<p>One of the most prevalent cancers and a significant cause of cancer-related death globally is colorectal cancer. Even with improvements in treatment strategies for CRC, effective patient outcomes depend on early detection and accurate classification. Histopathological imaging (HI) is an efficient modality that performs tissue examination under a microscope to identify abnormalities in CRC images. The primary challenge presented by CRC images is the presence of similar texturing and patterns. Although CNNs are efficient for CRC classification, they frequently exhibit limitations in learning global context. To overcome this limitation, this work proposes a CNN named Dynamic Quantum and Self-Distillation-based Network (DQSD-Net) for classifying CRC images. It is mainly built using the Multi-Scale Fusion (MSF) module, which consists of the Dynamic Quantum State Fusion (DQSF) method along with parallel convolution branches. The DQSF method is a novel quantum-inspired feature fusion approach that performs feature integration through learnable phase amplitude interactions between two feature maps. Additionally, two self-distillation methods (SD), Representation-Level Distillation (RLD) and Prediction-Level Distillation (PLD), are also employed to improve the performance of the DQSD-Net. In the RLD method, hierarchical feature refinement is performed from deeper layers to shallower layers through compressed feature maps. Conversely, the PLD method performs the progressive knowledge transfer from the primary classifier to the auxiliary classifiers. The experimental results indicate that DQSD-Net surpassed the recent architectures by achieving accuracies of 97.42% and 99.52% on the colorectal histology and NCT-CRC-HE-100&#xa0;K datasets, respectively. The ablation studies also validate the efficacy of the proposed DQSF and SD methods.</p>