The technological progress in the field of medical imaging in the last century has created unprecedented opportunities for non-invasive diagnosis, and established medical imaging as an integral part of medical and health system. One of the main areas of innovation representing these advances is the interdisciplinary field of medical image processing.
This rapidly developing field involves a wide range of processes from raw data acquisition to digital image transmission, which are the basis of complete data flow in modern medical imaging systems. Nowadays, these systems provide higher and higher resolution and faster acquisition time in terms of space and intensity dimensions, so as to produce a large number of high-quality original image data. These data must be processed and interpreted correctly in order to obtain accurate diagnosis results.
This paper focuses on the key areas of medical image processing, considers the environment of specific imaging modes, and discusses the main challenges and trends in this field.
The core field of medical image processing
There are many concepts and methods used to build the field of medical image processing, which focus on different aspects of its core area, as shown in Figure 1. These aspects form three main processes in this field – image formation, image computing and image management.
Figure 1. Structural classification of subject types in medical image processing.
The image forming process consists of data acquisition and image reconstruction steps, which are used to solve the problem of mathematical inversion. The purpose of image computing is to improve the interpretability of reconstructed images and extract clinically relevant information. Finally, image management processes the compression, archiving, retrieval and transmission of acquired images and derived information.
The first necessary step of image formation is to collect the original imaging data. The data contains raw information describing the captured physical quantities of each internal organ of the body. This information becomes the main subject of all subsequent image processing steps.
Different types of imaging modes can use different physical principles, which involves the detection of different physical quantities. For example, in digital radiography (DR) or computed tomography (CT), it is the energy of incident photons; In positron emission tomography (PET), it is the photon energy and its detection time; In magnetic resonance imaging (MRI), it is the parameter of RF signal emitted by excited atoms; In ultrasound, it is an echo parameter.
However, no matter what type of imaging mode, the data acquisition process can be subdivided into the detection of physical quantities, including the conversion of physical quantities into electrical signals, the preprocessing of collected signals, and the digitization of physical quantities. A general block diagram showing that all these steps are applicable to most medical imaging modes is shown in Figure 2.
Figure 2. General block diagram of data acquisition process.
Image reconstruction is a mathematical process of forming an image using the obtained original data. For multidimensional imaging, the process also includes a combination of multiple data sets captured at different angles or different time steps. This part of medical image processing solves the inversion problem, which is the basic topic in this field. There are two main algorithms for solving such problems – Analysis and iteration.
Typical examples of analytical methods include filtered back projection (FBP) widely used in tomography; Fourier transform (FT), which is particularly important in MRI; And delay superposition (DAS) beamforming, which is an indispensable technology in ultrasonic examination. These algorithms are sophisticated and efficient in terms of required processing power and computing time.
However, they are based on idealized models, so they have some obvious limitations, including their inability to deal with complex factors such as the statistical characteristics of measurement noise and the physics of imaging system.
The iterative algorithm overcomes these limitations and greatly improves the insensitivity to noise and the ability to reconstruct the optimal image from incomplete original data. The iterative method usually uses the system and statistical noise model to calculate the projection based on the initial target model. The difference between the calculated projection and the original data defines a new coefficient for updating the object model. This process is repeated using multiple iterative steps until the cost function of mapping estimated and true values is minimized, so as to integrate the reconstruction process into the final image.
There are many iterative methods, including maximum likelihood expectation maximization (MLEM), maximum a posteriori (map), algebraic reconstruction (ARC) and many other methods widely used in medical imaging mode.
Image calculation involves the calculation and mathematical method of reconstructed imaging data, which is used to extract clinical related information. These methods are used for enhancement, analysis and visualization of imaging results.
Image enhancement optimizes the transformed representation of an image to improve the interpretability of the information contained. The method can be subdivided into spatial domain and frequency domain.
Spatial domain technology works directly on image pixels, which is particularly useful for contrast optimization. These techniques usually rely on logarithm, histogram and power-law transformation. Frequency domain method adopts frequency transformation, which is most suitable for smoothing and sharpening images by applying different types of filters.
Using all these technologies can reduce noise and non-uniformity, optimize contrast, enhance edges, eliminate artifacts, and improve other related characteristics that are essential for subsequent image analysis and its accurate interpretation.
Image analysis is the core process of image computing. Its various methods can be divided into three categories: image segmentation, image registration and image quantization.
The image segmentation process divides the image into meaningful contours of different anatomical structures. Image registration ensures correct alignment of multiple images, which is particularly important for analyzing time changes or combining images obtained using different modes. The process of quantification determines the properties of the identified structure, such as volume, diameter, composition and other relevant anatomical or physiological information. All these processes directly affect the quality of imaging data and the accuracy of medical results.
The visualization process presents the image data as an intuitive representation of anatomical and physiological imaging information in a specific form on a defined dimension. Through direct interaction with the data, visualization can be carried out in the initial and intermediate stages of imaging analysis (for example, to assist in the segmentation and registration process), and the optimized results can be displayed in the final stage.
The last part of medical image processing involves the management of acquired information, including various technologies for image data storage, retrieval and transmission. Several standards and techniques have been developed to deal with all aspects of image management. For example, the medical imaging technology image archiving and transmission system (PACS) provides economical storage and access to images from multiple modes, while the medical digital imaging and communication (DICOM) standard is used to store and transmit medical images. The special technology of image compression and streaming realizes these tasks efficiently.
Challenges and trends
Medical imaging is a relatively conservative field. It usually takes more than ten years to transition from research to clinical application. However, its nature is complex, and it faces many challenges in all aspects of its scientific discipline, which steadily promotes the continuous development of new methods. These developments represent the main trends that can be determined in the core field of medical image processing.
The field of image acquisition benefits from the innovative hardware technology developed to improve the quality of raw data and enrich its information content. The integrated front-end solution enables faster scanning time, finer resolution and advanced architecture, such as ultrasound / mammography, CT / pet or PET / MRI combined system.
Fast and efficient iterative algorithm replaces the analysis method and is more and more used in image reconstruction. They can significantly improve the image quality of pet, reduce the X-ray dose in CT, and perform compression detection in MRI. The data-driven signal model is replacing the manually defined model to provide a better solution to the inversion problem based on limited or noisy data. The main research fields representing the trends and challenges of image reconstruction include system physical modeling and signal model development, optimization algorithms and image quality evaluation methods.
With the imaging hardware capturing more and more data, the algorithm becomes more and more complex, and people urgently need more efficient computing technology. This huge challenge can be solved by more powerful graphics processor and multiprocessing technology, providing a new opportunity for the transition from research to application.
The main trends and challenges associated with the transformation of image computing and image management cover many topics, some of which are shown in Figure 3.
Figure 3. Examples of major trend topics in today’s medical image computing.
The continuous development of new technologies related to all these topics has narrowed the gap between research and clinical application, promoted the integration of medical image processing field and physician workflow, and ensured more accurate and reliable imaging results.
ADI provides a variety of solutions to meet the most demanding medical imaging requirements for data acquisition electronic design, including dynamic range, resolution, accuracy, linearity and noise. Here are a few examples of such solutions developed to ensure the highest initial quality of raw imaging data.
A highly integrated analog front-end adas1256 with 256 channels is designed for Dr applications. The multi-channel data acquisition systems adas1135 and adas1134 with excellent linearity can maximize the image quality of CT applications. Multi channel ADCs ad9228, ad9637, ad9219 and ad9212 are optimized to have excellent dynamic performance and low power consumption, which can meet the requirements of pet. Pipelined ADC ad9656 provides MRI with excellent dynamic performance and low power consumption. The integrated receiver front-end ad9671 is designed for low-cost, low-power medical ultrasound applications requiring small-size packaging.
Medical image processing is a very complex interdisciplinary field, covering many scientific disciplines from mathematics, computer science to physics and medicine. This paper attempts to propose a simplified but well structured core domain framework, which represents the domain and its main themes, trends and challenges. Among them, the data acquisition process is one of the first and most important fields. It defines the initial quality level of the original data used in all subsequent stages of the medical image processing framework.