The field of data visualization is much broader than most people conceive of it, and exploring this breadth was one of our primary goals in compiling the projects described in “Beautiful Visualization.” In the following excerpt, Anders Persson of Linköping University in Sweden explains how radiological digital imaging methods allow medical practitioners to conduct “virtual autopsies” without the use of a scalpel or any other invasive instrument.
Warning: Some readers could find the forensic illustrations in this post too graphic for their tastes. You might want to skip this one if you’re squeamish.
The following was written by Anders Persson:
This chapter’s topic is extremely important to those who work in the field of medical information visualization. Emerging technologies are enabling visual representations and interaction techniques that take advantage of the human eye’s broad-bandwidth pathway into the mind, allowing users to see, explore, understand, and validate large amounts of complex information at once.
A striking feature of both clinical routine and medical research today is the overwhelming amount of information — particularly, information represented as images. Practitioners are dealing with ever-larger numbers of images (hundreds or thousands rather than dozens) and more complex, higher-dimensional information (vectors or tensors rather than scalar values, arranged in image volumes directly corresponding to the anatomy rather than flat images). However, they typically still use simple two-dimensional devices such as conventional monitors to review this overflow of images, one by one. As the bottleneck is no longer the acquisition of data, future progress will depend on the development of appropriate methods for handling and analyzing the information, as well as making it comprehensible to users. One of the most important issues for the future is the workflow. The entire chain from the acquisition of data until the point at which the clinician receives the diagnostic information must be optimized, and new methods must be validated.
Normally, performing this validation process on living patients has its limitations. It can in some cases be impossible to know if the acquired diagnostic information is correct as long as the patient is alive; the real gold standard is missing. Postmortem imaging has the potential to solve this problem.
The methodology of autopsy has not undergone any major transformation since its introduction in the middle of the 19th century. However, new radiological digital imaging methods, such as multidetector computed tomography (MDCT) and magnetic resonance imaging (MRI), have the potential to become the main diagnostic tools in clinical and forensic pathology in the future. Postmortem visualization may prove to be a crucial tool in shaping tomorrow’s healthcare, by validating new imaging technology and for quality assurance issues.
The importance of autopsy procedures leading to the establishment of the cause of death is well known. In forensic cases, the autopsy can provide key information and guide the criminal investigation. The decreasing trend in the frequency of autopsies over the past years has become a serious issue.
A recent addition to the autopsy workflow is the possibility of conducting postmortem imaging — in its 3D version, also called virtual autopsy (VA) — using MDCT or MRI data from scans of cadavers and with direct volume rendering (DVR) 3D techniques. At the foundation of the VA development are the modern imaging modalities that can generate large, high-quality datasets with submillimeter precision. Interactive visualization of these 3D datasets can provide valuable insight into the corpses and enables noninvasive diagnostic procedures. Efficient handling and analysis of the datasets is, however, problematic. For instance, in postmortem CT imaging, not being limited by a certain radiation dose per patient means the datasets can be generated with such a high resolution that they become difficult to handle in today’s archive retrieval and interactive visualization systems, specifically in the case of full body scans.
Several studies have shown the great potential of virtual autopsy in forensic investigations. This chapter will investigate several of the reasons for the rising interest in VA.
Impact on forensic work
The main questions to be assessed in examinations of the deceased are the cause and manner of death and the severity of injuries suffered, as well as the possibility of forensic reconstructions based on the obtained findings. Forensic documentation of postmortem findings is predominantly based on the same autopsy techniques and protocols that have been used for centuries. The main tools used are scalpels, verbal descriptions, and photographs.
A major disadvantage of this approach is that the documentation happens in a haphazard, subjective, and observer-dependent manner. Any findings that have not been documented are irreparably destroyed when the cadaver is sent to the crematory. Modern cross-sectional imaging techniques can overcome these shortcomings, as they provide datasets of cadavers that contain the findings in real dimensions and are storable for the future (Figures 18-1 and 18-2). The digitally acquired data can be referred to at any time as new questions arise, or may be sent to additional experts for a second opinion.
Some findings that are difficult to visualize in a conventional autopsy can easily be seen in a full body CT, such as air distribution within the body — e.g., in the pneumothorax, pneumopericardium, bloodstream (air embolism), and wound channels (Figure 18-3). A CT can also be invaluable for locating foreign objects such as metal fragments and bullets, which are of great importance for the forensic pathologist (Figure 18-4).
The virtual autopsy procedure
The Center for Medical Image Science and Visualization (CMIV) at Linköping University Hospital in Sweden, in collaboration with the Swedish National Board of Forensic Medicine, has developed a procedure for virtual autopsy that is now used routinely for forensic work. This method has been in use since 2003 and has been applied to over 300 cases so far (mostly homicides).
Our experience with VA has shown that full-body, high-resolution DVR visualizations are of great value in criminal investigations and for the validation of new technologies on living patients. Our work has focused on the total workflow for postmortem MDCT and on developing a new type of software that can visualize full-body datasets that could previously only be viewed in separate parts and with limited interactivity (Figures 18-5 to 18-7).
The traditional physical autopsy at CMIV is extended by adding the CT and MRI as VA activities. In most cases, the forensic pathologist comes to the crime scene and oversees the handling of the human cadaver, which is placed in a sealed body bag before being transported to the forensic department and put in cold storage. The following morning, a full-body dual source CT (DSCT) scan is performed at CMIV with a state-of-the-art SOMATOM Definition Flash scanner (from Siemens Medical Solutions in Germany). Currently, both single- and dual-energy modes are used for virtual autopsy cases; see Figure 18-8(a) and (b).
In selected cases, an MRI examination is also performed (using an Achieva 1.5T scanner, from Philips Medical Systems in The Netherlands). All children are routinely examined with MRI, because it offers superior visualization of the brain compared to DSCT (Figure 18-9). The cadaver remains in the body bag throughout the virtual autopsy procedure to ensure the security of technical evidence of forensic value, such as fibers and body fluids, and to avoid contamination.
Computed tomography: Use of dual energy CT
Dual energy CT (DECT) with two x-ray sources running simultaneously at different energies can acquire two datasets showing different attenuation levels. DECT allows additional information about the elementary chemical composition of CT-scanned material to be obtained. Compton scattering can be determined by using two different average photo energies, which correspond to two different tube voltages (80 and 140 kV).
In other words, x-ray absorption is energy-dependent — e.g., scanning an object with 80 kV results in a different attenuation than scanning it with 140 kV. This physics phenomenon can help to discriminate between materials with similar atomic numbers, such as calcium and iodine contrast. Colors can then be assigned according to changes in the CT numbers between the two energy settings, and the resulting color-mapped, dual-energy image can differentiate between calcifications and iodine contrast.
This technique can also be used to better visualize postmortem blood clots in vessels, and possibly bleeding in soft tissue. The material-specific difference in attenuation shown in the resulting image could facilitate classifications of different tissue types such as blood, soft tissue, tendons, and cartilage (Figure 18-10).
DECT has the potential to be an important diagnostic tool in the healthcare of tomorrow. However, further research needs to be done to explore this new technique. VA can speed up this research.
MRI: Use of synthetic magnetic resonance imaging
It is difficult to generate good contrast MRI images on dead, cold bodies — body temperature influences the MR relaxation times of all tissues, and hence clinically established protocols need to be adjusted for optimal image quality at any given temperature. This problem can be solved by measurement of the absolute MR tissue parameters for tissue characterization, T1, T2, and proton density (PD).
Since this can be difficult to implement on a clinical MRI scanner, a new approach has been invented at CMIV called synthetic MRI. In this approach, the three absolute parameters are translated into ordinary MR contrast images (Figures 18-11 and 18-12). A color scale can be used such that each tissue acquires a specific color composition depending on its MR tissue parameters, independent of body temperature. Since the MR parameters are absolute, an identical color transformation will lead to a specific color-to-tissue relation, and a visual segmentation of tissue. Especially for postmortem imaging, this is important, since the image contrast may vary dramatically with temperature (Figure 18-12).
Postmortem examinations do not suffer from motion artifacts, and high-resolution images can be obtained with a long scan time. An example is shown in Figure 18-13, which shows a head shot wound in 1.2 mm isotropic resolution. Since synthetic MRI is based on absolute values, it can be used to render 3D images with CT postprocessing software, resulting in the volume renderings displayed in Figures 18-13 and 18-14.
Visualization: image analysis
In preparation for the physical autopsy, the pathologist and the radiologist conduct a collaborative DVR session. They can obtain a clear survey of the entire body quickly, and localize fractures and air pockets. The full-body procedure permits fast localization of foreign objects such as metal fragments or bullets. Another important aspect is the high resolution of the data, which, in a seamless visualization, allows details such as dental information to be extracted for identification purposes (Figure 18-15). This can provide essential information in the early part of a police investigation. After scanning, the forensic personnel leave CMIV and start the conventional autopsy. Data from the collaborative DVR session is transferred to the forensic institute for them to use, and if more information is needed later, new contact with the radiologist is made.
An important added value of the virtual autopsy procedure is that the captured DSCT data is stored, which enables the procedure to be iterated. Often, findings during the physical autopsy lead to new questions that the VA can answer. The pathologist and the crime investigators can also — at any point during the investigation — re-examine the cadaver and search for additional information (Figure 18-16). Moreover, in crime scene investigations, new findings may require other hypotheses to be scrutinized by postmortal imaging.
VA is currently used as a complement to the autopsy procedure. It should, however, be noted that the workflow overhead introduced is minimal, as the time needed for the DSCT scan and visualization session is short in comparison to the physical autopsy, and that it can make the autopsy more efficient because the pathologist will have prior knowledge of the case before beginning the autopsy. That the cadaver remains in a sealed body bag throughout the VA procedure also secures technical evidence, such as fibers and body fluids, which in forensic cases may be of great importance.
Advantages and disadvantages of virtual autopsy
Let’s take a look at the advantages of VA compared with conventional autopsies:
– It is time-saving. The VA can be a complement to standard autopsies, enabling broad, systematic examinations of the whole body that are normally difficult and time consuming; for example, an examination of the entire bone structure or searching for the presence of air in the body (Figures 18-3 and 18-4).
– It is noninvasive. Once an invasive traditional autopsy has taken place, the body cannot be reassembled in its original state, thus precluding other forensic pathologists from conducting a fresh analysis on the same body (Figures 18-5 through 18-7).
– A traditional autopsy may be rejected by family members, perhaps due to religious beliefs that prohibit the desecration of the remains of a deceased person. For example, Orthodox Judaism prohibits disturbing dead bodies except when such action may save others, and decrees that practices such as organ removal should be avoided. Islam is likewise opposed to desecrating or even exposing the body of a deceased believer.
– Autopsy protocols and photographs used as evidence in criminal cases can be difficult for jurors to understand. VA visualizations are typically clearer (Figures 18-4 and 18-9).
– Storage of VA data poses few problems, whereas autopsy records such as tissue sections are difficult to store indefinitely (Figure 18-16).
– With potential global pandemics such as bird flu (avian influenza A) and swine flu (the H1N1 virus) posing an increasing threat, the practice of eviscerating the victims can pose serious health risks to coroners, pathologists, and medical examiners. With a VA, these risks are minimized.
However, virtual autopsies also have several shortcomings:
– For MDCT, soft tissue discrimination is low. Energy-resolved CT (DECT) has the potential to resolve this problem (Figure 18-10).
– The large amount of data produced is a problem to analyze, but better and faster postprocessing programs should solve this.
– MRI is a time-consuming investigation and not optimal on a cold body. Synthetic MRI is a promising alternative (Figure 18-14).
– Postmortem imaging with MDCT and MRI does not give any color documentation of the body. It may be possible to solve this issue with new volume-rendering 3D methods and body surface scanning (Figure 18-15).
– Macro morphology is absent (no histology and chemistry). This can be solved to a certain extent with MDCT guided biopsies or magnetic resonance spectroscopy (Figure 18-16).
– Circulation and possible bleeding points are difficult to visualize, although promising results have been achieved with postmortem angiography. As has been shown, postmortem CT angiography can be a feasible way to obtain more information from the VA (Figure 18-17).
– Postmortal gas can be difficult to distinguish from other types of gas (bowel gas, gas in wound channels, etc.). Therefore, it is important to execute the postmortem imaging examination soon after death has occurred (Figure 18-18).
The future for virtual autopsies
Both MDCT and MRI can be used for postmortem imaging. In principle, it is easy to visualize bone, gas, and metal with MDCT. However, it is important to be aware of not only the capabilities, but also the limitations of these technologies.
Visualization research in the future must include the overall aim of implementing a virtual autopsy workstation that includes everything needed to perform state-of-the-art virtual autopsies. Visualization tools to increase the quality and efficiency of virtual autopsy procedures need to be developed. Research and development efforts focusing on novel rendering and classification techniques are also needed to improve usability and to specifically address forensic questions. Another important goal is to establish designated protocols for the main forensic case categories.
The data analysis research includes the implementation of computer-aided diagnostic tools that can, once applied to the postmortem data, help search for and characterize relevant forensic findings. These tools can also deliver general information about the deceased individual such as height, body weight, sex, major injuries, foreign bodies (e.g., projectiles), and likely causes of death in an automatically generated preliminary, written virtual autopsy protocol.
When all of these tasks have been successfully addressed, the technology involved within all processes of a virtual autopsy can be improved to enable automation of the entire workflow. This will allow for virtual autopsies to be performed in large numbers within a reasonable time frame. This would be invaluable in handling incidents with significant numbers of victims such as those created by the tsunami catastrophe in Asia in 2004, where no autopsies were performed at all.
As terrorists improve their applied technologies day by day, it is unthinkable that forensic pathologists should not also be able to make use of emerging technologies in order to gather as much information as possible from their victims (Figure 18-19). In times where no one can really feel safe, we should not only focus on the prevention of catastrophe, but also prepare ourselves to handle disasters adequately when they do occur.
For a new era of digital autopsies to truly emerge, several forces must work in unison. Medical professionals and legal authorities must determine standard protocols for scanning and storing data. Legal systems around the world must accept the admissibility of imaging evidence in determining the cause and manner of death. Also, specialists in new fields such as postmortem radiology will need to be trained. Radiologists are typically trained to interpret images of living patients, but the dead often look different; severe trauma or the effects of decomposition can displace organs. Understanding these differences will require knowledge and expertise that does not exist on a widespread basis today.
Invasive autopsies will likely remain the norm for at least the next few years. However, in some cases, we may begin to see traditional autopsies being replaced by noninvasive virtual autopsies, with minimally invasive, image-guided tissue sampling conducted when necessary. Postmortem VA has the potential to gain high acceptance in the population compared with the traditional autopsy, making it possible to maintain high levels of quality control in forensic and traditional medicine.
The virtual autopsy is a newly developed procedure that will enhance the classic autopsy, giving it the capacity to achieve more reliable results. In some cases, the virtual autopsy could also replace the normal autopsy. Research on the unique aspects of postmortem radiology must, however, be undertaken to identify cases in which its use is most beneficial and to validate the new procedures. Clearly, the introduction of this new autopsy method is likely to have a major impact on forensic medicine, the judicial system, the police, and general medicine in the future.
- Book: Beautiful Visualization
- Excerpt: Redesigning the New York City subway map
- Data visualization primer: What they are and why they’re important
Get the “Beautiful Visualization” ebook for $9.99. Use discount code BLV99.
References and suggested reading
Donchin, Y., A.I. Rivkind, J. Bar-Ziv, J. Hiss, J. Almog, and M. Drescher. 1994. “Utility of postmortem computed tomography in trauma victims.” Journal of Trauma 37, no. 4: 552-555.
Etlik, Ö., O. Temizöz, A. Dogan, M. Kayan, H. Arslan, and Ö. Unal. 2004. “Three-dimensional volume rendering imaging in detection of bone fractures.” European Journal of General Medicine 1, no. 4: 48-52.
Jackowski, C. 2003. “Macroscopical and histological findings in comparison with CT- and MRI- examinations of isolated autopsy hearts.” Thesis, Institute of Forensic Medicine. O.-v.-G.-University of Magdeburg.
Jackowski, C., A. Persson, and M. Thali. 2008. “Whole body postmortem angiography with a high viscosity contrast agent solution using poly ethylene glycol (PEG) as contrast agent dissolver.” Journal of Forensic Sciences 53, no. 2: 465-468.
Jackowski, C., W. Schweitzer, M. Thali, K. Yen, E. Aghayev, M. Sonnenschein, P. Vock, and R. Dirnhofer. 2005. “Virtopsy: Postmortem imaging of the human heart in situ using MSCT and MRI.” Forensic Science International 149, no. 1: 11-23.
Jackowski, C., M. Sonnenschein, M. Thali, E. Aghayev, G. von Allmen, K. Yen, R. Dirnhofer, and P. Vock. 2005. “Virtopsy:
Kerner, T., G. Fritz, A. Unterberg, and K. Falke. 2003. “Pulmonary air embolism in severe head injury.” Resuscitation 56, no. 1: 111-115.
Ljung, P., C. Winskog, A. Persson, C. Lundstrom, and A. Ynnerman. 2006. “Full-body virtual autopsies using a state-of-the-art volume rendering pipeline.” IEEE Transactions on Visualization and Computer Graphics 12, no. 5: 869-876.
Oliver, W.R., A.S. Chancellor, M. Soltys, J. Symon, T. Cullip, J. Rosenman, R. Hellman, A. Boxwala, and W. Gormley. 1995. “Three-dimensional reconstruction of a bullet path: Validation by computed radiography.”Journal of Forensic Sciences, 40, no. 2: 321-324.
Ros, P.R., K.C. Li, P. Vo, H. Baer, and E.V. Staab. 1990. “Preautopsy magnetic resonance imaging: Initial experience.” Magnetic Resonance Imaging 8: 303-308.
Thali, M., W. Schweitzer, K. Yen, P. Vock, C. Ozdoba, E. Spielvogel, and R. Dirnhofer. 2003. “New horizons in forensic radiology: The 60-second digital autopsy-full-body examination of a gunshot victim by multislice computed tomography.” The American Journal of Forensic Medicine and Pathology 24: 22-27.
Thali, M., U. Taubenreuther, M. Karolczak, M. Braun, W. Brueschweiler, W. Kalender, and R. Dirnhofer. 2003. “Forensic microradiology: Micro-computed tomography (Micro-CT) and analysis of patterned injuries inside of bone.” Journal of Forensic Sciences 48, no. 6: 1336-1342.
Thali, M., K. Yen, W. Schweitzer, P. Vock, C. Boesch, C. Ozdoba, G. Schroth, M. Ith, M. Sonnenschein, T. Doernhoefer, E. Scheurer, T. Plattner, and R. Dirnhofer. 2003. “Virtopsy, a new imaging horizon in forensic pathology: Virtual autopsy by post-mortem multislice computed tomography (MSCT) and magnetic resonance imaging (MRI) — a feasibility study.” Journal of Forensic Sciences 48, no. 2: 386-403.
Yen, K., P. Vock, B. Tiefenthaler, G. Ranner, E. Scheurer, M. Thali, K. Zwygart, M. Sonnenschein, M. Wiltgen, and R. Dirnhofer. 2004. “Virtopsy: Forensic traumatology of the subcutaneous fatty tissue; Multislice Computed Tomography (MSCT) and Magnetic Resonance Imaging (MRI) as diagnostic tools.” Journal of Forensic Sciences 49, no. 4: 799-806.