The Value of Compressed Medical Images in Diagnostics: JPEG 2000 Lossy Compression of CT and MRI Images

Abstract

Background: Hospitals and clinical environments are rapidly moving toward the use of digital medical images. Digital images are acquired mainly for diagnostic reasons, important for the documentation of the patient and his/her disease, and offer an alternative to analog film-based systems. However, medical images occupy a substantial amount of memory space. Image compression is important in reducing image transmission time and storage space while maintaining relevant diagnostic information. Without image compression, images data would overload the network and make telemedicine impractical. The issue still remains of how much medical images can be compressed and still retain their diagnostic quality. Aim: The goal of this study was to determine the level to which medical images can be compressed in order for reconstructed images to be acceptable for diagnostic purposes. Methods: We showed compression and reconstruction of The Digital Imaging and Communication in Medicine (DICOM) images obtained from two different modalities: computed tomography (CT) and magnetic resonance imaging (MRI). Image compression and reconstruction was performed using JPEG 2000 compression standard which uses the biorthogonal CDF 9/7 wavelet for lossy compression. Results: Based on the results obtained from objective, Peak Signal Noise Ratio (PSNR) and Mean Square Error (MSE), and subjective measures, we showed that a MRI image can be compressed to 45:1 (0,1768 bpp) without losing its diagnostic value, while a CT image can be compressed to 32:1 (0,25 bpp). The correlation between each objective and subjective measure was found. Conclusion: JPEG 2000 compression standard which uses the biorthogonal CDF 9/7 wavelet for lossy compression can be used to achieve efficient compression of CT and MRI images without compromising their diagnostic quality.