This study aimed to explore the evaluation of Adriamycin-loaded microspheres in the treatment of liver cancer under DenseNet-based magnetic resonance imaging (MRI) image classification algorithm. According to different treatment methods, the research objects were classified into a normal saline (saline) group, a doxorubicin raw material (DOX) group, and a chitosan cross-linked pectin-doxorubicin conjugate macromolecular (CS-PDC-M) group. DenseNet’s migration learning was employed to analyze the dynamic enhanced MRI characteristics and classify the MRI images. The CS-PDC-M-targeted nanotransfer system was examined with its apparent morphology, drug absorption, and cytotoxicity. Tumor volume was monitored using MRI, and alanine aminotransferase (ALT) and creatine kinase isoenzyme (CK-MB) values were detected. Results showed that the classification accuracy of liver cancer MRI image based on DenseNet model reached 80% at the arterial hepatobiliary stage. The DOX and CS-PDC-M group had obviously smaller tumor volume than that of the saline group with a statistical meaning. The mortality in the DOX group was 30%, while there was no death in the saline and CS-PDC-M groups. Compared with the saline and CS-PDC-M groups, ALT and CK-MB from the DOX group increased substantially . Therefore, DOX had an inhibitory effect on tumor but damaged the heart and liver. DOX was used to construct CS-PDC-M that could maintain the original treatment effect of DOX and inhibit its side effects on the body, so CS-PDC-M had a clinical application value. In conclusion, Adriamycin-loaded microspheres could not only maintain the original therapeutic effect of Adriamycin but also inhibit its toxic and side effects on the body. The DenseNet model was applied in the liver cancer MRI dynamic image classification algorithm, and the normalization algorithm could improve the accuracy of the liver cancer microvessel classification, thus promoting the diagnostic efficiency of liver cancer diagnosis, which had clinical application value.

中文翻译：

---

基于 DenseNet 的 MRI 图像分类，用于检测治疗肝癌前后的差异

本研究旨在探讨在基于 DenseNet 的磁共振成像 (MRI) 图像分类算法下阿霉素微球治疗肝癌的评价。根据治疗方法不同，将研究对象分为生理盐水（生理盐水）组、阿霉素原料（DOX）组和壳聚糖交联果胶-阿霉素共轭大分子（CS-PDC-M）组。DenseNet 的迁移学习被用来分析动态增强的 MRI 特征并对 MRI 图像进行分类。研究了 CS-PDC-M 靶向纳米转移系统的表观形态、药物吸收和细胞毒性。使用 MRI 监测肿瘤体积，并检测丙氨酸转氨酶 (ALT) 和肌酸激酶同工酶 (CK-MB) 值。结果表明，基于DenseNet模型的肝癌MRI图像在动脉肝胆分期分类准确率达到80%。DOX和CS-PDC-M组肿瘤体积明显小于生理盐水组具有统计意义。DOX 组的死亡率为 30%，而盐水和 CS-PDC-M 组没有死亡。与生理盐水和 CS-PDC-M 组相比，DOX 组的 ALT 和 CK-MB 显着增加. 因此，DOX对肿瘤有抑制作用，但对心脏和肝脏有损害。用DOX构建CS-PDC-M，既能保持DOX原有的治疗效果，又能抑制其对机体的副作用，因此CS-PDC-M具有临床应用价值。综上所述，载有阿霉素的微球既能保持阿霉素原有的治疗作用，又能抑制其对机体的毒副作用。将DenseNet模型应用于肝癌MRI动态图像分类算法中，归一化算法可以提高肝癌微血管分类的准确率，从而提升肝癌诊断的诊断效率，具有临床应用价值。