Cong Liu

A feasibility study of automating radiotherapy planning with large language model agents

Abstract Objective. Radiotherapy planning requires significant expertise to balance tumor control and organ-at-risk (OAR) sparing. Automated planning can improve both efficiency and quality. This study introduces GPT-Plan, a novel multi-agent system powered by the GPT-4 family of large language models (LLMs), for automating the iterative radiotherapy plan optimization. Approach. GPT-Plan uses LLM-driven agents, mimicking the collaborative clinical workflow of a dosimetrist and physicist, to iteratively generate and evaluate text-based radiotherapy plans based on predefined criteria. Supporting tools assist the agents by leveraging historical plans, mitigating LLM hallucinations, and balancing exploration and exploitation. Performance was evaluated on 12 lung (IMRT) and 5 cervical (VMAT) cancer cases, benchmarked against the ECHO auto-planning method and manual plans. The impact of historical plan retrieval on efficiency was also assessed. Results. For IMRT lung cancer cases, GPT-Plan generated high-quality plans, demonstrating superior target coverage and homogeneity compared to ECHO while maintaining comparable or better OAR sparing. For VMAT cervical cancer cases, plan quality was comparable to a senior physicist and consistently superior to a junior physicist, particularly for OAR sparing. Retrieving historical plans significantly reduced the number of required optimization iterations for lung cases (p < 0.01) and yielded iteration counts comparable to those of the senior physicist for cervical cases (p = 0.313). Occasional LLM hallucinations have been mitigated by self-reflection mechanisms. One limitation was the inaccuracy of vision-based LLMs in interpreting dose images. Significance. This pioneering study demonstrates the feasibility of automating radiotherapy planning using LLM-powered agents for complex treatment decision-making tasks. While challenges remain in addressing LLM limitations, ongoing advancements hold potential for further refining and expanding GPT-Plan’s capabilities.

Cell type-specific genotoxicity in estrogen-exposed ovarian and fallopian epithelium

Abstract Background Loss of the genomic stability jeopardize genome stability and promote malignancies. A fraction of ovarian cancer (OvCa) arises from pathological mutations of DNA repair genes that result in highly mutagenic genomes. However, it remains elusive why the ovarian epithelial cells are particularly susceptible to the malfunction of genome surveillance system. Methods To explore the genotoxic responses in the unique context of microenvironment for ovarian epithelium that is periodically exposed to high-level steroid hormones, we examined estrogen-induced DNA damage by immunofluorescence in OvCa cell lines, animal and human samples. Results We found that OvCa cells are burdened with high levels of endogenous DNA damage that is not correlated with genomic replication. The elevation of damage burden is attributable to the excessive concentration of bioactive estrogen instead of its chemomimetic derivative (tamoxifen). Induction of DNA lesions by estrogen is dependent on the expression of hormone receptors, and occurs in G1 and non-G1 phases of cell cycle. Moreover, depletion of homologous recombination (HR) genes (BRCA1 and BRCA2) exacerbated the genotoxicity of estrogen, highlighting the role of HR to counteract hormone-induced genome instability. Finally, the estrogen-induced DNA damage was reproduced in the epithelial compartments of both ovarian and fallopian tubes. Conclusions Taken together, our study disclose that estrogen-induced genotoxicity and HR deficiency perturb the genome stability of ovarian and fallopian epithelial cells, representing microenvironmental and genetic risk factors, respectively.