Digital diagnostic images from subtraction angiography, computer tomography and ultrasound grabbed at 640 x 560 pixel and 320 x 280 pixel, eight-bit resolution were compressed at seven different levels with the 'lossy' JPEG algorithm in order to decrease the computer magnetic memory space required for their storage. After reconstruction, the deformations caused by the process were assessed (i) qualitatively: a perception test was performed twice, by five radiologists, on a total of 14 images compressed at various levels and (ii) quantitatively: the average grey level value per pixel was calculated at each compression level from each system's residual images obtained by subtracting the compressed image from the original and taking the absolute value of the result. It was concluded that images compressed at a ratio of about 10:1 were indistinguishable from the originals and hence could be considered acceptable for clinical use.