This study aimed to analyze the diagnostic value of magnetic resonance imaging (MRI) combined with computed tomography (CT) based on the improved algorithm for acquired immune deficiency syndrome (AIDS) combined with spinal tuberculosis, so as to provide an effective reference theory for the clinical application of imaging diagnosis. The ResNet and Inception network structure were combined to form an improved convolutional neural grid classification (CNGC) algorithm. The improved algorithm was applied to AIDS patients with spinal tuberculosis to evaluate its diagnostic effect and value. 50 patients with AIDS and spinal tuberculosis and 50 patients with spinal tumors were selected for MRI and CT scans. The results showed that the accuracy, specificity, and sensitivity of the improved ResNet-Inception algorithm were much higher than those of the ResNet18 and GoogLeNet algorithms. In addition, compared with the training losses of ResNet18 and GoogLeNet algorithms that converged after 700 times, the ResNet-Inception algorithm only required 350 times to achieve convergence. Based on the improved ResNet-Inception algorithm, the false positive rate (FPR) of AIDS combined with spinal tuberculosis in imaging examinations was about 4%, which was greatly lower than that of the other two traditional algorithms (
P
<
0.05
). It can be seen that the improved ResNet-Inception algorithm was more valuable than traditional algorithms for MRI combined with CT in the diagnosis of AIDS combined with spinal tuberculosis and showed great clinical value for the diagnosis of the disease and the differential diagnosis of spinal tumors.