In radiation oncology, probably more than in other medical specialties, the design and implementation of electronic information systems occur due to the need to reduce and eliminate human error in treatment – such as during the transfer of information from treatment planning to treatment delivery [6]. Although there is certainly a trend toward paperless or paper-light departmental organization, this is still work in progress in most units: digitalization touches several issues from legal, practical to know-how. While it has been demonstrated that a purely digital workflow is possible [7], this is not the case in many centers. The level of digital information available varies, heavily dependent on department-specific configurations. Costs and logistical and technical issues have been perceived as barriers in the implementation of a purely digital workflow [8]. Simply reproducing previous paper-based workflows can lead to unsafe processes and inefficient information flow [9]. On the other hand, key data in radiation oncology (e.g., treatment plans, 3D imaging) is not replaceable with paper so a transfer from a hybrid to a paperless system seems intuitive.

Similar to the level of digitalization, integration can be implemented to various extents. The purpose has been clear for decades now, to capture and store data consistently in order to rearrange and display the data as needed [10]. From a machine learning point of view, a strong fragmentation with redundancies may be present within an RO department or hospital. Heterogeneity of computer systems is ubiquitous and has multiple healthcare domain-specific causes [11]: departmental organization within a hospital, medical devices with inflexible built-in software, either general or niche vendors, selecting the “best” software for each specific purpose ignoring the big picture, as well as legacy applications [12, 13]. Often, third-party providers supply picture archiving and communication systems (PACS) or systems specific for laboratories or, for example, pathology reports.

These issues have been recognized a long time ago and have not been resolved to date. Radiation oncology units are typically embedded in, or at least associated with a broader healthcare system. Often, some information is associated and stored in a hospital information system (HIS) before the patient reaches a radiation oncology unit. This would include general demographics and patient identifiers (date of birth, gender, etc.), whose uniqueness is decisive for assuring the consistency between databases.

Integration is not only an issue between a radiation oncology unit and the hospital. It is nontrivial among different suppliers within a radiation oncology unit. IHE-RO stands for Integrating the Healthcare Enterprise-Radiation Oncology (IHE-RO) and is an ASTRO-sponsored initiative for improving the functionality of the radiation oncology clinic. The IHE-RO task force develops IHE integration profiles, which specify how industry standards are to be used to address specific clinical problems and ambiguities [14]. This process has succeeded in defining standards that have helped advance radiation oncology-specific integration [15].

Problems with integration are complicated by patient data confidentiality and security [16, 17]. A universal database where all relevant information is well structured and available is hardly to be expected. There are however attempts to centralize and integrate most relevant information by several vendors. Patient management systems, designed around the requirements of a radiation oncology unit, are available such as ARIA [18] by Varian and MOSAIQ [19] by Elekta. In theory, integration with third-party applications is not a problem; in practice, we see this is not always the case. Unfortunately, a vendor-independent radiation oncology-specific data model, i.e., xml standard or similar, is still not established.

Some departments are involved in developing their own models and systems, these with various capabilities, ranging from patient management to radiation delivery.

Innumerable levels and forms of information are present in radiation oncology and may be approachable with data mining, operations research, and machine learning [20]. These range from nonstructured free-text in writing (physician notes) to highly structured digital information (e.g., imaging and radiation dose distribution) (Fig. 5.2) [15].

A modern radiation oncology unit has been described to have three distinct computer systems: clinical medical electronic record, a computerized treatment planning system, and a record and verify system (R&V) [21]. Possibly to a lesser extent, but one should mention the quality assurance applications too. The electronic medical record may be integrated into a broader patient management system (PMS). Most of the time, the clinical medical records are distributed in the hospital PMS and the RO expert system, i.e., the R&V system. The information structure may vary from one system to another one. For example, an R&V system can have for historical reasons two SQL [22] database parts, at which the imaging part is still organized in a folder tree structure.

At this stage, the process of diagnosis and staging as well as the treatment prescription should be concluded and defined. The treatment prescription includes the description of the target volume as well as dose, including fractionation. Additionally, criteria for organs at risk are defined to enable the treatment planning process to commence. Images from the CT scanner, together with defined targets and organs at risk, provide the basis for treatment planning.

Practically, all hospitals have some form of centralized patient management system, the extent of which may vary significantly among institutions. Several vendors provide systems; these include solutions by Siemens, SAP, and many others.

A few vendors have specialized in integrating electronic medical records into a patient management system within a radiation oncology setting. The two most prominent providers are Varian with their product ARIA [18] and Elekta with their product MOSAIQ [19].