We propose a new method to measure the thickness of very thin single layers in helical computed tomography images. This method involves using the density profile of a longitudinal image to obtain the image width. First, we examined the images of the step density distribution of block phantoms to clarify the step spread function and to decide the spread width (L) of longitudinal images. It was found that the spread width is half of a sum of the X-ray beam collimation and the table feed distance when 180 degrees linear interpolation is used. Next, we derived the density profile of a single layer by taking the density distributions of a single layer to be the composite of step density distributions of different step heights. An important relation is made clear that the image width of the longitudinal image for a single layer of thickness (D) is equal to D+2L whether the shape is symmetrical or not. From this relation the thickness can be measured by detecting the image width with a known value of the spread width of longitudinal images. In conditions of beam collimation of 1.0 mm, table feed distance of 1.0 mm, and reconstruction interval of 0.5 mm, measurement accuracy of this method using the density profile of a longitudinal image is estimated to be within 0.05 mm. This method is much more effective than the conventional method using a gray level image for measuring thicknesses less than 0.6 mm. In addition, this method appears to be more consistent than the conventional method, as indicated by a lower standard deviation for figures obtained with this method than for those obtained with the conventional method. We applied this method to the measurement of the thickness of thin cortical bones.