Our research group has previously developed a novel label-free flow cytometry method based on diffractive imaging, which utilizes a time delay integration (TDI) camera to capture diffraction images and employs machine learning algorithms for cell identification. However, the detection throughput is limited by the scanning frequency of the TDI camera. To address this limitation, we designed an TDI camera optimization scheme to increase the scanning frequency and verify its practical effectiveness. In this study, we optimized the timing control of the TDI camera, successfully increasing its scanning frequency from 50 kHz to 100 kHz. In the validation experiments, after capturing diffraction images with the optimized camera, we extracted feature values using the Gray-level co-occurrence matrix (GLCM) and conducted machine learning training with support vector machine (SVM) and random forest (RF) classifiers. The classifiers were used to distinguish between cultured normal liver cells and hepatocarcinoma HepG2 cells, and to classify three lung cancer cell lines (A549, NCI-H378, and NCI-H446) in a three-class identification task, achieving test set recognition accuracies of 94.14% and 95.20%, respectively. Our optimized system not only doubled the cell flow rate but also ensured the acquisition of images that meet the recognition requirements. This innovation provides a novel technical support for high-speed imaging, with significant scientific and practical value.