The authors present in this chapter an overview on evaluation of medical image compression. The different methodologies used in the literature are presented. Subjective evaluation uses some a priori knowledge such as the judgment of experts or the ability to realize a correct diagnosis. Objective evaluation generally takes into account the value of metrics: the PSNR is an example of such a criterion. The goal of hybrid evaluation is to realize a reliable judgment while having a simple computation. The authors discuss on the benefits and drawbacks of these approaches. The European Project called OTELO in which they were involved, gives feedback on ultrasound image compression.
Medical imaging is an important and a powerful technique whose goal is to facilitate the expert diagnosis. Many image processing algorithms can be used within this context such as: image filtering, compression, segmentation, interpretation or retrieval... One important issue concerns the evaluation of different image processing results for the medical expert: as for example, image filtering can improve the image quality but can disturb the ability of a medical expert to make a diagnosis.
The proposed chapter deals with the particular field of the evaluation of medical image compression. Image compression for medical applications is an important topic as many image acquisitions are transmitted and stored for a further analysis. In this context, we want to minimize the size of the compressed image while keeping a sufficient quality for the diagnosis. Several evaluation methods have been proposed in the state of the art. We propose in this chapter to make an overview of these approaches. We present different evaluation techniques for either an expert in image processing or for a medical expert. We discuss the advantages and drawbacks of each method. The European OTELO project (Delgorge et al., 2005) in which we have been involved, provides a good experience feedback in the evaluation of medical image compression.
The main objective of the European OTELO project (mObile Tele-Echography with an ultra Light rObot) was to develop a robotic tele-echographic system. A light weight robot holds and moves a real probe on a distant patient according to the expert gesture and permits an image acquisition using a standard ultrasound device (see Figure 1). Ultrasound images constitute the only feedback information available to the medical expert to remotely control the distant robotized system. The expert controls the remote probe holder robot by using a dedicated input device and based on the quality of the received information. The diagnosis made by the specialist strongly depends on the quality of these images. An important task also concerns the evaluation of the quality of the compressed images. Many experimental results are presented in this chapter in order to illustrate the behaviors of the different evaluation methods.
the OTELO tele-echographic system
We can distinguish three types of evaluation methods in the state of the art. The first one concerns the subjective evaluation. The quality of a compression result, for any medical types of images, is traditionally evaluated by considering a visual test where many experts examine a large set of images and score each one based on their quality or the ability to make a correct diagnosis. Second, we present the objective evaluation methods. Many statistical criteria have been proposed in the literature to automatically evaluate different compression results of a single image. We propose to compare many of them and discuss their efficiency. Subjective and objective evaluation methods have many advantages and drawbacks. These two approaches are complementary. That is why many works propose an hybrid approach. These methods are presented in the main trust of this article. The goal of these methods is to make a reliable judgment (similar to a medical expert) while using some statistical criteria to make the evaluation of a large set of compression results possible. The future trends in the domain are then proposed and a conclusion is given.Top
Image compression is an important issue in medical imaging as the distant visualization of medical images is now possible through high bandwidth networks for different applications (discussion between experts on a difficult case, storage of medical images of a patient...) and as telemedecine becomes an emergent technology nowadays (Delgorge et al., 2005).
Image compression is an image processing algorithm whose objective is to decrease the size of storage of the image while preserving as much as possible its visual quality (see Figure 2). Even if this definition is quite simple, the main problem is to evaluate the quality of a compression result. As for example, the result given in Figure 2 is clearly not very good but it is difficult to say if this quality would be satisfactory for a medical expert to make a diagnosis.
Example of a JPEG compression result (ratio 5%) of an ultrasound image
Key Terms in this Chapter
Subjective Evaluation: It is a quantitative or a qualitative evaluation involving experts or some a priori knowledge.
Human Visual System (HVS): It refers to the visual perception of humans that is simulated by researchers in evaluation.
Objective Evaluation: It is a quantitative evaluation generally based on statistical criteria. None a priori knowledge is used for the evaluation.
Receiver Operating Characteristic (ROC): ROC curves provides tools to analysis the behavior of models, algorithms or human judgments.
Peak Signal-to-Noise Ratio (PSNR): The PSNR is most commonly used as a measure of quality of reconstruction in image compression.
Fusion: Combination of different data in order to improve decision making.
Hybrid Evaluation: The goal of this approach is to obtain a judgment as reliable as the subjective one while having an easy computation.