Generative AI in Medicine

Artificial Intelligence for Genetic Cancer Risk Assessment in Gynecologic Oncology: A Review of the Current Landscape and Future Directions
Roxanna Haghighat, Sarah R Levi, Melissa K Frey

Abstract:

Hereditary cancer syndromes are associated with up to 25% of ovarian and 5% of endometrial cancers, yet rates of genetic testing and counseling remain low. Artificial intelligence (AI) offers new opportunities to streamline risk assessment, enhance gene variant interpretation, and expand access to genetic counseling. This narrative review synthesizes current evidence on AI applications in gynecologic cancer genetic risk assessment, including chatbot-based risk assessment, natural language processing of electronic records, and machine-learning approaches to variant classification. We highlight key challenges, including data bias, privacy, and implementation barriers, and outline future directions for AI technologies in gynecologic cancer genetic risk assessment.

Annotation

The finding: This review highlights that AI has significant potential to enhance hereditary cancer risk assessment in gynecologic oncology, particularly through chatbot-based screening, natural language processing of electronic medical records, and machine-learning support for variant interpretation. Existing studies indicate these tools can identify patients who meet NCCN criteria, simplify referrals, and potentially increase access where genetic counseling resources are limited. At the same time, the review stresses that current evidence is still early stage, and AI functions best as a supplement to clinician- and counselor-led care rather than a replacement.

Strengths and weaknesses: A major strength is that this article provides a broad, clinically useful overview of the current AI landscape across various stages of genetic risk assessment, from screening to counseling to variant interpretation. It also effectively discusses implementation challenges such as data bias, privacy, generalizability, and workforce limitations, making this review more balanced than solely technology-focused summaries. Its main weakness is that it is a narrative review rather than a systematic review of outcomes, so the conclusions are limited by the quality and heterogeneity of the included studies. Many of these underlying studies are feasibility studies, single-system studies, or conducted in selected populations, which restricts external validity and makes it difficult to determine whether AI improves long-term clinical outcomes.

Relevance: For C-L psychiatrists, this article is relevant because hereditary gynecologic cancer syndromes and related interventions often carry major psychological consequences, including anxiety, decisional conflict, anticipatory distress, reproductive concerns, and adjustment to prophylactic surgery or cancer diagnosis. As AI tools increasingly shape how patients are identified and counseled about genetic risk, psychiatrists working in oncology, women’s mental health, and medically complex settings will need to understand both the promise and the limits of these tools. The review is especially relevant to C-L psychiatry because it highlights systems-level questions—equity, access, communication, trust, and implementation—that directly affect patient experience and interdisciplinary care.