Tumor Data Analysis Based On Multi-Dimensional And Multi-Scale Imaging

Human global cancer incidence and mortality rates are increasing due to the influence of genetic,environmental,and dietary factors.The complexity of cancer hinders accurate diagnosis,determination of tumor biological boundaries,and further leads to poor prognosis.With the rapid development of synchrotron radiation and X-ray detectors,synchrotron based X-ray microtomography（SR-XPCT）has been used for pathological micro-morphological studies of tumors and other major diseases.SR-XPCT has successfully visualized tumor growth and development in three dimensions at the micron level and has become an important tool for tumor pathological examination and evaluation.In addition,Fourier transform infrared spectroscopy（FTIR）has been widely used for the analysis of tumor-related mechanisms and the discovery of biomacromolecular markers,such as lung,brain,thyroid,breast,prostate,and endometrial cancers.In this study,we developed an effective method for multidimensional and cross-scale correlation analysis based on tumor tissue and blood samples.By combining including SR-XPCT,SR-FTIR with clinical H&E and pathological examination,the tumor tissues were applied for demonstration and presentation.Among them,SR-XPCT utilizes microscopic morphological features of tumor tissues for quantitative analysis,and the diameter of calcified spots in most breast cancer tissues is in the range of 10-20μm,with a microcalcification density of 12.78%.The volume ratio of tumor microvascular network in lung cancer tissues was 14.83%.The diagnosis of the degree of tumor deterioration can provide clinicians with more morphological information and effectively avoid misdiagnosis of fine pathological diagnosis of tumor characteristic structures by mutual confirmation of micromorphology on in situ digital sections and stained tissue sections of tumor tissues.Infrared spectroscopy measurements of tumor tissues using SR-FTIR revealed that the increase in absorbance with vibrational functional groups was consistent with an increase in the concentration of certain specific biocomponents,which could help to further identify potential biomarkers of tumors.Additional blood samples collected from different cancer patients for attenuated total reflection-Fourier transform infrared spectroscopy（ATR-FTIR）measurements revealed that most of the characteristic peaks appeared in the 1000-1200 cm^-1 and 1500-1700 cm^-1 regions,mainly from stretching and bending vibrations of carbohydrates and ribose.Finally,various multivariate statistical analysis methods and machine learning algorithms were used to further compare the various differences between the IR spectra of cancer patients and healthy groups,and to develop a model for analyzing and evaluating the classification of different cancer species using IR spectroscopy.For our in-situ serum infrared spectroscopy data,a 2D-SD-IR-based feature dataset was constructed by setting the reference point to 1000 cm^-1and quantitative extraction of absorbance and wave number shifts of vibrational peaks for comprehensive cancer diagnosis,and the proposed machine learning method for cancer diagnosis using 2D-SD-IR-based feature dataset achieved 100%classification accuracy and revealed The relationship between different blood test data was revealed.Based on the diagnostic data from our experimental scheme,combined with histopathology and blood biochemistry tests,a more comprehensive and accurate diagnosis and understanding of soft biomedical specimens can be achieved.In conclusion,the proposed cross-scale and multimodal data acquisition and analysis method has great potential application in the development of tumor evaluation criteria.