Workflow Challenges of Enterprise Imaging: HIMSS-SIIM Collaborative White Paper

There are several ways a modality worklist equivalent can be created. Some institutions use a health level 7 (HL7) admission, discharge, transfer (ADT) interface to mimic a modality worklist based on encounters. In this instance, the interface engine is used to perform an ADT to order transformation, effectively mimicking an order-based DICOM modality worklist to devices. Since DICOM modality worklists do not support visit entries, the order transformation allows the devices to select an actual order. The end result of this transformation is that the imager is able to select patient demographic information in a manner similar to how DICOM modality worklists are used. Images can be associated with the modality worklist entry either before or after imaging is performed.

Typically, the modality worklist can be accessed BEFORE imaging by “connected” or “smart” devices. Examples of these devices include smartphones, tablets, or other networked imaging devices. In this scenario, the imager could select the patient from the worklist before taking an image. Once the imaging is complete, the image is sent to the imaging archive along with patient demographic information. While this workflow may be ideal, it may limit hospital use of certain consumer technology such as digital cameras.

Another option would be to apply the demographic information to the image AFTER it is obtained. In this case, because the information is applied after the images are obtained, the same workflow considerations related to patient identification that are described above must be accounted for. In this scenario, the correctly identified images could be uploaded from within the patient context within the electronic medical record. In addition, this option requires providers to perform the additional steps required to upload images at a separate time point. There is a risk that image upload will not be performed by busy providers unless it is tied to image acquisition.

It is likely that a number of these scenarios may be used to upload data. If that occurs, it may be useful to develop a confidence score for the reliability of the data. System-entered data (such as data generated from a modality worklist) would have a higher confidence score compared to manually entered data.

Length and width measurements are key criteria in many disease processes. Simple measurements allow health care providers to quickly assess how a disease process changes over time [4, 5, 12, 18, 22]. The ability to perform a measurement is a key feature of DICOM-based imaging. For example, in radiologic images, one can measure the diameter of a tumor as the exact size of each pixel is known. In addition, certain DICOM-based studies, such as Doppler waveforms on echocardiography, allow the user to measure time. In the cases where the size information is not known, DICOM images can be calibrated using an imaged ruler so that measurements can still be obtained. This model should prove to be useful in non-DICOM imaging. Because of this, best practice might be that rulers be included in all images that do not have inherent size information. The standard use of a ruler helps to mitigate against differences in the appearance of a lesion due to zoom factor and the distance between the camera and the lesion.

Standardizing color values has not been addressed in the traditional imaging services. This is because most radiologic and cardiologic are imaged using shades of gray. When color is used, such as in nuclear medicine or color Doppler imaging, the hue is not as important as the intensity. This principle is not true in other specialties. In specialties such as dermatology, pathology, and wound care, creating reproducible colors is important [4, 5, 9, 15, 17, 23, 24]. However, reproducible color is challenging due to differences in lighting (incandescent versus fluorescent versus natural), shadowing, and camera settings.

Standardizing color may be difficult. Perhaps the best method for standardization is through the use of a medical photographer. A medical photographer can obtain images in a photo studio, controlling the light, background, and even camera settings. By having a standard protocol for imaging, the differences at each imaging time point are lessened [2, 4, 25]. While the use of medical photographers may be possible in an ambulatory setting, it is unlikely that this practice will translate to the inpatient setting, the operating room, or the emergency department. Therefore, other strategies for standardizing color should be considered.

Other potential options include incorporating a color wheel, grayscale, or white card within the photograph. Once photographed, a computer could then automatically calibrate the color to the accepted standard. There are several disadvantages of this system. First, each of these solutions must be printed, either on paper or plastic. In order to ensure consistency of the process, high-quality printers would have to be used. This would add cost to the hospital. Like printed identifiers, these printed color wheels could only be used in certain settings. However, both the cost and location issues could be mitigated by identifying the specialties where reproducible color is needed. The remainder of specialties could produce images without color calibration. Additionally, including items such as a color wheel, patient identification sticker, and ruler all add to the space required in the image. In order to account for this space, the image field of view may have to be increased thus decreasing the available space for the clinical region of interest.

It is also important to understand the color correction features of modern digital cameras and the potential effect that they may have on medical photographs [4, 5, 17, 24‐26]. Consumer camera manufacturers have developed a number of features that modify images automatically. While features such as red-eye removal, blemish correction, and image filters enhance personal photographs, they may mask medical conditions and provide unpredictable and unreproducible content for health care providers. In general, these features should be turned off when possible.

Standard positioning is crucial to many of the traditional imaging services. Because radiologists and cardiologists identify abnormalities based on pattern recognition, they rely on a standard appearance of body parts. This standard positioning allows them to quickly distinguish normal from abnormal. In specialties where imaging is not used for diagnosis, positioning may be less important. The major benefits of standard positioning in these specialties include the ability to correlate abnormalities with anatomic landmarks and the ability to document an abnormality in the same way at multiple time points [2, 4, 25]. Even if standard positioning is not needed in many specialties, it may be useful to annotate the location or patient positioning on the image for future reference.

In the setting of enterprise imaging, certain pieces of information increase in importance. Perhaps the most important piece of information for searching and sorting across specialties is the body part imaged. For example, a patient involved in a car accident may sustain a right knee injury. As part of his or her care, a photograph of injury is obtained in the emergency room. The patient then travels to radiology where an X-ray and MRI are obtained. Finally, the patient is brought to the operating room where he or she undergoes arthroscopic surgery. The images from all of these studies are related and should act (or at least have the potential to act) as relevant comparisons.

Radiology has a certain structure based on DICOM standard body parts [27]. While this structure works for radiology, it does not work in the setting of enterprise imaging. In some cases, the naming structure is too specific while in others, it is not specific enough. An example of a term that is too specific is the humerus. While this term makes sense in for an X-ray of the upper arm, it does not make sense for a picture of the skin of the same location. On the other hand, the term abdomen is used when, for surgical or pathologic images, this is not specific enough. In these specialties, it is important to refer to the specific organ being imaged, such as the liver or pancreas.

We recommend that a standard ontology be adopted for the purpose of body part mapping. The selected ontology should allow for synonyms and have a relational structure so that body parts can be considered parent and children terms. Thus, humerus and upper arm are considered to be synonymous. At the same time, humerus, forearm, hand, and fingers are all part of the upper extremity.

In radiology, study-level metadata is applied directly from the order. Perhaps the most important metadata applied to a radiology study is the procedure description. This one field may contain information relating to up to four variables. For example, the procedure description “RAD Hand 2-3V Right” tells the provider that the imaging study is (1) a radiograph (RAD) containing (2) 2–3 views of the (3) right (4) hand.

While the procedure description has worked in radiology, there are several limitations in the setting of enterprise imaging. First, there is no standard way to create a procedure description. This is best exemplified in the naming of a chest X-ray. Depending on the setting, this procedure can be named Rad Chest, Chest X-ray, Chest XR, or CXR among others. The American College of Radiology has identified this problem in creating the dose index registry [28, 29].

This problem is compounded by sites creating study-specific orders. For example, an MR enterography is a special type of MRI of the abdomen and pelvis performed to diagnose and monitor inflammatory bowel disease. Creating a study-specific order is useful for several reasons. It allows health care providers to order a specific study for a specific indication. A radiologist can then create a structured report based on this specific order. Finally, with study-specific orders, practices can better understand the ordering patterns of their providers. While there are benefits of this practice, the biggest disadvantage is the explosion of orderable procedures within the electronic medical record. The ACR has addressed this issue by asking providers to map their procedures to standard procedures identified in the RadLex playbook [28‐31].

As we move to enterprise imaging, it is possible that certain imaging workflows are encounter based, not order-based. In this scenario, there must be an efficient way to create a standard procedure description. Each procedure description should contain at least two pieces of information, the study type and the body part imaged including the side of the body. The study type helps to differentiate similar studies within the same department. This allows an ultraviolet photograph of the face be different from a visible light photograph of the face [19, 32]. Given the complexity of searching and creating hanging protocols in an enterprise viewer, it may make sense for the body part to be duplicated in a separate field.

In many current radiology and cardiology PACS, the department DICOM field is of limited value. As all of the images in the hospital come together, this field becomes crucial. It is important that a dermatologist can find all of the dermatology images quickly and distinguish them from radiologic images of the same body part. Image viewers must be able to allow users to search and sort based on this field.

Like the body part field, this field should be standardized. The biggest challenge in standardizing the department metadata is identifying the correct level of granularity for each specialty. While in most instances, it makes sense to use named divisions such as cardiology and gastroenterology in the department field, in other specialties such as radiology, this may not make sense.

Because traditional imaging modalities represent a significant capital expenditure, medical imaging has been performed exclusively by medical practices. However, this paradigm is shifting. As digital cameras become increasingly prevalent, many new imaging providers exist. This includes patients taking photographs of themselves to share with a health care provider. As patients begin to upload their own images to an electronic medical record or enterprise imaging archive, there will likely be a need to be able to tag studies as either patient obtained or provider obtained. The difference may be important for quality assurance purposes, liability concerns, and even reporting related to meaningful use.

There are unique privacy concerns that exist with certain types of photographs. It is recognized that institutions may choose to limit access to certain types of images such as images related to sexual assault or child abuse. In these scenarios, there are several options vendors may choose to allow institutions to manage permission to view these images. Potential options are listed below.

First, electronic medical record systems may be used to manage the permissions to view images. In this case, permissions may need to be applied for all images related to a specific department such as plastic surgery or to images related to a specific procedure description such as sexual abuse imaging from the emergency department. In each scenario, it is important that users have access to view the report within the electronic medical record, but not view the images related to the report. Because users could access images either via the electronic medical record or the image viewer, users should be restricted in both systems. However, ideally, only one system is required to manage permissions.

A second potential mechanism for sensitive images may be to request permission on a case-by-case basis. If vendors and institutions select this option, a user would have to select a reason why he or she needs to view the restricted images. These reasons would be recorded and could later be audited. Both the viewer and electronic medical record system would identify and flag users who frequently view restricted images. These users could then be audited to ensure that they are viewing restricted images in line with their patient care duties. Ideally, a permissions-based system could be coupled with an override capability so that certain departments such as plastic surgery are not over-burdened in an attempt to view images relevant to their practice.

Another challenge related to restricted images is image sharing. In the setting where sensitive images are considered part of the medical record, hospitals need to determine mechanisms for the types of images they are willing to share and the permission required to share these images. It is possible that hospitals choose to flag certain images and require a separate permission/consent to share these images with other institutions or providers.

In the analog era, images were printed on physical media. Because there was a direct cost associated with each printed image (both for film and for paper), fewer images were obtained and printed. In the digital era, most of the direct costs have been removed and images exist as a collection of bits and bytes on a server. While there are costs associated with storage, this cost is often small per megabyte and removed from the imager.

In the consumer market, we have devices that that can take and store thousands of images. Many of us now have archives of thousands of personal digital photographs. As such, we have an expectation of keeping everything. In the medical world, this may not be needed or appropriate. While this decision will likely play out in individual hospitals, through legal case precedent, and possibly legislation, it is important to think of this concept ahead of time.

Several states have legislated (or mandate through legal precedent) that radiology images be maintained for a certain amount of time. While most hospitals maintain the radiology images obtained at their institution, they do not archive or keep all images across the enterprise. In fact, each department or division maintains their images differently. In many cases, this even varies within a department. For example, in radiology, some images or series are rejected at the modality (such as due to motion) and never sent to the archive. In ultrasound, the technologist may select the images he or she wants to capture during live imaging. In most cases, the entire study is not recorded. Other specialties have a similar workflow. Specialists may review all of the images at the end of a procedure and select the key images that document the finding or the procedure. The other lower quality or less representative images may be discarded.

The use of video as a method of documentation will likely explode in the coming years. This new data source has the potential to cause a significant increase in the size of the imaging archive. As video is obtained, hospitals may choose to place a limit on the length of the video or the types of exams or procedures that are recorded. It should be noted that video editing may be overly burdensome. Technology that allows providers to prospectively and retrospectively select segments to record should be encouraged. For example, an endoscopist should be able to push a button or press a foot pedal to begin recording. In addition, he or she should also be able to press another button or foot pedal to capture the previous 1–3 min of video. Finally, the process of video editing is not without risk. Hospitals will have to determine who is allowed to edit the video, which elements are required to sufficiently document each procedure, and how procedures or examinations are selected for recording.

As cameras improve, the file size for each image is increasing in size. This has the potential to cause a significant increase in the size of the imaging archive. In radiologic imaging, the U.S. Food and Drug Administration (FDA) requires that when an image is displayed, it be labeled with a message stating if irreversible compression has been applied and with approximately what compression ratio [36]. In addition, the FDA states in the Mammography Quality Standards Act that irreversible compression of digital mammograms for retention, transmission, or final Interpretation is not allowed, although irreversibly compressed images may be used for images from prior studies [37]. The ACR-AAPM-SIIM technical standard for the electronic practice of medical imaging makes no statement on the type or amount of compression that is allowed for a medical image to be called “diagnostic [34].” The European Society of Radiology suggests that image compression may be acceptable for certain specialties or in certain imaging types such as teleradiology or in long-term image storage and unacceptable in other specialties or for other imaging types such as in mammography, radiation oncology, or surgical navigation [38]. It thus follows that allowing for lossy images in certain scenarios or with certain file types may be allowed. If vendors enable multiple methods of image compression, it may allow to improve viewer performance and reduce storage requirements.

If lossy images are allowed, there should be a balance between the amount of compression and the subsequent image quality. With photographic images, the degree of lossiness is not normally expressed in a compression ratio but a standard specific quality factor. For example, JPEG lossy with a quality factor of 85 or better is seen as “high quality” or “visually lossless” to most people. If medical photographs were saved using a high-quality factor, significant space savings may be achieved without apparent loss of image quality.

File format is closely related to image fidelity. Some file formats, like joint photographic experts group (JPEG), can be considered lossy, while others, like tagged image format (TIF) or RAW formats, are considered lossless. It should be noted that lossless file formats such as RAW may not be standard across vendors. As hospitals move toward enterprise imaging, they may choose to select lossy formats for standard image storage. This decision may be based on several factors including file size, industry standards, the ubiquity of format, and the capabilities of enterprise image viewers. In addition, files can be viewed natively or wrapped using an industry standard format such as DICOM. The discussion of these options extends beyond the scope of this white paper. Further information on the subject can be found in the white paper entitled Technical Challenges of Enterprise Imaging.