Ocular tumors represent a class of diseases characterized by complex diagnosis and treatment decisions that heavily rely on physician expertise. Treatment selection depends not only on tumor size and location but is also closely tied to metastasis risk determined by genetic profiling. In clinical practice, integrating vast amounts of clinical, imaging, pathological, and genomic data to make optimal decisions poses a significant challenge for any physician. This leads to substantial variations in treatment decisions across different medical centers and among different doctors. Artificial intelligence, particularly deep learning technologies, has demonstrated potential to surpass human experts in medical image analysis, natural language processing, and predictive modeling. It can uncover deep patterns within high-dimensional, complex multimodal data that are difficult for the human eye to discern. Therefore, applying AI to ocular tumor diagnosis and treatment to build an intelligent decision support system holds significant clinical value and practical relevance. This paper aims to systematically review the research framework and development pathways for AI-driven personalized treatment decision support systems for ocular tumors. It designs and preliminarily validates such a system. By integrating patients' multidimensional data, this system provides clinicians with dynamic, quantitative, and visualizable decision references. Its core objective is to realize precision medicine through personalized treatment plans, thereby enhancing the overall standard of ocular tumor diagnosis and treatment. A combined methodology of bibliometric analysis and systematic review is employed. Relevant literature published in core databases over the past five years is retrieved and organized, focusing on keywords such as “AI,” “ocular tumors,” and “decision support” to analyze research trends and technological hotspots. High-quality studies were subjected to in-depth content analysis, comparing and summarizing dimensions such as research design, data sources, algorithmic models, validation methods, and system integration. This enabled the identification of commonalities, strengths, and limitations within existing research paradigms, facilitating logical extrapolation and forward-looking projections.Current findings indicate that deep learning-based prognostic prediction models demonstrate significant advantages. Multimodal data fusion has been proven to effectively enhance model performance. Regarding treatment recommendations, preliminary simulation studies show high consistency (>90%) between AI proposals and expert decisions. However, these findings also reveal critical challenges: most models rely on single-center retrospective validation, raising questions about their generalizability; inconsistent data standards and barriers to data sharing severely constrain progress; and insufficient model interpretability coupled with a lack of prospective clinical utility validation remain core obstacles to clinical implementation. The development of AI-driven DSS for ocular tumors represents an inevitable trend toward precision medicine with immense potential. Future research must advance simultaneously on two fronts: technological breakthroughs and clinical implementation. Clinically, the immediate priority is conducting large-scale, multicenter prospective randomized controlled trials to confirm the system's clinical value in improving patient outcomes and its cost-effectiveness. Ultimately, seamless integration and widespread adoption within real-world diagnostic and treatment workflows will be achieved through deep multidisciplinary collaboration.