Upload
preston-james
View
216
Download
2
Tags:
Embed Size (px)
Citation preview
e-DIOLA Scan2009
Outline e-DIOLA Scan• Motivation:
– SON– Statistical Analysis
• A-Scan• Design Approach• Features • Block Diagram:
– Higher level Block Diagram– Lower level Block Diagram
• Algorithm & Simulation• Devices selection• Hybrid Design Model:
– Hardware– Firmware– Software– Interfacing
• System Validation– Simulation-based Testing– Audio Signal-based Testing
• Limitations• Demo• Future Work
• WHY• WHAT• HOW• System Validation• Limitations• Demo• Future Work
Statement of need• Recent Studies show a remarkable prevalence of ophthalmic
diseases.
• Cataracts leading cause of blindness in the world
• Cataract surgery involves substitution of natural lens by an artificial (IOL)
• Lens power is a function of eye dimensions and curvature values.
• Accuracy level is crucial
• In order to motivate the local market & adapt to changes in the global industry, we will follow a progressive approach to reach our prospective goal:
Enhance A-scan biometry systems
A-Scan Biometry Systems
Physicians Opinion
A-Scan Probe
1/10/2008 Graduation Project - Milestone I 4
Ophthalmic A-scan Uses US to measure the length (dimension) of the eye. Ophthalmologist uses these measurements to evaluate eyes prior to cataract surgery. provides a one-dimensional image of deflections of varying amplitude along a baseline
Height : intensity of the returning impulse Position: indicates time of receiving the echo
IOL : implanted lens – replaces existing crystalline lens since it is clouded over by a cataractDimensions:
Axial Length (AL)Anterior Chamber Depth (ACD)Length Thickness (LT)Vitreous Cavity Depth (VCD)
Techniques: Applanation/Contact ↔ Immersion
Design Approach
We propose an enhanced relatively-low-cost A-scan Biometry system which connects via USB cable to any Windows XP/Windows Vista laptop or desktop computer
Featuresadded
Automatic Diagnosis
best scan selection
optimization of A-Scan Gain & TGC
Adjustable Layer-specific velocity
misalignment & corneal compression check
Manual & automatic selection of IOL formula
Manual labeling
Adjustments for dense cataract & trauma
Manual TGC adjustment
Accurate measurements of AL, LT, VD,&VCD
IOL power calculation
Lens correction factorContinuous Display Freeze
Thresholding all IOL formulae
Automatic & Manual modes
Fixed and TGC
Ribbon Display
Compare IOL formula
3 modes
Avg. & Std
Accurate Peak detection
Block diagrams
• High level BD• Low level BDs
– Acquisition– Preprocessing– TGC
• How we calculated it (each interface & attenuation)• Function
Higher-Level Block Diagram
Lower-Level Block Diagram
Signal Acquisition Module:
Input: Signal from probeOutput: Digitized signal
Pre-processing Module:
• Input: Digitized Signal• Output: Amplified signal
Region Cornea Aqueous Lens Vitreous Retina-Sclera
Attenuation at 9MHz 0.234 0.091 1.82 0.091 0.364
Max Distance (mm) 0.8901098 4.6153846 5.9109893 22.12088 2.03077
Attenuation (dB) 0.208285 0.419999 10.75800 2.013000 0.739200
Classification Module
• Input: Amplified Signal
• Output: Thresholded spikesTime
Voltage
Threshold
Mode Selection
1. Auto-detection & Signal Validation • Input: Spikes from eye Interfaces• Output: identified peaks, while: • Valid Spikes are sent to Eye Measurement Module.• Abnormal Spikes are sent to Auto-diagnosis.• Invalid Spikes are sent to the Reset to rescan.
Mode Selection
B. Manual Mode & Gates:• Input: Spikes from eye Interfaces• Output: identified peaks
Eye Calculations Module
• Input: Time & Amplitude Arrays from Manual or Automatic modes
• Output: Eye Measurements (AL, VCD, LT, VCD)
IOL Measurement Module
• Input: IOL Formula from Manual or Automatic modes
• Output: Calculated IOL Power
Display Module
• Patient Name• Date of Test• Eye Mode (phakic, Aphakic, pseudophakic)• Axial Length• ACD• VCD• Identified Interface-Peaks
Auto-diagnosis Module
• Input: Abnormal Spikes from Automatic Mode• Output:
Non-Axial Scanning
• Input: Spikes from Manual Mode • Output: Tumor Location
θ = cos-1 And the radius at which the tumor is located by:
R = L sin (θ)
Algorithm and Simulation
Algorithm and Simulation0 0.5 1 1.5 2 2.5
x 10-5
-50
0
50
Am
plit
ude
Returning Echo
0 0.5 1 1.5 2 2.5
x 10-5
-500
0
500
Am
plit
ude
Echo after fixed gain
0 0.5 1 1.5 2 2.5
x 10-5
-50
0
50
Am
plit
ude
Echo after TGC
0 0.5 1 1.5 2 2.5
x 10-5
-500
0
500
Am
plit
ude
Echo after fixed gain & TGC
0 0.5 1 1.5 2 2.5
x 10-5
0
200
400
Am
plit
ude
Enveloping
0 0.5 1 1.5 2 2.5
x 10-5
0
2000
4000
Am
plit
ude
No. of Samples
Extracted Spikes
Device Selection• MCU
ASIC DBB µP and µC DSP chipChip count 1 >1 1 1Flexibility None limited programmable programmableDesign time long mdeium short shortPower consumption low Medium-high medium Medium-highProcessing speed High high Low-medium Mediu-highReliability high Low-medium high HighDevelopment cost high medium low LowProduction cost low high Low-medium Low-mdeium
Criteria Required ProvidedEfficient data flow Real time data transmission YESComputational power Medium HighResolution Medium (8bits) High (up to 12 bits)Processing speed High HighMemory size Medium MediumSignal type Mixed signal (Analog and digital) Mixed signalAvailability of a full set of development tools and supports
C compiler, linker, logic analyzer, development assistance, real time debugging tool
All available with the kit
Device Selection• Java
– Java is simple– Java is object-oriented– Java is platform-independent– Java is distributed– Java is interpreted– Java is secure– Java is robust– Java is multithreaded
Devices Selection
• Supporting Tools– Keil + SiliconLabs– MatLaB– NetBeans– Orcad– Proteus– Edge
Data Collection
Device SelectionHybrid comparison table
Standalone system Microcontroller - PC based Software
Advantages1-No compatibility issue 1-Friendly UI 1- Friendly UI2-Independent on electrical problems 2-Independent processing 2-Faster processing3- No interfacing problems 3-Extended data storage 3-Extended data storage
4-More acquired knowledge(e.g.: power supply)
4-More compact 4-More features
5-More acceptable 5-More Features by software 5-More compact (probe + CD)
6-Needs little training 6-Portable(battery) 6-Inexpensive7-Inexpensive
Disadvantages
1-Difficult to implement 1-Dependent on PC or laptop 1-Specific requirements of processor
2-Limited data storage 2-Restricted by complex multitasking
2-Higher risk of computer freezing or limitation
3-Not friendly software 3-Compatibility issues 3-Compatibility issues4-Limited processing 4-Dependent on laptop power
supply4-Upgrading
5-Integrated printer (so more complex) 5-Specific requirements of laptop or PC
5-Dependent on laptop power supply
6-Viruses’ problem(no windows network )
Votes 28.3% 51.7% 20%
Hybrid model diagram
Hardware
Probe
Digital Oscilloscope
Pulser
T/R SwitchVcn Vcp Pulser BPF Probe Simulation
Case 1 high low operating 0 Vpulse from the Pulser circuit
Case 2 low highZero input
signal
signal from the
probe
receives signal from tissue
Enveloping
Returning Echo
Time Envelope-Detector Response to a realistic A-Scan Pulse
Time
Voltage
Bandpass Filter (anti-aliasing)
• Band-lited signal undersampling• Bessel Filter:
– Sharper cutoff frequencies.– Removing ripples in the pass band.– Eliminating overshooting and ringing problems
Badpass Filter
Frequency (MHz)
OutputOutput Amplitude
(V)
25 2 V
4.38 MHz 0.5V
Power Supply
Required Obtained Required Obtained Required Obtained
25 V 15 V
3.3 V
- 15 V 5 V
Overall schematic from ORCAD
• Hyperlinks to each part’s proteus simulation• Bandpass:
– Bandlimited signal not nyquist– Why Bessel– Illustration image – Simulation only 2 pic (4. MHz and 25 MHz)
• Power Supply output voltage– 25,15,-15,5,3.2 V
Firmware Flowchart
Firmware performance analysis
• Blue bars picture• Last 2 rows of performance analysis table• Memory distribution part
Firmware performance analysis
Firmware performance analysisCode Segment Total # of
Instructions# of Calls msec Percentage per Scan
Time
Initializations 64 1 0.087 0.2%
Trigger 8 1 -- ~0%
ADC Interrupt Service Routine 51 260 0.916 4.0%
UART Interrupt Service Routine 128 -- -- --
UART Transmission 755 1 0.416 1.8%
Amplification (FIXED) 78 1 2.311 10.1%
Amplification (TGC) 385 1 12.570 54.7%
Peak Detection 1037 1 3.784 16.4%
Thresholding 369 1 0. 267 1.2%
Check for normality 2602 1 0. 129 1.2%
Main 591 1 0. 170 1.5%
Mathematical Calculations 307 -- -- --
Dynamic Allocation 320 70 1.451 6.7%
Total Time/ scan 6694 -- 26.70 100%
Total Time/ 10 scans -- -- 251.76 --
Firmware performance analysis
Memory Distribution
• • Code Space: 14138 Bytes• RAM space• Internal: 65 Bytes• External: 3509 Bytes
Java Flow chart
Interfacing High level protocol• Java-to-MCU packet image
0xAA
0x00
Ltype DC M/A Type
0xAA
0x01
a
b
c
d
0xAA
0x02
RST STP F/C
Initial Data TGC Control Word
Interfacing High level protocol• MCU-to-Java packet images not kalam
HEADER
# of SPIKES
STATUS
DATA
…
…
…
0xCC
STATUS
Data packet Status Packet
Validation
Acquisition ADC MCU Laptop
Simulation-basedSample-based
Overall Testing
Sample of initial testing of FIrmware
Normal
Sample of initial testing of FIrmware
Aphakic no lens
Sample of initial testing of FIrmware
Mild Asteroid
Sample of initial testing of software by simultion
Sample of initial testing of software by simultion
Coat’s disease
Sample of initial testing of software by simultion
Testing by audio signal
0 50 100 150 200 2500
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Am
plitu
de
No. of Samples
Analog Audio Signal - Normal Scan
Demo goes here =)
LimitationsLimitation Reason
unable to acquire a real ultrasound echo signal
to hardware deficiencies
unable to detect overlapping envelopes that are not separated by at least 1 sample (0.29usec)
limited resolution
perform the designed corneal compression test
limited resolution
the angle algorithm was not implemented as Java code and stayed as pseudo code due to.
time restriction
automatic labeling: the system is unable to transfer to the manual labeling mode in case of more spikes in the retina region.
for phantom materials shortage of materials
Wireless probe design and difficulty to purchase high level electronic components small enough to be on a probe PCB
Future work -Wireless
Future work -Phantom
• Interpretation Eye Interface Velocity (m/s) Density (kg/m3) Acoustic impedance
(MRayl)Attenuation Coefficient
at 9 MHZ (dB/mm)Equivalent Material
Cornea 1620 1062+5 34 0.234 Donated human cornea
Aqueous 1500 1040 1.56 0.091 Saltwater
Lens 1620 or 1641 1000-1050 1.67125 1.82 Lucite
Cow’s eye: 1140+33
Vitreous 1520 or 1532 1005.3 1.534 0.091 Water
Retina 1540 0.364 biopolymer
hydrogels
Sclera 1613 -1622 1000 1.6175 0.7826 ---
Future work -Phantom
• Problem Definition• System Background• Function of Parameters• Model
Project CostItem Cost (L.E)
A-Scan Probe 10,000
Microcontroller Kit 233
PCB 660
USB Cable 12
Notebook 1,500
FTDI chip (USB-Serial IC) 30
Components 70
Total 12,505
Timeplan – Gantt Chart
Thank you