Compression of video-otoscope images for tele-otology: a pilot study

This pilot study was designed to determine the maximum level of compression of digitised images of the eardrum and to develop assessment protocols. The JPEG algorithm was used to compress fifteen 1.44 MB images to different sizes. As an objective assessment, the RMS errors between the original and compressed images were calculated. Two assessors graded image quality and recorded their observations of clinically significant abnormalities, which were compared to a gold-standard. It was found that RMS error increased markedly when images were compressed beyond about 20 KB, and at least 90% of the images quality were graded as being of good quality until image size went below 15KB. Agreement with the gold-standard in the identification of abnormalities was >70% for the two assessors, but there was a wide range of sensitivity and specificity values. In both cases there was no relationship with the level of image compression. It was concluded that images of the eardrum can be compressed using JPEG to about 20 KB before image quality is affected, but further studies require a higher quality of original images and a well understood protocol for assessment before further conclusions can be made on the effect of image compression on the ability to detect clinically significant abnormalities.