Siemens Medical Solutions (Erlangen, Germany) has announced the
formation of a new division, Image and Knowledge Management (IKM),
which combines several information technology groups, including the
computer-aided diagnosis (CAD) group.
Applied Radiology
spoke with Alok Gupta, PhD, MBA, Vice President of Siemens'
Computer-Aided Diagnosis and Knowledge Solutions (Malvern, PA),
about the change and the state of CAD technology in the United
States.
Applied Radiology:
Why was the new IKM division created?
Alok Gupta:
At Siemens Medical Solutions, we have a long-term vision that the
data and knowledge contained in information technology¨C¨Csuch as
patient data and clinical knowledge¨C¨Ccan help to improve quality
and manage costs of healthcare because they improve the efficiency
and accuracy of clinical processes. The IKM division was formed by
combining 3 Siemens businesses: radiology information system (RIS)
and PACS;
syngo
software platform; and the CAD group. Each of these groups were
looking at different knowledge management and intelligent data
processing and image processing activities. Combining these
businesses shows our commitment to strengthening these
technologies.
AR:
What are the main goals and objectives for the CAD and Knowledge
Solutions group?
AG:
We were formed 4 years ago to look for information and knowledge
around clinical processes that can be brought together in
applications that are used across the patient-care continuum to
include early detection, accurate diagnosis, therapy decisions, and
follow-up. These applications must be integrated in our overall
product portfolio. For example, our applications are seamlessly
integrated into our PACS and multimodality workstations.
AR:
Which products are in this group's portfolio?
AG:
Our group includes anything related to CAD, ie, colon polyp and
lung polyp detection applications. This group conceived them, did
the research, did the laboratory premarket approval (PMA), and
brought them to market, including integrating them into the
application.
AR:
One new product from Siemens is the
syngo
Lung CAD. How is this different from the previous product,
syngo
LungCARE NEV (nodule enhanced viewing)?
AG:
The
syngo
LungCARE NEV is a first-generation product that it is em-bedded in
the clinical application
syngo
LungCARE CT (computed tomography), which offers a workflow that is
a first step toward our CAD product. The
syngo
Lung CAD (Figure 1) is a generation beyond NEV. It has gone through
clinical testing and validation through the PMA process, and we
have published clinical results and claims.
AR:
What is the minimum nodule size that
syngo
Lung CAD can detect?
AG:
The product has been validated in our published results for solid
nodules 3 mm. We chose the lower nodule size limit with the
intention that radiologists will institute their own cutoffs based
on clinical context.
AR:
What other CAD products does Siemens currently have on the
market?
AG:
The
syngo
Colonography PEV (polyp-enhanced viewing) is embedded in our
syngo
Colonography application. This is a complete colon examination
application with 2 added features: detection (with identification
and magnification of the polyps or polyplike structures) and
measurement of the structures.
We also have a product in echocardiography,
syngo
Auto Ejection Fraction (EF). It can be used to automatically detect
and track the heart walls in the very difficult scenario of cardiac
ultrasound. The system can quantify several measurements
automatically, including EF.
We also have a separate product called
syngo
TrueD, which is a complete application for evaluating therapy
response. We can look at longitudinal studies using
positron-emission tomography (PET)/CT from mulitple time points,
and we can combine them with complete support for registration,
mapping regions that correspond, and quantitative tabulations.
AR:
Which other CAD products are in development?
AG:
The
syngo
MammoCAD is under PMA review and is not yet available in the United
States. We have a product for pulmonary embolism detection or
filling defect detection that is also in review.
AR:
In what direction do you see CAD going in the future?
AG:
In oncology, we are targeting more specific diseases and organs,
but we are also looking at what I call " whole-body" CAD. CT,
magnetic resonance imaging, PET, and single-photon emission
computed tomography collect a lot of data, and getting through that
data is increasingly difficult, especially if many areas are to be
examined. So we are focusing on whole-body CAD with CAD tools for
navigation and disease detection.
AR:
Do you believe CAD will move beyond detection and toward disease
classification?
AG:
We are taking a very broad view of that. I would say it's moving
more toward diagnosis. In fact, we are not just looking at images.
We use images as one source of information. When you need to
characterize a lesion, we believe that you need more information.
You need to also know the clinical history of the patient, other
laboratory tests, blood data, and, potentially, genetic
information. Our concept is a little unique in that we look
holistically at all relevant data that go into diagnostic
determinations. Because we are talking about a decision aid, we can
essentially present a doctor with a certain probability of a
certain disease so that the patient management can be directed in a
certain direction. Our concept is broad, including all this other
data, and we are working on several areas in oncology and
cardiology to look at disease as disease, not just as a single
modality.
AR:
What are the major obstacles to the development of new CAD
products?
AG:
If you try to build products in this area, there are limitations on
the technology of how accurately one can carry out a task. There is
only a certain degree to which you can solve the problem. Given
that technical limitation, the primary driver for developing CAD is
the availability of high-quality ground-truth data. CAD systems are
specifically trained from evidence-based prior data. This data must
be very high quality and comprehensive, and it must provide enough
background information so that you can validate a CAD system
completely. Before you put a CAD system on the market, you must
know quite well that it will perform to its specifications in an
unknown clinical environment. That data is the fundamental heart of
building CAD systems, and you must have confidence to build the
right algorithms.
AR:
How hard is it to gain the acceptance of physician end-users?
AG:
When products truly help physicians sift through large amounts of
data, they must be integrated almost seamlessly into the workflow.
I cannot ask a physician to go to a second workstation to use CAD.
But if CAD is available in the right place in the PACS environment
where they do their reads and it is only 1 or 2 mouse clicks away,
it will overcome the acceptance hurdle.
AR:
Is reimbursement an issue?
AG:
With products that relate to a large quantity of data that try to
increase accuracy and efficiency, the systems have to be embedded
in the workflow in such a way that the value for the doctor or the
practice is that they are able to increase patient throughput. So
while the CAD systems may not increase reimbursement directly, the
argument should be that they should use CAD because it improves
overall efficiency.
AR:
What would you say is Siemens' view of the future of CAD?
AG:
The difference in the way we view CAD at Siemens is that we look
holistically at all data. That is really how we see the future,
that we have to look at diseases and all data that is relevant to
diagnosis.