Transcript DIET Imaging

DIET – An emerging noninvasive, portable and lowcost approach to breast
cancer screening
Prof J. Geoffrey Chase
Univ of Canterbury
Dept of Mechanical Engineering, Centre for Bio-Engineering
Christchurch, New Zealand
Overview
1.
Background
2.
DIET Technology
3.
Clinical results
4.
Summary
The Problem
 Breast cancer was the most common cause of female cancer
death in 1999
 Over the period 1972 to 1997, the annual number of breast
cancer deaths increased from 427 to 643[1]
 Breast cancer is over represented among Maori (in NZ) and
other ethnic groups worldwide
[1] NZ Ministry of Health, 2002
Breast Cancer Screening
Reduces Mortality (Tabar et al, 2003)
• Screening = More early detection
+20%
• No real difference
• Results due to improving care
Why?
Low Compliance & Access
What you don’t see can kill you! [Breast Screen Aotearoa]
Goal is 70% every 2 years, but we get ~60% -- 40% missing!
• Predominant compliance rates in the US and EU range from 50-80% based on many factors
• Eligible populations (over 50 years) are growing demographically for next 10-20 years
• Certain sub-groups have very low screening rates and thus much higher mortality
Source:
NZ
Ministry favor
of Health,
Trends and
• Occurrence rates don’t
seem to
particularly
any group
Projections 2002.
Screening

Screening has resulted is fewer deaths and earlier treatment

Concerns of screening
 Patient dose
 When should women start screening
 Costs
 Resources
 Impacts of false positives (in a gold-standard test that can only be
done infrequently at older ages)

However, fundamentally, mammography is a scarce and
increasingly costly resource …
 One thus cannot simply go get a repeat test to confirm, especially if
you are younger
Current Diagnostic Techniques

Palpation/Clinical Breast Examination (CBE)



Mammography




Mammography
High level of operator skills
Significant (sensor) noise
MRI




only modality recommended for screening
X-rays, breast compression (painful)
Lower compliance due to pain, cost and access
Ultrasound



Very subjective
Primary form of diagnosis  Most mammograms occur after being
found this way first
Very accurate, but expensive
Very few can be served
No screening programs in most developing nations!
Cost and X-Ray dose are major factors
MRI
Problems with Existing Techniques

Currently, predominant breast cancer screening methods are:
 Uncomfortable
 Subject the patient to doses of radiation
 Require expensive, location specific equipment and clinical staff.
 They thus have relatively limited throughput (not enough capacity)

They are also low contrast as cancerous tissue density varies only ~5-10%
from healthy tissue

Coupled with resulting low compliance rates the average tumour size detected
is ~1cm = 10x larger than possible
What’s Needed?
 An all new approach
 Must be clinically and commercially feasible
 Must address compliance (w/ screening) issues
 Must offer high throughput in terms of speed to test and access
 An ideal design list would include:





Low cost equipment with no need for specialist technician
Portable
No X-Ray dose
Equal efficacy (1cm detection) compared to mammography
Greater comfort (no compression)
 Succinctly: less invasive, low cost screening system with
more objective diagnosis (less ‘human error’)
The DIET Concept

DIET = Digital Imaging-based Elasto-Tomography

Can we meet all these needs?
Governors Bay, Christchurch
Sunset over Southern Alps, Christchurch
The DIET System Concept
Advantages of the DIET Concept

Screening from a younger age (no radiation dose)

Possible to build a history (every year!)

Less painful alternative (equals higher compliance)

Accuracy (initial target 1cm)

Portability and ease of use (no specialised technician and no loss of
compliance due to travel)

Scalability (will improve as silicon technology used improves)

Should be low cost (low-cost technologies used)
Prototype Development
Laboratory Hardware
1st Clinical prototype
DIET device at Canterbury BC
So, does it work??

If we could measure surface motions could we detect cancerous
lesions, from surface data only?
Lake Mathieson, Mirror Lakes
West Coast of S. Island
Lindis Pass and into Wanaka
Cental Otago, S. Island
So, how does it work? And does it?
 3 main steps:
 Vibration and image capture
 Surface motion: (a) tracking and (b) characterisation
 Diagnose based on surface motion
 Proof of concept clinical trials as part of ergonomic and comfort
trials shown
 Multiple volunteers, multiple frequencies of actuation
 Test of technology and proof of clinical concept
Step 1: Vibration & Image Capture

Vibrate breast from underneath and capture digital image sequences
from 5 cameras
Silicone gel with fiducial markers
Simulation of human breast
Step 2a: Surface Motion Tracking

Markers tracked in 2D, then 3D correspondences
Camera 1

Camera 2
Full 3D reconstruction
Optical flow skin tracking (better resolution)
Fiducial tracking
Skin tracking
Step 2b: Motion characterisation
Z-amplitude
Z-phase

Motion data is
parameterised to allow
diagnostic processing

Elliptical path characterises
motion in 3D (at each
frequency)

Motion flows around a hard
tumor like water around a
rock changing amplitude
and phase asymmetrically
Step 3: Diagnosis from tumour effect
on motion



Silicone phantom with 20mm tumour at 6 o’clock
Motion images alone can potentially give yes/no answer about
tumour inclusion
Readily automated ... Entirely objective
Re Amplitude
z-phase
In vivo results (preliminary)
 Breast from ongoing trial
 30mm tumour between 1-2 o’clock, left outer upper quadrant
 Same effects as in phantoms are observed
TOP
RIGHT
LEFT
BOTTOM
Rez-amplitude
Z-phase
Some more visual outcomes
 Both breasts of each subject shown at one frequency
30mm @ 1:30 o’clock
20 + 8mm @ 2:30 o’clock
11mm @ 10:30 o’clock
 120mm not shown, as it is almost entire breast
DIET human trial summary

Ergonomics/Calibration trial

18 Subjects, age 49.3 (SD 8.5)

36 breasts imaged (both on each subject)

4 breasts with malignant tumors (11, 30, 20+8, 120mm)

2 breasts with benign cysts (clusters of 10-21mm, )

30 healthy breasts

Range of sizes and shapes see below.
First Uses:
A 3 word case
 Under-age
 Under-served
 Under-equipped
Plenty of “room” for a new modality like this
A Brief Summary

DIET is an all new approach to breast cancer screening that offers several
potential advantages over current methods

Initial simulation and experimental proof of concept studies showed that it
might be possible to achieve realistic screening (~1cm inclusion size
detection)

The main imaging and reconstruction steps are technologically challenging

Initial proof of concept experiments on silicone phantoms have been
successful in identifying inclusions both via reconstruction and from
disturbances in surface motion
Acknowledgements
Prof Geoff Chase Dr Geoff Shaw
Arnaud Milsant
Jerome Rouze
Richard Brown
Dr Thomas Desaive
Ashton Peters
Wili Berger
Dr. Richard Wien & Dr. Larry Ray
Christina Starfinger
Rodney Elliot
Crispen Berg
Ben Petit
Michael
Wiertlewski
DIET Project Team 2004
DIET Project Team 2005
Fabrice Jandet
Shig Kinoshita
Edouard Ravni
Anthony Hii
Dr. Chris Hann
Stefan Wortmann
Questions?