If you would like to learn more about the IAEA’s work, sign up for our weekly updates containing our most important news, multimedia and more.
Information Technology
Nuclear Medicine
Introduction to Information Technology
According to Moore’s law, new released memory chips possess twice the capacity of prior chips, which means that memory size increases exponentially. The exponential growth of computer capabilities has a very important implication for the management of a Nuclear Medicine department as the increase in technical capabilities also needs to be transformed to a more productive working environment. In the field of Nuclear Medicine this would affect the various applications of computers in Nuclear Medicine:image processing, data acquisition, the storage of data as well as the use of databases. Digital image processing has developed simultaneously with the increasing capabilities of computers and contributes to the steady progress in the field of nuclear physics. In addition, modern Nuclear Medicine data acquisition devices have considerable computing power embedded in them that perform tasks to improve images. Also, very common devices like SPECT (Single Photon Emission Computed Tomography) and PET (Positron Emission Tomography) are affected by the progress of information technology. With the increase in computing power, topics like data analysis, data storage and security become more and more important. A hospital information system is a large, distributed database where data are collected from many different sources like clinical laboratories, , nuclear medicine departments and financial departments. With more acquired and collected information, databases need to be adapted and developed together with the increasing capabilities of computers.
The IAEA reference book Nuclear Medicine Physics provides a comprehensive overview of the topic of nuclear medicine in general. In addition, more specific topics such as image processing, data acquisition, data storage, databases and information security are also covered.
Imaging informatics and DICOM
Until the end of the last century, the vast majority of medical imaging examinations used film as a medium for image capture, display and storage. The digital image revolution in medical diagnostic imaging began in the 1970s with the invention of the computed tomography (CT) scanner and has continued to gain momentum. As a result, digital medical imaging has become the preferred method for medical imaging.
Important principles
Digital images are composed of individual picture elements (pixels), which typify a single point in the image and are usually represented by a series of numbers. Mathematically, image pixels can be seen as digital or discrete functions that give the value of the image at each point in the image. Each pixel can only have values at corresponding pixel locations. A computer can only represent digital functions, where both the independent and dependent variables are digital. The computer is able to construct an image suitable for display to a human observer out of this function.
In general, all modern imaging equipment will capture images in a format compatible with the DICOM (Digital Imaging and Communications in Medicine) standard, which was developed for medical device intercommunication, and the storage and transmission of medical images. The standard preserves fidelity and metadata on each image as for example specific patient identification, documentation of the used equipment, the examination type and date, image characteristics and the details of image acquisition. The DICOM standard can properly preserve cross-sectional imaging data from modern tomographic imaging systems such as CT, MRI, positron emission tomography (PET) and single photon emission computed tomography (SPECT). It then allows processing and reconstruction of the images using various software tools, which can enhance depiction and visualization well beyond that possible with the originally captured image data.
Specific and detailed information about digital imaging, information technology and its application for medical purposes is covered in the IAEA Human Health Series No.28 “Worldwide Implementation of Digital Imaging in Radiology”. In addition, the IAEA book on Nuclear Medicine Physics provides a complementary overview of the topic of nuclear medicine.
PACS-RIS system
Digital medical images are typically managed in a dedicated image network and database known as a Picture Archiving and Communications System (PACS), whereas all other patient information is typically recorded in a Radiology Information System (RIS). PACS is used for the storage, distribution and review of medical images and RIS is an information system where patients are registered, examinations are scheduled, and reports are recorded, stored and distributed.
Important principles
PACS and RIS need to work seamlessly togetherto enhance productivity in medical departments resulting in faster study turnaround times for patients and clinicians. The recent widespread deployment of PACS has revolutionized medical care through medical image communication as images and reports are immediately available for physicians within a hospital. Large countries with relatively scattered medical facilities can be serviced from major centres by picture archiving and communication systems in large networks. Centres with limited expertise in specific imaging modalities, or in less common diseases, are able to obtain expert opinions from subspecialists located in major centres. Such opinions are often obtained during a live conversation, either via voice over IP or telephone and can dramatically and rapidly alter the management of any patient.
There are still many technical problems inherent in PACS installations that require intense preparation and troubleshooting, particularly in the planning and initial implementation phases. While it is possible to organize and manage a digital imaging service using PACS alone, the absence of a RIS will mean significant limitations. Patient and room scheduling, workload balancing and the automatic scheduling of patient lists on imaging equipment are more difficult without a RIS. Furthermore, it is very difficult to manage image data for the same patient across more than one imaging centre as a RIS reduces the chance of duplicates and incorrect patient data entry.
Information about PACS and RIS as well as their application in hospitals is given in the IAEA Human Health Series No.28 “Worldwide Implementation of Digital Imaging in Radiology”. In addition, the IAEA book on Nuclear Medicine Physics provides a complementary overview to the topic of nuclear medicine.
Security in Information Technology
Digital imaging systems are also included in a radiation safety and information safety program. Radiation protection is inclusive of all aspects of medical imaging, including digital imaging systems. Digital systems have methods of automatically recording the exposures and delivered doses, which allows storing and reporting these doses for the calculation of effective dose and the monitoring of the dose delivery. Information safety on the other hand includes network safety, the studying of data quality, data reconciliation and data integrity.
Important principles
In non-digital medical imaging environments, the security of imaging data is innately associated with the physical security of the storage medium of the images. Physical security can be achieved through controlled access to sites where imaging systems and films are stored. Digital imaging systems require physical and operational security. Physical security includes protection of the computer equipment used to control the imaging systems and to manage, transmit, store, review and interpret medical images. This approach is similar to that used for non-digital images. Operational security involves limiting system access to authorized and recognized users. Usually, a superuser is tasked with administering the list of authorized users, their passwords and login identity codes. This must be maintained regularly for every individual user, and shared common passwords and login identity codes are not available. In addition, the system should be safe from unauthorized external access and hacking. The systems must be protected from viruses, spyware and other malicious software as it can reduce workstation and network performance, cause the workstation or RIS (Radiology Information System) and PACS (Picture Archiving and Communications System) servers to crash, corrupt the RIS or PACS database and covertly acquire and distribute sensible data externally. Malicious software can be introduced through external CD’s, flash drives, or web sites that have associated spyware or viruses. Coordination with the workstation vendors and the local IT department is required to ensure that appropriate antivirus software is installed, functional and regularly updated helps to protect a system from such software.
Information about security in information technology for hospitals is covered in the IAEA Human Health Series No.28 “Worldwide Implementation of Digital Imaging in Radiology”. In addition, the IAEA book on Nuclear Medicine Physics provides a complementary overview on the topic of Nuclear Medicine.
Tracer studies
In broad terms all diagnostic nuclear medicine methods are tracer studies. However, the term is generally used for those methods not utilizing imaging equipment but instead based on measurement of whole body retention or collection and measurement of biological samples such as plasma, urine etc.. Tracer studies are also an important tool in biomedical research, where tracer studies can be used to develop kinetic models, which in turn provide quantitative analysis.
Important principles
The principle of a tracer study is that it allows the analysis of a substance and its interactions in the body through the labeling of the substance with a radionuclide in a manner that does not alter the substances original properties. The main steps in a tracer study include selection and administration of the radiotracer, sampling, sample preparation and measurement. In diagnostic applications the tracer substance (radiopharmaceutical) is selected by its appropriateness for the study being applied, and the procedures for performing the study should be in accordance with the recommendations from professional organizations (international or national). Samples (e.g. blood, urine, …) are collected and prepared in accordance with the study objectives. The selection of measurement method and instrument should be based on the properties of the radioactive sample such as the radionuclide, expected activity, sample volume etc. Sample preparation may include careful dilution and measurement of volume and/or weight. The result of the measurement is generally expressed in terms of count-rate, which many times has to be corrected for background registrations, dead-time losses, decay of the radionuclide and the sensitivity of the instrument in order to calculate absolute activity or relative to a standard. A regular programme for quality control of the instruments should be established.
The most commonly used equipment for sample measurements include the gamma-counter and the liquid scintillation counter. Also external probes (e.g. for thyroid uptake measurement, surgical probes for detection of sentinel node), the gamma-camera (bile malabsorption) and the PET-camera (biomedical research in general) are used.
The web-book Basic Physics of Nuclear Medicine includes some basic information on compartment modeling. Details on the instrumentation used for measurements as well as the content of a quality control programme are presented in an IAEA document: Quality control of nuclear medicine instruments.
Machine Learning in Nuclear Medicine
Modern healthcare consistently gathers large amounts of electronic data, with a significant portion being imaging data. For instance, modern PET/CT systems generate gigabytes of data per patient study. As a result, new data handling and evaluation methods are needed that go beyond traditional software processing capabilities. Machine learning offers a promising solution for managing medical big data.
Important Principles
Machine learning involves techniques for handling vast amounts of data, which can identify and learn from patterns within datasets. Each machine learning algorithm has a logical structure composed of a model, a fitness measurement, and an optimizer. In this structure, the model predicts new information from the data and is the outcome of any machine learning training process, while the optimizer generates the model in a way that maximizes its fitness value. Machine learning can be divided into supervised and unsupervised methods. Supervised machine learning uses labeled reference data to build a predictive model, while unsupervised machine learning works with unlabeled data and can be described as a clustering approach, useful for identifying patterns and relationships within the data.
Machine learning methods can also be classified based on data extraction and analysis, with the main methods being shallow learning and deep learning. Shallow learning methods rely on engineered features, while deep learning methods build on automatically understood, multi-layer data representation. Deep learning generally outperforms shallow learning approaches, as it can analyze data at different complexity levels. In healthcare applications, deep learning is still underutilized, as its greatest potential lies in data with complex, hierarchical structures. However, deep learning shows promise in synthesizing artificial CT from MRI data or for attenuation correction of PET images (Papp L. et al., 2018).
The publication "Personalizing Medicine Through Hybrid Imaging and Medical Big Data Analysis" by Papp L. et al., published in the journal Frontiers in Physics, offers a concise introduction to the role and applications of machine learning in the field of nuclear medicine. Additionally, the IAEA reference book on Nuclear Medicine Physics provides a comprehensive overview of nuclear medicine in general.