Alkali treatment is an effective method to separate and activate Euglenoid-derived paramylon; However, it may alter other properties such as molecular Size, Crystallinity Packing, Hydrogen-Bond Network, and Conformation of Beta-1,3-glucan backbones. Developed and validated a smart analysis System to track the Structural Evolution of alkali-treated Euglena polysaccharide based on empirical multi-modal characterisation Data. Paramylon-rich polysaccharides were treated with NaOH at 0.1, 0.3, 0.5, 0.7, and 0.9 mol L-1At 45 degrees Celsius for 2, 6, and 12 hours; the untreated groups were set as control. Produced 192 records from four biological extraction batches, each with three replicate analyses. Each entry included FT-IR spectrum, NMR-informed linkages, SEC-MALS molecular-weight values, XRD crystallinity coefficients, Congo-red conformer assessments, as well as particle morphology parameters. A modality-aware fusion model was trained to predict weight-average molecular weight, a composite structural-integrity index, trihelices retention, Beta-1,3 linkages retention and four different Structural States. Empirical data present a series of alkali-responsiveness stages. The structural stability was reduced by 42% and reached 83.8% after being added with 0.1mol/L, respectively.-1for 2 h, 0.626 at 0.5 mol L-1for 6 h, and 0.354 at 0.9 mol L-1For 12 hours. The corresponding MW values decreased as follows: 356 kDa → 331; 204 kDa; 82 kDa, triple-helix retention decreased from 100% to 91.2%; 67.2%; 39.4%. The proposed fusion method obtained an Mw RMSE of 8.9 kDa, an S_int RMSE of 0.024, a macro-F1 of 0.942, an AUROC of 0.985, and an expected calibration error of 0.031 compared with the FTIR-alone, NMR-alone, SEC-XRD alone or morphology-only early-fusion baselines. Ablated experiments confirmed that cross-modal attention, treatment embedding, structural-consistency loss, missing-modality masking, and calibration all contributed individually to the final results. The system maps the heterogeneous experimental results to a state-aware structure-evolution diagram and can perform batch screening, confirmatory measurement selection, and interpretability assessment of alkali-treatment degree. The figure sequence organisation structure is not empirical evidence but empirical supplements; Descriptors curves, three-dimensional structured landscapes, Latent state topologies, prediction performances, ablative behaviours, calibration methods, error distributions, case-level interpretations are presented separately. The moderate treatment zone has shown greater ambiguity; crystallinity and conformations decreased before molecular size collapse was complete. Calibrated fusion output, which includes both the state labels and the verification indications. Therefore, it can be connected with confirmation experiments of SEC-MALs, XRD, Congo-red and NMR techniques when the sample is close to the swelling-transition interface.
