The urinary sediment examination is carried out by microscopic inspection of laboratory technicians. Particles contained in urine samples are classified and counted, then the results are reported to doctors. These results can help diagnose diseases affecting the kidney, ureter, and bladder. Image processing techniques are expected to automate the microscopic inspection. However, the application to practical use has been prevented by the fact that the particles in urinary sediment have various shapes, sizes, and colors (even those of the same kind). In this paper, in order to automate the examination, a new image segmentation method and a hierarchical modular neural network (HMNN) are proposed for the classification of the images captured by flow cytometry. Proposed segmentation method can segment not only colored particles but also transparent ones by using both of gray level and its differential values. HMNN enables accurate classification of urinary sediment images by using small neural network modules hierarchically. Experimental results using our segmentation method showed its effectiveness. We compared the classification accuracy when using the HMNN to that with a general single neural network (SNN) and found that the classification accuracy is higher than when using a SNN. Finally, we applied the proposed method to automatically measure the density of the particles of each kind. Its measurement results were compared to those of technicians’. The match rates of the automated measurement results and technicians’ results were more than 93% and we concluded that the proposed methods enable automation of screening-purpose examination by jointly using chemical tests.