final presentation

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

1

Intracranial Hematoma Detection

G.K.I. Abayarathna : 050001A

Gartheeban Ganeshapillai : 050131V

E.D.R. Kumara : 050234N

W.M.D. Soysa : 050440R

Supervised By: Dr. A. A. Pasqual

Project Presentation

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

3

What – Whom - How

Intracranial Hematoma (IH) is a highly fatal, albeit treatable, secondary injury.

Traumatic Brain Injury (TBI), the most prevailing cause of fatality in an accident, is one of the prime causes of IH.

Post-surgery complications is another prominent cause for IH, which requires continuous monitoring.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

4

Incidence

In USA, on average 2 out of 1000 peoplesuffer by IH.

In Sri Lanka, on average 10 patients undergo surgery for IH, daily at National Hospital, Colombo alone, most of them transferred from peripheries only after showing external symptoms.

There are only 4 teaching hospitals with the facilities to treat IH, in Sri Lanka.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

5

Solution!

Affordable,

Portable,

Safety abiding, device to assist in

on-site detection,

continuous monitoring, and

prioritizing the CT scan

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

7

Objective

Indicating the existence of extra-cerebral hematomas

Deliverables

Indication

Visual imaging

Comparison and analysis

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

8

Problem

Multiple claims address the problem with disparate technologies, and we set

out to find the most viable!

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

9

Tasks

1. Research to understand the problem and background

2. Experiment to find viable methods for detection, ascertainment and visualization

3. Development of a prototype – proof of concept

4. Design of semi-automatic algorithms

5. Enhance the accuracy, performance and implement machine-learning on the availability of adequate data

6. Field test

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

11

Intracranial Hematoma (IH)Intracranial bleeding occurs when a blood

vessel within the skull is ruptured or leaks.

Leads to increases in intracranial pressure.

Severe increases in intracranial pressure can cause potentially deadly brain.

Early identification, prior to neurological deterioration, is the key to successful surgical treatment.

Continuous monitoring and serial scans are used

to detect delayed hematomas.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

12

Types of IH

Extra-cerebral.

Epidural hematoma

Subdural hematoma

Subarachnoid hematoma

Intra-cerebral.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

13

Between 750 – 850 nm wavelength

Water has a low absorption

Oxyhemoglobin (HbO2) and deoxyhemoglobin (Hb) have detectable absorption.

Greater concentration of hemoglobin in an acute hematoma

Extra vascular blood absorbs NIR light more than normal brain tissue due to greater hemoglobin concentration.

Near Infra-red (NIR)

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

14

Bidirectional Associative Memory (BAM)

A recurrent neural network.

Used to establish hetero-association.

Popular for recognizing predetermined patterns such as characters.

Adopted for post detection integration –to identify and filter out improbable detections.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

15

Others

Optical Density

o▲OD = log (I1/I2)

o▲ OD= log Io/Il - log Io/Ih

= log Il/Ih

Threshold detection

Pre-detection integration

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

17

iHD Terminal Sensor & Data

Acquisition Subsystem

Processing & Controlling Subsystem

Input Output Subsystem

Power Management Subsystem

Communication Subsystem

Firmware

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

18

iHD Mobile Application

Data collection and Management

Normalization

OD Calculation

Threshold detection

Post detection integration

Visualization

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

20

Continuous scan

Continuous scan over injured side and healthy side.

Approximately,200 x 400 samples in 7 secondsover 16cm.

Automatically stopped after 10 seconds

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

21

Localized scan

On four predefined locations on the head.

A : Frontal area

B : Temporal area

C : Parietal area

D : Occipital area

First on the left side, followed by the right side.

400 samples per scan.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

22

Sampling

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

23

Transfer

Over any media supporting serial profile.

Eg: RS232, Bluetooth.

Layered Communication architecture.

Communication protocol specification.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

25

Data management

Additional data such as Record ID, MO ID, etc are stored.

Data is stored in XML.

Maximum compatibility and extensibility.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

26

Analysis

Normalization -used to account for unequal number of samples.

Threshold based semi-automatic detection

BAM based filtering

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

27

Design - algorithms

Direct visual comparison with reference scans

Optical Density (OD) calculations

Adopting radar system principles

Instead of few localize scans, continuous scan over whole length

For each subject scan calculate corresponding reference value from reference scans by averaging

Variable threshold to detect a hematoma for a given probability of error.

Post detection integration to eliminate improbable cases

Threshold detection to identify hematomas semi-automatically

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

28

Visualization

Intensity based color maps.

Semi-automatic detection and visualization.

Approximate estimation of the location of hematoma.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

30

Safety Following FDA regulations

Heating Cell damage is probable above 41˚C.

Cerebrum temperature is 35˚C, which gives 6˚C buffer.

NIR of 50 mW/sq cm contributes to 0.5˚C (sunlight 50mW/sq cm).

LED Output Light output of an LED is 30% of the input

850 nm LED uses 50 mA continuously. Around 40mW per LED.

Use of PWM 50 mW/sq cm applied for 0.5 s will be able to penetrate more than

25 mW/sq cm applied for the full 1 second.

However it generates the same amount of heat.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

31

Best practices followed …

Modular design in hardwareIsolated sensor, LED, communication and I/O subsystems

Rapid prototyping for enclosureElegant, user-friendly design

Design of communication protocol and abstractionTo structure the communication

Use of .Net 3.5 platformFor broader availability, usability, and code reuse

Use of OOPCode reuse, organization, and maintainability

Hosted at http://fyp.theebgar.net

Project repository, Knowledge base, Public Gateway

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

32

Results

700 mV through healthy side on reception, with the noise floor at 50 mV, maximum of 1.2 OD can be identified.

Hair absorption is high, needs higher power transmission

5 minutes for complete scan:Each scan takes 10 seconds, and data-transfer is instantaneous

Analysis in iHDMA is almost instantaneous taking less than 2 seconds for calculations on 400 samples.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

33

Clinical Trials To assess the efficacy of the iHD

To obtain the statistical data of the area and concentration of hemoglobin

Subjects : patients undergoing surgery due to IH.

Scheduled on 23rd May, at National Hospital, accident ward

Testing duration : 5 minutes

Procedure: Scanning before undergoing the surgery and a follow up trial after the surgery is conducted with a duration of another 5 minutes, when possible. 2 areal scans done in each side of the head of the patient, and another 8 single scans on specific locations.

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

34

Financial Statement

Prototype cost

Item Cost

Bluetooth module 7000

Microcontroller 500

Sensor and LED 2000

Components 300

PCB 300

Dual cell 1500

Enclosure 16000

Miscellaneous 1000

28600

Total R&D cost

Item Cost

Components 5000

Enclosure 25000

Imported components 20000

Optical filters 55000

Miscellaneous 5000

110000

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

36

Problems faced

Power and safety

FDA regulations

Hair absorption

Ambient noise

Compactness of the device

Detection and false alarm

Field testing

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

37

Future directions

2D solution : Using an array of LEDs and sensors

Compact design : Using surface mount components and aluminum body

Modulated signal and matched filter : To improve accuracy

Quick results : Performing basic calculations in the microcontroller

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

38

Glimpses

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

39

DISCUSSION

Department of Electronic and Telecommunication Engineering, University of Moratuwa

Final Year Project Presentation – Y05 May 22, 2009

40

THANK YOU