AI in breast cancer treatment: Promise, progress and gaps

Breast cancer remains the most commonly diagnosed cancer among women worldwide, with millions of cases each year and increasingly complex treatment pathways requiring personalized care. AI technologies, ranging from machine learning models to conversational chatbots, are now being deployed to support clinical decision-making, patient education, and long-term monitoring. However, the integration of AI into patient-centered care remains limited and uneven, according to a new study published in the International Journal of Environmental Research and Public Health.

The study, titled "Applications of Artificial Intelligence (AI) in Breast Cancer Care Delivery and Education: A Scoping Review," maps the growing use of AI technologies across the post-diagnosis breast cancer care pathway, revealing both rapid innovation and major structural gaps in how these tools are applied.

AI dominates clinical decision-making but remains provider-focused

The review analyzed 54 studies published between 2016 and 2024, drawn from a pool of nearly 4,000 records, to understand how AI is being applied after a breast cancer diagnosis. The findings show that machine learning overwhelmingly dominates the field, accounting for more than 80 percent of AI applications.

These tools are primarily designed to assist healthcare providers rather than patients. Around 83 percent of AI systems identified in the study are clinician-facing, supporting tasks such as recurrence prediction, survival forecasting, treatment planning, and workflow optimization.

One of the most prominent applications is recurrence prediction, particularly during follow-up and surveillance. Nearly half of all studies focus on this stage, where AI models analyze clinical, imaging, and genomic data to estimate the likelihood of cancer returning. These predictive systems are intended to guide monitoring strategies and inform treatment decisions, helping clinicians tailor care to individual risk profiles.

AI is also widely used in treatment planning. Models are being developed to predict long-term survival outcomes, assess disease progression risks, and optimize clinical workflows. Some systems forecast hospital stays after surgery, while others automate aspects of radiotherapy planning or extract relevant clinical data from patient records to streamline decision-making.

During treatment delivery, AI tools are applied to predict adverse effects and treatment responses. Neural networks analyze electronic health records to anticipate complications, while other models evaluate genetic mutations to guide therapy choices. These applications highlight AI's growing role in enhancing precision medicine by aligning treatments with patient-specific characteristics.

Despite these advances, the study notes that most AI systems remain in early development stages. The majority are based on retrospective data and lack real-world validation, limiting their immediate clinical impact. Integration into routine care workflows is still rare, and few studies explore how clinicians interpret or act on AI-generated insights.

Patient-facing AI tools lag behind despite growing demand

While provider-focused applications dominate, the study identifies a smaller but growing segment of patient-facing AI tools. Only 17 percent of studies examine technologies designed to directly support patients, highlighting a significant imbalance in how AI is deployed.

These tools fall into two main categories: conversational agents and generative AI systems. Conversational agents, often rule-based chatbots, are used in structured settings such as pre-treatment education or symptom monitoring during therapy. For example, some systems guide patients through biopsy preparation or collect real-time feedback on treatment side effects.

Generative AI, particularly large language models like ChatGPT, is emerging as a tool for patient education during survivorship. These systems can answer complex questions about breast cancer, treatment options, and recovery, often providing more detailed and accessible information than traditional search engines.

The study highlights the potential of these tools to improve health literacy and patient engagement. By translating complex medical information into understandable language, AI systems can help patients make informed decisions and better manage their care.

However, the findings also raise concerns about reliability and safety. Generative AI systems are not constrained by predefined rules and may produce inaccurate or inconsistent information. This creates risks in high-stakes medical contexts where misinformation can have serious consequences.

Another critical issue is the growing trend of unsupervised AI use. Patients are increasingly turning to publicly available AI tools for health advice without clinical oversight. The study notes that this phenomenon remains largely unexamined in existing research, representing a major gap in understanding how AI influences patient behavior and decision-making.

Even in structured applications, patient-facing tools are rarely integrated into broader care pathways. Most are tested in controlled environments rather than real-world clinical settings, limiting their ability to deliver consistent and meaningful impact.

Uneven development leaves gaps in survivorship and palliative care

The study reveals significant disparities in how AI is distributed across different stages of breast cancer care. While follow-up and surveillance receive the most attention, other critical phases remain underdeveloped.

Survivorship care, which addresses the long-term physical and psychological needs of patients, accounts for only a small share of AI applications. Although generative AI tools show promise in this area, their use is still limited and largely focused on information delivery rather than comprehensive care support.

The most striking gap is in palliative care. The review finds no studies applying AI to end-of-life care for breast cancer patients, despite the complexity and importance of this stage. This absence highlights a broader issue in AI development, where technological innovation tends to focus on measurable clinical outcomes rather than holistic patient needs.

Geographic disparities further complicate the picture. Nearly three-quarters of the studies originate from high-income countries, reflecting differences in infrastructure, funding, and data availability. This concentration raises concerns about the generalizability of AI tools, particularly in low-resource settings where healthcare systems and patient populations differ significantly.

The reliance on data from specific populations also introduces the risk of algorithmic bias. AI systems trained on homogeneous datasets may perform poorly when applied to diverse groups, potentially exacerbating existing health inequalities.

Rapid growth signals potential but raises governance challenges

The study documents a sharp increase in AI research in breast cancer care over recent years, with the majority of studies published after 2020. This growth reflects broader trends in healthcare innovation, driven by advances in computing power, data availability, and the rise of generative AI technologies.

Machine learning models remain the backbone of AI applications, particularly in predictive analytics. However, newer technologies such as large language models are beginning to reshape patient-facing care, introducing new opportunities and challenges.

The findings underscore the need for stronger governance frameworks to guide AI integration. Policymakers and healthcare systems must distinguish between clinically validated tools and general-purpose AI systems, ensuring that each is used appropriately within care pathways.

The study also calls for more prospective research to evaluate AI tools in real-world settings. Most existing studies focus on model development rather than implementation, leaving critical questions unanswered about effectiveness, usability, and patient outcomes.

Last but not least, issues such as data privacy, informed consent, and accountability for AI-generated decisions must be addressed as these technologies become more widespread. Ensuring transparency and fairness in AI systems is essential to building trust among both clinicians and patients.