is essential.
Ethical and regulatory concerns: The use of AI raises issues around algorithmic bias, lack of transparency in decision-making, and the risk of deepening existing inequalities. There is an ethical dilemma in relying on AI for critical healthcare decisions in settings where human oversight may be limited. Regulatory frameworks for AI in healthcare are still in early stages in most African countries, and the absence of regulation raises concerns about the quality and safety of AI applications. Data protection laws may be less stringent or poorly enforced, posing risks to patient privacy and consent. South Africa's Protection of Personal Information Act (POPIA), modeled in part on the EU's GDPR, represents one example of how legal standards for consent, transparency, and responsible data use can be embedded into national policy.
African-led AI initiatives have gained significant momentum in recent years. The African Institute for Mathematical Sciences (AIMS) has been instrumental in fostering AI research across the continent, with researchers developing AI models that analyze large datasets for early cancer detection and improving diagnostic accuracy in resource-limited settings. The University of Cape Town's Data Science for Social Impact Research Group has applied machine learning techniques to analyze cancer data and predict patient outcomes. These initiatives demonstrate growing capacity within Africa to lead AI research that is relevant to the continent's unique needs.
Locally developed tools: In Kenya, researchers from Strathmore University developed an AI-based tool for early detection of cervical cancer, using machine learning to analyze medical images and identify early signs of cancer. In Nigeria, the Health AI Initiative has developed AI-powered software that assists oncologists in planning treatment protocols for cancer patients, including those with Type II EC. This software is designed to work within the specific healthcare infrastructure available in Nigerian hospitals. Additionally, African developers are creating mobile-based AI applications to improve accessibility in remote areas, allowing healthcare workers to input patient data and receive AI-generated diagnostic or treatment recommendations.
Global collaborations: The AI4D Africa program, supported by the International Development Research Centre (IDRC) and the Swedish International Development Cooperation Agency (Sida), has funded numerous projects leveraging AI to address health challenges including cancer. A collaboration between the University of Lagos and the Massachusetts Institute of Technology (MIT) focuses on developing AI-driven diagnostic tools for cancer detection. These partnerships provide African researchers with access to advanced technology and expertise while ensuring that solutions are relevant to the African context.
Indigenous knowledge integration: A unique opportunity for innovation in Africa lies in incorporating indigenous knowledge systems into AI-driven healthcare. Traditional medical practices and understanding of local disease patterns can enhance AI model effectiveness. AI systems can be trained using data that includes information on traditional remedies and local health practices, allowing culturally appropriate recommendations. Africa's rich genetic diversity presents a further opportunity: genomic studies in African populations may uncover novel genetic markers associated with Type II EC that are absent in other populations, leading to new diagnostic methods or targeted therapies specific to African patients.
Alignment with Sustainable Development Goals: The authors map AI-powered advances in Type II EC care onto multiple SDGs. SDG 3 (Good Health and Well-being) is supported by AI tools that detect aggressive cancers at earlier stages and develop tailored therapies to improve survival rates. SDG 5 (Gender Equality) benefits because endometrial cancer affects women exclusively, and AI tools can make screening and diagnostics more accessible, helping bridge gender-based health disparities. SDG 9 (Industry, Innovation and Infrastructure) is advanced through investments in digital health infrastructure such as high-speed internet, cloud computing, and data storage needed to support AI applications.
Reducing inequalities and building partnerships: Under SDG 10 (Reduced Inequalities), the authors argue that incorporating African-specific genetic and epidemiological data into AI training can develop more equitable models that provide effective care for African patients. SDG 17 (Partnerships for the Goals) calls for collaboration between governments, international health organizations, research institutions, and private sector stakeholders. International organizations can assist AI integration in Africa by investing in infrastructure, capacity building through training programs, developing local AI talent, and creating Afrocentric regulatory frameworks.
Policy recommendations: The authors call on African policymakers to invest in technological infrastructure (high-speed internet, data storage), promote development of local AI talent through funded education and training programs, and establish regulatory frameworks ensuring safe and ethical AI use. These frameworks must address data privacy, security, and patient consent, along with standards for validation and approval of clinical AI algorithms. Public-private partnerships are essential to drive innovation and scale AI solutions continent-wide. The establishment of regional centers of excellence and pan-African networks for cross-border cooperation in AI research are also recommended.
Sustainable implementation strategies: Sustainable AI development requires multiple approaches: regional centers of excellence as hubs for innovation, collaboration between African countries to share knowledge and best practices, and involvement of local communities in AI development and deployment. The authors emphasize that AI technologies must be culturally appropriate and equitably distributed. They envision a future where AI facilitates telemedicine and mobile health (mHealth) initiatives, integrates with genomics and bioinformatics for personalized treatment, and enables context-specific solutions trained on local datasets reflecting Africa's genetic and environmental diversity.
Conclusion: The authors conclude that with continued research, investment, and collaboration, AI can revolutionize how Type II EC is diagnosed, treated, and managed across Africa. AI-driven innovations can lead to earlier diagnoses, more personalized treatment plans, and improved survival rates. The successful implementation of AI in Africa could serve as a model for other low-resource settings globally, ensuring that advances in medical technology benefit all populations regardless of geography or socioeconomic status.