2. Identification of
the exact make
and model of an
orthopaedic
implant prior to
a revision
surgery can be
challenging.
Current
identification
procedure is
manual and
time
consuming.
Further time
lapse in
contacting that
particular
implant
manufacturer
to confirm the
make and
model of the
implant.
Leads to delay
in treatment
thus requiring
extra hospital
bed occupancy.
DEFINITION
3. We propose developing an automated implant
recognition mobile application based on computer vision
and image pattern recognition technology to identify the
make and model of an implant by analysing the
radiographs and matching the same with the database
provided by the vendors.
AIMS AND OBJECTIVES
4. The correct identification of orthopaedic devices
is an important element of pre-operative
planning which should facilitate implant extraction
during revision arthroplasty.
IDENTIFICATION
5. CLOUD APPLICATION
The application will be based on cloud (publically
available software as a service) services hence allowing
the orthopaedic surgeons to accurately identify an
implant and also make sure that parts are still currently
available as occasionally implants may have been
withdrawn from the market.
6. The patient outcome can be affected.
Wrong size or improper match can lead to soft-tissue irritation,
impingement, pain, decreased motion, poor function, and/or runaway wear.
A second operation may be required.
IMPORTANTS of IDENTIFICATION
7. Surgeon may not able to recognize the
implant as he has never used the particular
type of TKR.
REASONS FOR DELAY
IN IDENTIFICATION
Operating notes of the primary surgery are
not available to determine the model and
manufacturer of the implant.
8. PROPOSED METHOD
Identification and segmentation of the implant on the digital X-Ray
Estimation of multiple view geometric model that is used for data
association
Improve X-ray visual quality by using image processing methods
Search and match the Implant image in the database
Recognition Model
Step 1
Step 2
9. The System would be able to specifically extract a specific geometry from the X-ray based on the
difference in the color density
IMAGE EXTRACTION
12. IMAGE RECOGNITION
API’s specify how the software components of the proposed application interact with
each other.
One or a combination of API’s that help to recognize the implant accurately from a
large database of radiographs will be used to design a fully functional application in the
initial phase.
Then a specialized and advanced API will be developed in the implant identification
application.
Analyse image recognition techniques to understand the underlying technologies
based on an interface , Application Programming interface (API).
13. STEP 6
Testing and
prototyping
STEP 1
Research all
existing APIs
that currently
perform image
recognition
STEP 2
Check the
relevance and
prioritize the
APIs through
trial and error
to understand
potential of
each
technology
STEP 3
Design and
develop a
custom API
that will aid in
Implant
Recognition
STEP 4
Launch a free
service
STEP 5
Build
manufacturer
imparted
database
STEPS INVOLVED
14. Each manufacturer brands their models with unique
grooves, tapers and design constraints to create
differentiation.
RECOGNITION FEATURES
Ratio of Length of femoral peg to the tibial peg.
Length and width of femoral component.
Specific Metallic landmarks to anchor point of the
polyethylene.
Angle of tibial base plate.
Ratio of width of tibial base plate to length of tibial peg.
15. IMAGE RECOGNITION TOOLS
Our study takes into
account the existent
technologies such as
Facebook,
Pictoria, Imagga,
Google images.
There is an API
currently available that
can be directly applied
to build an implant
recognition system.
Facebook’s image tagging
algorithms to store unique
information with each image
is the starting point to help
build an intelligent system
that in combination with
image processing and
development of a custom
implant recognition API.
16. artificial intelligence software capable
Until now, computer vision has largely been limited to recognizing
individual objects.
Recognizing and describing the content of photographs and videos
with far greater accuracy than ever before
Even mimicking human levels of understanding.
17. Facial recognition (or face recognition) is a type of biometric software application that can
identify a specific individual in a digital image by analyzing and comparing patterns.
Facial Recognition
Ans: A biometric is a unique, measurable characteristic of a human being that can
be used to automatically recognize an individual or verify an individual identity.
Biometrics can measure both physiological and behavioral characteristics.
1. What are biometrics?
18. CAPTURING OF IMAGE
BY STANDARD VIDEO CAMERAS
The image is optical in characteristics and may be thought of as a collection of a large
number of bright and dark areas representing the picture details
. In other words the picture information is a function of two variables: Time and Space.
It would require infinite number of channels to transmit optical information
corresponding to picture elements simultaneously. There is practical difficulty in
transmitting all information simultaneously so we use a method called scanning.
19. The 3 main components of face recognition systems, they are as follows
COMPONENTS OF FACE RECOGNITION SYSTEMS
Enrollment module
Database
Identification module
20. Each face has distinguishable nodal
points that represent the peaks
and valleys termed as features of a
face.
The system measures these 80
nodal points
Ratios Measured
Distance between the eyes
Width of the nose
Depth of the eye sockets
The shape of the cheekbones
The length of the jaw line
The FACE ID is
mapped with
individual and higher
IDENTITY is created
This is stored as
an UNIQUE FACE
ID by application
Points are transferred
to database as an
algorithm of numbers
Application reads
geometry of FACE
(as plotted in grid above)
1 2 3
FACEBOOK FACIAL RECOGNITION
21. RGB
HSV(hue, saturation, and value)
Binary
YCbCr
CIE XYZ
CMYK
Color Model
HSV and YCbCr color models are used for Maximum Accuracy of the Face and
feature Extraction. I have Used HSV to detect Face from Image
22. GOOGLE IMAGES
Uploading an Implant radiograph on Google images gave the following results
Google’s Neural recognition methods based on Mathematical models can
be used for object recognition.
Google Images API cannot be directly used due to privacy
issues.
Manufacturer’s name who had previously uploaded the same image on
the internet.
Similar radiographs that looked 70-80% similar
Search results did not reflect any implant photographs. Although,
they are available over the Google Image server.
23. Prototype development cost: 120 hrs @ average of approx. £15/hr
Researchers’ salary : 6 months @ average of approx. £ 1000/month
Application development cost
Revision Cost
Server Maintenance
Senior developers’ salary: 3 months @ average of approx. £ 2000/month
Total
Detailed budget items Amount (£)
1800
6000
10000
3000
1500
6000
45400
Cost statement
24. Revision surgeries
Revision surgeries may need to be carried out by hospitals
without access to index surgery operating notes.
Patients may approach the most convenient not necessarily
the same surgeon for a revision surgery.
25. CONCLUSION
There is an urgent need to have a robust and accurate system
for identification of orthopaedic implants.
The dependency upon surgeon’s experience, hospital
facilities and archiving of records can be avoided with the
use of a single application that allows multiple
manufacturers to contribute to a database of catalogue of
their products.