3 Medical Imaging
Radiology gave to the physician a look of the inside the human body without the need for surgery. Images diagnostics has evolved over time, with the introduction of new imaging tools, protocols and imaging technologies based on the use of ionizing radiation and other physical principles. The current diagnostic equipment are accurate, smaller and emits less radiation than before. Moreover, the most recent tools allow to acquire diagnostic data in digital form, enabling data post-processing using dedicated software. Diagnostic images have gone from having only a diagnostic purpose to being often the foundation for treatment planning. At the root of modern digital diagnostics is the DICOM standard (Digital Imaging and COmunication in Medicine). The standard is freely available online and it can be freely used 27. On this Standard is based the whole chain of modern patient management, from the acquisition of clinical and anamnestic data to diagnostic images, up to the transfer of data between different health workers and institution 25.
The human organism is a three-dimensional object oriented in space. In
anatomy we describe the human organism located in the Anatomical
Position.
Scansion of the human organism can be performed with different
techniques and tools. The scan can represent only the surface of the
organism, or also capture its internal organs.
Computer Tomography (CT) is a technique that uses beams of X-rays that
pass through the human body and are detected by a scanner, that evaluate
the energy absorbed during the path, and therefore the density of the
tissues (in Hunsfield Unit). This measurement is performed from various
angles on the same plane, and the readings are processed by software,
which returns an image that is a map of the density of the scanned
volume (slice).
The slice is a grid of voxel (whose resolution depends on various
factors, including the device used and the FOV (Field Of View), always
according to the ALARA principle), in which each voxel is represented
in a shade of gray proportional to the attenuation of the X ray, by the
tissue founded in the represented volume. The slices are acquired, one
after the other, at a thickness (slice thickness) and a distance
(pitch, 23) that depend on the used tomograph and the
implemented scanning protocol. The acquired slices are then ordered to
form a series, and each slice is interpolated with the subsequent,
allowing to obtain a volume.
The volume thus constituted can be studied on arbitrary planes of space,
thanks to multiplanar reconstruction algorithms (MRP). The volumetric
representation thus created is of great importance, because it is the
base of modern radiological diagnosis and therapy.
CBCT is a technique that uses the same physical principles as CT, but
presents differences due to scan procedure and reconstruction algorithms
12, 13.
The CBCT scanner uses a conical beam of X-rays, which is rotated around
the region to be scanned (ROI - region of interest) and is
detected by a sensor, placed orthogonal to the beam. The conical beam
allows to acquire a volume during the revolution, so the scans are
faster and with a lower dose of radiation absorbed by the patient than a
conventional CT 14.
CBCT equipment is smaller and mechanically less complex than
conventional CTs; these characteristics, associated with the scanning
speed and the lower cost, have made it widespread, especially in the
diagnostic of the oromaxillofacial region. The reduced FOV allows to
reduce the absorption of radiation and to provide high resolution 3D
reconstructions. Interesting is also the fact that these devices, given
their small size, have been used as an intraoperative aid during surgery
15 and during endodontic therapy procedures in selected
cases 16.
The diagnostic images obtained from the CBCT scan is the result of a
first processing step, which allow to map in a three-dimensional space
the data collected by the detector 11. The images collected
are processed with algorithms to reduce artifacts due to the scanning
process. The scattering noise is prominent in the unprocessed images,
so it is reduced through an image blur process (blurring or
smoothing), finding a compromise with the preservation of the spatial
resolution of the images (which is higher at low levels of blur, ie.
when the image is well defined, edge sharpening).
Of relevance, especially in the implant surgery planning, is the
assessment of bone quality in the implant site. Historically, bone
density, measured in Hounsfield Unit (HU), has been the most used
parameter to describe bone quality. Today’s CBCT scanners do not give
Hounsfield’s scale images, but in a gray scale (Gray Value, GV) which
is the representation of the attenuation of the X-ray signal by the
tissues found in the path of the beam.
It was shown how the grayscale representation is dependent on the used
scanner, as well as being highly variable even between different scans
acquired with the same device. Differences in the representation of the
density can be highlighted between the central and the peripheral area
of an image, both on the horizontal and vertical axis. To partially
obviate to this problem, the scanner must be calibrated with test
samples provided by the manufacturer; variability remains even after
calibration 17, but is considered acceptable for normal
procedures. However, the search for algorithms and techniques to reduce
scanning artifacts is one of the fields in which innovations are most
frequently presented.
CBCT scaning device use a lower slice-thickness than conventional CT,
improving the representation of small details, as the endodontic canal.
Magnetic Resonance Imaging (MRI) uses magnetic fields to orientate
water’s molecules within the organism and to detect their position,
through the energy emitted when they return to their initial
configuration, after the magnetic field has been switched off. The scans
are repeated to intensify the signal. The image obtained is in
grayscale, with the richest areas of water showing a greater intensity
(white) than areas with little water (black). MR equipment uses
different parameters for detection, with protocols that allow to
suppress or increase the signal intensity of a tissue.
MRI does not emit ionizing radiation, but it can be risky for the
patient with metal implants or pacemakers, so these conditions must be
assessed before subjecting the patient to the scan.
This imaging technique allows good visualization of the body’s soft
tissues, but bone margins are difficult to detect, especially at the
level of alveolar processes. Some groups have however started to explore
the potential of MRI also in the diagnosis of bone pathology
18 and its microstructure 19.
Surface scan makes possible to digitize the body surface and the
cavities accessible by the scanning equipment.
In the dental field various optical scanning devices have spread,
replacing the plaster impressions. These scanners allow to obtain
digital impressions of the patient’s arches, which are important for the
subsequent therapy, that very often makes use of CAD/CAM devices for the
design and manufacturing of temporary restoration and prosthesis.
There are several scanning technologies, but the most used in the
medical field are techniques that reconstruct a three-dimensional
surface starting from a series of images of the region to be digitized.
The instruments required for scanning usually consists of an image
acquisition device, a computer and data processing software. The
software assists the operator in the images acquisition, which are often
processed in real time to provide the clinician with an instant view of
the model being acquired. The three-dimensional model thus created can
also contain the color of the scanned tissue (textures)
20.
The acquired models can be used in combination with models made by CT or
MRI, where the oral tissues are less defined than the models obtained
with surface scanners, and may present artifacts due to the presence of
prostheses or metal restorations, in addition to not having information
of the surface textures. This combination allows the creation of high
quality models to accurately plan orthognathic surgery procedures
21, 22.
Digital reconstructions of real objects can also be performed by means
of a camera, with a technique called Photogrammetry. This consists in
acquiring a series of partially superimposed pictures around the object
to be scanned; the images are then processed by algorithms to give a
digital 3D model 117.
Digital images share the problem of partial volume effect. It
consists in the fact that a voxel can represent a single shade of gray,
so if in the volume of a voxel there is more than one tissue, perhaps at
different densities, the density represented in the voxel will be a
weighted average of the densities present in the voxel volume. A
decrease in the size of the voxels (and therefore an increase in the
number of voxels with the same volume) makes it possible to partially
compensate for the problem.
This problem is always to be taken into account during image
segmentation, because it affects the quality of the details and the
discrimination of the boundary of different tissues.
In optical scans, a frequent problem is the presence of noise on the surface of the scan. This is due to various causes, but this problem can be solved with algorithms that calculate the average of the curvature of the surface and remove the points that deviate from it (the variance in the distribution of points is often an adjustable parameter in the cleaning algorithm ).
DICOM files contain sensitive patient data, such as general information,
medical history and diseases, as well as medical images. These files are
often sent to colleagues for advice, shared for research purposes or
shown to students during teaching. There are online libraries where
diagnostic image sets are loaded for research purposes, and these are
freely accessible on the web 24. To allow a more secure
management of images preserving the possibility of sharing, several
methods have been proposed.
Essential is the protection of patient data in the network in which the
data is collected. Computers and servers should be protected by a
firewall, while data should only be sent via a VPN. But this is not
enough to protect the data when they have to leave the original network.
This is why the data must be anonymized or encrypted.
Anonymization consists of removing entries containing sensitive data
from the DICOM file. However, the right balance between the data removed
for security and those to be maintained must be found, because items
such as date of birth, sex and date of execution of the scan are
important both in the subsequent study of the images and for their
management.
Anonymization is not always necessary if images are to be kept in a
private and protected archive, but should be used when the file is
spread and there is a risk to the patient’s privacy.
Several software, open-source or commercial, have been proposed for the
scope, for example DICOM Confidential 46.
Anonymisation is an irreversible process, because is impossible to
retrieve data after that they have been removed from the DICOM file.
Encryption is essentially the process of translating data into another
form to prevent them from being easily understood.
Replacing every letter of a word with the following letter in the
alphabet (hello -> ifmmp) is an example of symmetric key
cryptography, because the key (replacing every letter of the message
with the next to encrypt and with the previous one to decrypt) is the
same for both parties that exchange the message. The asymmetric key
algorithms work with a key for encryption (public key) and a key for
decryption (private key).
RSA is an asymmetric key algorithm, whereas AES is a symmetric
key algorithm; both are supported by the DICOM Standard. The choice of
the algorithm must however be made not only looking at security, but
also at the time of encryption-decryption. RSA is currently considered
very safe, but it is also slower to calculate; for this reason RSA is
used to generate keys, which is a process that is performed
infrequently, while AES is used with RSA keys to encode data
25.
Encryption is a completely reversible process, because there is no
loss of data during the process.
There are many ways to exchange data, and this flexibility comes at the
risk that the data is somehow changed without permission. To solve this
problem we can use another tool borrowed from cryptography: the hash
function. The hash function is essentially an algorithm that takes as
input an arbitrary sequence of bits, and give as output a standard
sequence of bit. When the input is changed, there is a very high
probability that the output will change.
If, for example, in a CT the value of one or more pixels is changed, or
the date of acquisition is changed, when the hash is recomputed the
output will not match the original hash, which indicates that the data
are corrupted.
QuickHash GUI is a cross-platform open-source software with a graphical
interface, which allows to create hash of files and folders and to check
their authenticity 65.
When two operators exchange data, who receives the data must be sure of the validity of the sender. With digital data, this confirmation is given by the digital signature. The digital signature is a sequence of characters that is issued by trustworthy entities after ascertaining the credentials of the applicant. The recipient of the data will then be able to discriminate the valid sender relying on his signature.