The occurrence and development of malignant tumors have become a global public health event, which poses an important threat to human life and health. According to the cancer data released by the International Agency for Research on Cancer of the World Health Organization, there were 19.3 million new cancer cases in the world in 2020, including 9.96 million deaths caused by cancer, and 3 million cases in China, accounting for 30% of the total number, ranking first in the world^[1]. With the rapid development of molecular imaging and clinical diagnostic technology, how to accurately identify the process of tumor occurrence and development, tumor molecular typing, and so on, so as to achieve accurate diagnosis and treatment has become an urgent clinical and basic research problem^[2].

In order to reveal the molecular mechanisms related to tumor growth and explore novel biomarkers and regulators related to metastasis and drug resistance, multi-omics technology strategies have emerged, including genomic, epigenetic, transcriptomic, proteomic and metabolomic studies of homogenization systems^[3⇓~5]. However, these methods can not provide accurate images of the spatial information and cell distribution related to the tumor and its microenvironment. The occurrence and development of tumors undergo an extremely complex process, showing a high degree of spatial and temporal heterogeneity. Tumor cells with different locations in the same disease and different lesions and tumors may have great molecular biological differences. This heterogeneity directly affects the biological behavior, metastatic potential, and response to therapy of tumors^[6]. Therefore, accurate resolution of tumor spatiotemporal heterogeneity and visual analysis of molecular information are essential to achieve precise diagnosis and individualized treatment.

In order to make up for the shortcomings of traditional omics technology, spatial omics technology has been developed rapidly in recent years. Spatial omics technologies (including spatial transcriptome, spatial epigenetics, spatial proteome and spatial metabolome) aim to combine multiple molecular information and in situ spatial distribution information to reveal and clarify the complexity and diversity of physiological and pathological conditions of organisms at the spatial level^[7⇓⇓~10]. Spatial omics technology can be used to determine a variety of cell types and their spatial and temporal distribution, (molecular or metabolic) typing, and the interaction network of a variety of molecules, so as to obtain a more accurate panorama of tumor biology, which provides an important molecular basis for exploring the occurrence and development of malignant tumors, finding potential drug therapeutic targets, and developing new drugs^[7⇓⇓~10]. However, spatiomics methods generally focus on single cells, multicellular assemblies and tissue specimens in vitro, which lacks important in vivo information of organisms. Unlike spatial multi-omics techniques, most clinical imaging techniques have the advantage of non-invasiveness, which can accurately evaluate the biophysical characteristics of tumors, such as size, morphology, elasticity, location, metabolism, and blood perfusion^[11]. In recent years, the complementary advantages of spatiomics technology and clinical imaging technology have been used to integrate the two technologies, as well as the cross-disciplinary advantages of biology, physics, pathology, biophysics, imagomics, etc.It has been applied to the precise diagnosis of tumors, providing a new method and strategy for exploring the spatiotemporal heterogeneity of tumor morphological structure and panoramic molecular spectrum, and for lesion prediction.

The key problem to achieve precise diagnosis and treatment of tumors is how to comprehensively recognize the spatiotemporal heterogeneity of tumors and their microenvironment. From the perspective of time and space, tumors and their microenvironment are subject to precise molecular regulation and transcriptional regulation between cells, and have complex cell organization, stratification, evolution and structural characteristics, which directly affect the fate and development of cancer cells. When a variety of cell clusters in the body are stimulated by external stimuli, including environmental pollution, drug therapy, etc., a variety of biomolecules carry out reprogramming process, resulting in changes in metabolism, localization and immune microenvironment between tumor cells and stromal cells, which can not be detected and analyzed by traditional multi-omics technology^[12]. The introduction of spatiomics technology can help to obtain the spatial information corresponding to the whole gene transcriptome, proteome, metabolome and multiple biomolecules, that is, to carry out qualitative, quantitative and multi-dimensional imaging studies at the same time, so as to analyze the spatio-temporal characteristics of tumors and their microenvironments^[13,14].

Spatial omics technology has been widely used in precise analysis of tumors. In the study of cancer cell remodeling, heterogeneity and tumor microenvironment, taking lung adenocarcinoma as an example, the reason for its high mortality is the heterogeneity at the cellular, histological and molecular levels, and there are often multiple molecular subtypes within a single tumor. In order to deeply understand the cellular and molecular composition of highly heterogeneous specimens, Wang et al. Used spatial transcriptome technology and immunofluorescence analysis to explore the cellular composition and regulatory mechanism of tumor microenvironment driving the evolution of histocytic subtypes.Macrophages are the most abundant cell type in lung adenocarcinoma, and hypoxia-induced regulatory networks play an important role in determining subtype evolution. Different subtypes show significant intra-tumor heterogeneity in the dedifferentiated state, including having different tumor microenvironment subsets^[15]. These molecular heterogeneous landscapes of cancer progression provide important insights into potential therapeutic targets for strongly aggressive lung adenocarcinoma.

In the study of cancer development and metastasis, taking liver cancer as an example, its metastasis and recurrence is a huge problem in the overall diagnosis and treatment of liver cancer, and there is a lack of panoramic description of the gene expression profile and mutation profile of liver cancer, which leads to the lack of accurate clinical basis for the evaluation of liver cancer metastasis. Sun et al. Selected specimens of primary, metastatic and recurrent foci from patients with hepatocellular carcinoma to carry out spatial multi-omics study, and depicted a panoramic image of the spatio-temporal evolution of hepatocellular carcinoma metastasis.The molecular mechanism and evolution trajectory of clonal selection of liver cancer metastasis were found, which provided important data support for the development of markers and therapeutic targets suitable for clinical research to predict liver cancer metastasis^[16]. Similarly, in the clinical diagnosis of lung cancer, it is difficult to distinguish between multiple primary lung cancer and intrapulmonary metastasis, which puzzles its diagnosis and treatment. Wang et al. Used the spatial multi-omics strategy to deeply analyze the cell composition, spatial composition and characteristic molecular expression of the tumor microenvironment of different lesions, and screened out the genes highly expressed in multiple primary lung cancers, including CLDN2, and the specific cell type type II alveolar cells (CLDN2 significantly highly expressed)^[17].

In the study of immunotherapy for tumors, tumor heterogeneity significantly affects the effect of immunotherapy and its clinical prognosis. Song et al. Selected tissue samples from patients with advanced non-small cell lung cancer treated with bispecific antibodies to carry out spatial multi-omics studies to obtain spatial transcriptome and proteome information in tumor regions and stroma regions.It was found that there were significant differences in gene and protein expression between the two regions, and 18 regional characteristic proteins were screened, which could be used as spatial markers to predict PD1/PD-L1 and CTLA-4 bispecific antibody immunotherapy^[18]. Among them, spatial markers in the stromal region showed higher predictive power for immunotherapy.

With the development of technology, fusion strategies based on spatial multi-omics have emerged, including spatial epigenome-spatial transcriptome joint ^[19], spatial transcriptome-single cell sequencing joint ^[20~22], spatial transcriptome-spatial metabolome joint ^[23], spatial metabolome-spatial proteome joint ^[24] and spatial multi-omics data integration ^[25]. The combination of spatial transcriptome and single cell sequencing is one of the commonly used combination methods, which can accurately quantify the spatial heterogeneity of gene expression in tumor cells at the single cell level by using the combined gene phenotype and spatial distribution information at the single cell level, thus facilitating the acquisition of new cancer subtypes and the analysis of their spatial metastasis characteristics and trend ^[20~22]. In addition to the combination of simple methods, scientists have also developed a microarray-based spatial transcriptome method, which uses microarray Spots to reveal gene expression patterns in space, each of which can capture transcriptome information from multiple adjacent cells.Multimodal intersection analysis (MIA) was used for data mining analysis of microarray point single cell sequencing and spatial transcriptome data. Through this method, taking primary pancreatic cancer as an example, a new feature of the complex composition of cancer cells and their microenvironment is revealed, that is, the spatial distribution of various types of cell subsets is limited (including ductal cells, macrophages, etc.).However, it was significantly co-enriched with other cell types in space, and the co-localization of inflammatory fibroblasts and cancer cells expressing stress response gene modules was also found to be ^[26].

In addition, the researchers used adjacent tumor sections of gastric cancer tissues to carry out spatial metabolome and spatial transcriptome studies, and found and located metabolic and gene pathways (arginine and proline metabolism, fatty acid anabolism pathway, etc.) With differential changes in cancer cells, as well as metabolic reprogramming of immune cells.That is to say, glutamine and polyunsaturated fatty acid metabolism are significantly up-regulated in the tumor boundary area, which provides data for immune escape and precise therapeutic target screening in the development of gastric cancer^[23]. The researchers combined the spatial metabolic composition image and the spatial protein composition image based on trypsin enzymolysis to explore the molecular mechanism of breast cancer tumor proliferation and malignant lesions affected by environmental pollutants, and found the spatial heterogeneity of lipids and proteins driven by the morphological changes of breast tumor cells induced by bisphenol S exposure. Among them, low-dose bisphenol S (10 μg/kg body weight/day) altered the spatial distribution of ceramide-sphingomyelin signaling pathway, chromosome stability-related proteins and cell proliferation related to the central necrotic area of tumors, and produced significant tumor proliferation effects; High doses of bisphenol S (100 μg/kg body weight/day) altered the spatial distribution of proteins associated with the stability of nucleic acid structures in the peritumoral region and significantly accelerated the progression of the tumor (Fig. 1)^[24].

In general, spatiomics and various spatiomics combined technologies can characterize the cell types, immunity and interstitial microenvironment of tumors and their microenvironment in situ, thus constructing the interaction network of metabolism/protein/gene, which provides a frontier perspective and an important analysis tool for the further study of tumor progression.

The process of accurate clinical diagnosis and preoperative evaluation of tumors mainly depends on imaging examinations, including ultrasound diagnosis, magnetic resonance imaging (MRI), positron emission computed tomography (PET/CT), etc., that is, by confirming the location, size, extent of invasion, and whether there is (distal) metastasis of the lesion, to provide clinical information for the formulation of personalized treatment plans for patients^[27].

In addition, in view of the large amount of clinical image data generated, image feature extraction, image transformation, information mining and analysis can be carried out through automated, high-throughput and in-depth learning strategies.At the same time, combined with the individual information of patients, the prediction model of disease occurrence and development is constructed, which plays an important guiding role in the differential diagnosis, recurrence and prognosis evaluation of benign and malignant tumors^[28,29]. Taking the imagomics diagnosis results of lung cancer and head and neck cancer as examples, the researchers selected CT and CT/PET images, extracted characteristic data to quantify the intensity, texture and shape of tumor images, and found that these quantitative indicators had the ability to predict the therapeutic effect, and carried out prognostic ability evaluation and prediction model construction. At the same time, compared with genomics data, these prognostic imagomics features can capture tumor internal heterogeneity and correlate with gene expression patterns. This new prognostic phenotype screened by imagomics can be directly applied to clinical practice and plays an important role in diagnosis and treatment decision-making^[30]. There are only two treatment strategies for non-small cell lung cancer, including tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs). The generation of new treatment strategies depends on the search for biomarkers that can change dynamically during treatment. The researchers proposed a deep learning model based on ¹⁸F-FDG-PET/CT, which can quantitatively predict the EGFR mutation status with high accuracy and significantly distinguish the above two treatment strategies: for patients treated with EGFR-TKIs, the deep learning score (EGFR-DLS) is significantly positively correlated with longer progression-free survival; For patients treated with ICIs, EGFR-DLS was significantly negatively associated with durable clinical benefit and longer progression-free survival, that is, EGFR-DLS can identify patients with non-small cell lung cancer who are sensitive to both treatments and evaluate the efficacy^[31]. At present, clinical imaging technology is difficult to obtain accurate tumor staging, and can not provide more molecular information.

In 2021, The Innovation published a multimodal fusion strategy based on spatiomics, that is, to solve The problems of clinical medicine through the integration of spatiomics and clinical imaging, which is expected to become an important method for the future generation of biochemical analysis and precise diagnosis and treatment of tumors (Fig. 2)^[32]. Most biological and chemical imaging techniques rely on "intrinsic properties" to visualize the target compounds and their interactions in the sample, such as element chemical bond information, mass-to-charge ratio, molecular fragments, ion mobility and functional group information, and most of them are in vitro detection without in vivo/in vivo information. Clinical imaging techniques mostly use non-invasive methods to carry out detection and analysis, which provide a large number of biophysical characteristics of organisms in vivo, and quickly obtain diagnostic results to guide clinical research, but can not obtain more comprehensive and diverse molecular information. The integration of spatiomics and clinical imaging technology can obtain more "hidden" physical, chemical and biological spatio-temporal information from complex biological processes.It provides a more valuable perspective for the study of biological processes with higher spatial resolution, thus helping to clarify the occurrence, diagnosis, personalized treatment and prognosis of diseases, which is a real close integration of clinical and basic research^[32]. Specifically, the fusion research of spatiomics and clinical imaging technology is mainly applicable to the following contents: cell heterogeneity research: the fusion technology can be used to explore cell types and their States (interaction, nutrient molecular gradient, oxygen supply conditions, etc.) And relative spatial location.So as to reveal the interaction and regulation between molecule-molecule, cell-cell and cell-stroma, which is conducive to in-depth understanding of the functional changes of tumors and their metastatic tissues and organs; Discovery of new mechanisms of tumorigenesis, development and metastasis: Based on spatiomics-clinical imagingomics data, combined with imaging mass spectrometry, pathology and other research strategies, new biological hypotheses can be further obtained, including new signaling pathways and molecular networks, cell subsets, etc^[32,33].

With the development of imaging technology and data integration methods, the fusion technology of spatiomics and clinical imagomics has shown great potential in biochemistry, clinical medicine and other fields.It includes the metabolism and spatial distribution of drugs, the identification of molecular phenotypes related to tumor heterogeneity, the screening of target organs exposed to environmental pollutants, and the screening of biomarkers related to disease progression. At present, the emerging fusion technologies of spatiomics and clinical imagomics include the fusion of spatiomics and mature clinical imaging technologies, such as MRI, PET, pathological imaging, clinical ultrasound, and the integration of multi-modal data^{[32⇓⇓~35]}. Take the clinical ultrasound-guided spatio-omics study as an example: the spatio-temporal heterogeneity of tumors provides an escape mechanism for breast cancer cells, thus hindering the study of tumor progression. Ultrasound diagnostic technique has high accuracy but low chemical specificity. By combining with spatiomics technology, which can provide rich molecular spatial information, scientists have proposed a non-invasive ultrasound elastography guided molecular spatial imaging strategy (UEg-MSI).This strategy integrates tumor physical and biochemical characteristics acquired from in vivo and in vitro imaging, shows elasto-histopathological-metabolic fingerprints and related molecular interaction information, and reveals the complete multi-focal spatiotemporal heterogeneity of breast spontaneous tumors in the early, middle and late stages. Primary lesions of the breast located in the neck and chest were found to be progressively more malignant according to increasing elasticity scores, from hyperplasia to nests of tumor cells, increasing areas of necrosis, and changes in regional metabolic heterogeneity and spatiotemporal reprogramming. At the same time, the study confirmed the distinct characteristics of tumor progression in breast tumors, that is, the synthesis of fatty acids from top to bottom is mainly concentrated in the core cancer nest area in the middle and late stages of tumors and in the peripheral microdomain in the early stage of tumors. The UEg-MSI strategy expands the application of spatiomics technology, while also improving ultrasound-mediated tumor diagnostic capabilities, providing new insights into the early detection of cancer and the occurrence of metastasis (Fig. 3)^[35]. To sum up, the fusion technology of spatiomics and clinical imagomics integrates the advantages of high sensitivity, high throughput, high specificity and high spatial resolution, and has high complementarity and flexibility in two-dimensional and three-dimensional multi-modality imaging.

The integration of spatiomics and clinical imaging technology is actually the integration of basic imagomics and clinical medical technology, and its greatest advantage is that it can visualize the distribution of substance molecules on the spatio-temporal scale.It is urgent to apply it to the research of scientific problems related to frontier clinical medicine or basic medicine, so as to truly promote and promote the scientific research of tumor-related problems in the field of clinical medicine. It should be pointed out that the fusion technology is still in the initial stage of development. Although the integration of multiple technologies is necessary, not every scientific problem needs fusion technology, and specific research strategies need to be adopted for specific scientific problems. In addition, the integration of technologies is not a simple list of methods, but a necessity to integrate and obtain key complementary information. Spatiomics and clinical imaging technology, data integration and fusion need continuous improvement and development. The difficulties and key points of research are mainly reflected in the following aspects: (1) developing innovative biomedical engineering technology to more realistically simulate the whole process of tumor evolution.Including the invasion, infiltration and metastasis between tumor cells, stromal cells and tumor-stromal cells, focusing on the development of methods and strategies related to the preparation, assembly, manipulation and biological visualization of (tumor) organoids; (2) Develop new methods such as data integration, image registration, preprocessing and noise reduction of multi-image omics to enhance the mining and interpretation capabilities of hidden spatio-temporal evolution data; (3) Construct a tumor evolution database based on spatial omics and clinical imaging technology, realize the resource sharing of multi-regional, multi-modal data and multi-spatio-temporal visualization information of different tumors, promote the discovery of new biomarkers related to tumor diagnosis and the screening and evaluation of innovative drugs, and provide new ideas and methods for early diagnosis and treatment of diseases.