PROST is a framework for quantitative identification of spatially variable genes (SVGs) and domain detection in spatial transcriptomics. It consists of two modules: the PROST Index (PI) and the PROST Neural Network (PNN). The PI module quantitatively characterizes spatial variations in gene expression patterns without statistical hypotheses, while the PNN module uses a self-attention mechanism to integrate spatial and transcriptional information for unsupervised clustering of spatial domains. PROST outperforms existing methods in SVG identification and domain segmentation across various spatial resolutions, including multicellular and cellular levels. The PI score enables prioritization of spatial expression variations, facilitating biological insights. PROST is flexible and robust, capable of analyzing diverse spatial transcriptomic data. It effectively identifies spatial domains and SVGs with clear spatial expression patterns and biological interpretations. PROST also demonstrates strong performance in identifying spatial patterns in mouse embryogenesis at single-cell resolution. The framework is scalable and efficient, with potential for future analysis of larger datasets. PROST's PI module provides a robust index for quantifying spatial gene expression patterns, considering tissue context and expression profiles. The PNN module enhances spatial dependency learning, enabling accurate tissue segmentation. PROST's flexibility allows it to be applied independently or integrated into existing workflows. Future improvements could involve integrating histological images and spatial gene expression profiles for tissue domain determination. PROST has the potential to enhance spatial domain segmentation, particularly in complex tissues, by leveraging reference atlases and transfer learning techniques.PROST is a framework for quantitative identification of spatially variable genes (SVGs) and domain detection in spatial transcriptomics. It consists of two modules: the PROST Index (PI) and the PROST Neural Network (PNN). The PI module quantitatively characterizes spatial variations in gene expression patterns without statistical hypotheses, while the PNN module uses a self-attention mechanism to integrate spatial and transcriptional information for unsupervised clustering of spatial domains. PROST outperforms existing methods in SVG identification and domain segmentation across various spatial resolutions, including multicellular and cellular levels. The PI score enables prioritization of spatial expression variations, facilitating biological insights. PROST is flexible and robust, capable of analyzing diverse spatial transcriptomic data. It effectively identifies spatial domains and SVGs with clear spatial expression patterns and biological interpretations. PROST also demonstrates strong performance in identifying spatial patterns in mouse embryogenesis at single-cell resolution. The framework is scalable and efficient, with potential for future analysis of larger datasets. PROST's PI module provides a robust index for quantifying spatial gene expression patterns, considering tissue context and expression profiles. The PNN module enhances spatial dependency learning, enabling accurate tissue segmentation. PROST's flexibility allows it to be applied independently or integrated into existing workflows. Future improvements could involve integrating histological images and spatial gene expression profiles for tissue domain determination. PROST has the potential to enhance spatial domain segmentation, particularly in complex tissues, by leveraging reference atlases and transfer learning techniques.