OpenHolter project: D.I.Y. Electrocardiography using Arduino and Django

Transcript

1. The OpenHolter Project: D.I.Y. Cardiometry with Arduino and Django. Roberto Rosario robertorosario.com
2. Who am I?
3. Who am I? Hint: Not him.
4. Who am I?
5. Who am I?
6. Who am I?
7. DISCLAIMER I’m not a doctor or a medical practitioner. This talk is about my personal experiences and covers my basic understanding about some topics using Internet sources. I share this as an introduction of these topics only. This talk is not meant to be a scientifically accurate document, do not use it or reference it as such. There many things in this talk that can be and should have been made differently/better. The device shown in this talk is not meant or claims to diagnose or cure heart diseases or conditions. I’m not endorsing that you build this device. The material in this talk doesn’t substitute a visit to your doctor or the years of study they endure to become doctors or the years of experience they endure to become experts. If you decide use the information shown in this talk or build the device shown in this talk, you do so at your own risk and assume any and all consequences of such decision, relieving me of any and all liabilities.
8. What is Cardiometry? Cardio = Heart Metry = Measurement
9. What is Electro Cardiometry? Electro = Electricity Cardio = Heart Metry = Measurement
10. What is an EKG (or ECG) Acronym: Electro Cardio Gram Electric Heart Diagram
11. Sample EKG. Wikipedia, Public Domain
12. Cardiac pacemaker. Wikipedia, CC BY-SA 2.5
13. Cardiac pacemaker. Wikipedia, CC BY-SA 3.0
14. ECG principle. Wikipedia, CC by 3.0
15. Sinus Rhythm. Wikipedia, Public domain
16. Sample EKG. Wikipedia, Public Domain
17. EKG leads. Wikipedia, Public Domain
18. Precordial leads. Wikipedia, Public Domain
19. Limb leads. Wikipedia, Public Domain
20. Limb leads. Wikipedia, Public Domain
21. Limb leads. Wikipedia, Public Domain
22. Sinus Rhythm. Wikipedia, Public domain
23. Sample EKG. Wikipedia, Public Domain
24. Vintage EKG machine. Wikipedia, Public Domain
25. Modern EKG machine. Wikipedia, CC by 3.0
26. What is a Holter? EKG + Time EKG Histogram
27. Holter. Wikipedia GNU Free Documentation License
28. Holter. Wikipedia GNU Free Documentation License
29. Holter. Wikipedia GNU Free Documentation License
30. Canine Holter. Wikipedia, Public Domain.
31. Holter. Wikipedia GNU Free Documentation License
32. Why I decided to do this?
33. Queue flashback!
41. Problems • An EKG is just a snapshot of your heart at a specific moment. • A moment when you might not have symptoms. • Monitoring your heart requires a Holter. • You need an appointment and referral for a Holter study. • Holters fail. You have to start all over again. • Holters only work for 24 hours.
42. Realisation • An EKG is ADC sampling. • Holter is an EKG with storage and battery. Or a data logger. • I had all the knowledge to build an Arduino based data logger. • I could work on it 1 hour a day during my recovery.
46. Challenges • Use off the shelves parts only. Orderwise known as “stuff-I-could-get-on-eBay” strategy. • Cheap. • Simple. • Use standard medical cables and electrodes. • Modular or breakout boards.
47. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
49. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
50. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
51. Bio Interface - Needs • No voltage but differential voltage reading • Signal amplification • Adaptive signal filtering • Analog to digital conversion
52. Bio Interface - Analog Devices AD8232 • $10-$20 USD • Built-in voltage comparators • Built-in amplifier • Built-in filtering • Power management • Cons: No ADC
53. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
54. Real Time Clock - Needs • Small • Battery or SuperCap backed up • Easy interface
55. Real Time Clock - Dallas DS3231 • $1 USD • I2C Interface • Unlike DS1302 and DS1307, has temperature compensated crystal and crystal oscillator • Has battery input and powerless timekeeping
56. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
57. Microcontroller unit - Needs • Low power • Analog to digital interface • Faster than bio interface sampling rate
58. Microcontroller unit - Arduino Nano • Clones at $4 USD • Atmel Atmega 328 at 16 MHz • Analog and digital pins • 32 KB Flash (minus 2KB bootloader) • I2C & SPI communications • UART TTL to USB
59. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
60. Power management - Needs • 5 volt output • High size and weight to power ratio • Low heat
61. Power management - Custom • GTL 18650 5300mAh Lithium Ion ($2 USD) • Lithium Ion charger module w/ output pins ($0.35 USD) • 3.7v to 5v voltage inverter ($0.50 USD)
62. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
63. Graphical User Interface - Needs • Small • Low power or low energy mode • Easy to interface
64. Graphical User Interface - Solomon Systech SD1306 • $5 USD • Multiple communication modes including I2C • OLED • Visibility • Pixel based • 2 x Push buttons
65. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
66. Storage - Needs • Fast • Detachable • Computer interface compatible
67. Storage - SDCard & module • $3 USD (card) + $1 USD (reader) • Cons: Fast modes need license. • SPI bus mode. This bus type supports only a 3.3-volt interface. This is the only bus type that does not require a host license. • 5V to 3V converter • 5V to 3V level converter
68. Communications SPI I2C or I2C
69. What is SPI The Serial Peripheral Interface bus (SPI) is a synchronous serial communication interface specification used for short distance communication, primarily in embedded systems. The interface was developed by Motorola in the late eighties and has become a de facto standard. Typical applications include Secure Digital cards and liquid crystal displays.
70. What is SPI SPI devices communicate in full duplex mode using a master-slave architecture with a single master.
71. What is I2C Short for Inter-IC, a type of bus designed by Philips Semiconductors in the early 1980s. Is used to connect integrated circuits (ICs).
72. What is I2C I2C is a multi-master bus, which means that multiple chips can be connected to the same bus and each one can act as a master by initiating a data transfer.
73. Block diagram MCU Bio Interface RTC Storage Graphical User Interface Power Management
74. Block diagram - Communications MCU Bio Interface RTC Storage Graphical User Interface Power Management Analog Digital I2C Digital I2C SPI
75. Organic interface - Connectors • Many types and subtypes ◦ Snap ◦ Banana ◦ Suction ◦ Clip ◦ Limb clips ◦ Alligator • Easiest to work and with the most pros: Snap
76. Organic interface - Connectors
77. Organic interface - Leads • Build your own ◦ Source connectors ◦ Good wire gauge • Included leads ◦ Poor in all regards • Lead + electrode ◦ Not feasible
78. Organic interface - Leads • Professional ones ◦ Best connectors ◦ Best wire gauge ◦ Tangle resistant ◦ Labeled ◦ $20 USD ◦ Easy to interface
79. Organic interface - Electrodes • BIO Protech T815 ◦ Cloth ◦ Extended wear (day) ◦ Sweat resistant ◦ Cons: Painful to remove • BIO Protech T716 ◦ Foam ◦ Cons: Not water resistant • $20 USD for bag of 50
80. Organic interface - Electrodes • BIO Protech T815 ◦ Cloth ◦ Extended wear (day) ◦ Sweat resistant ◦ Cons: Painful to remove • BIO Protech T716 ◦ Foam ◦ Cons: Not water resistant
81. Firmware • Modular • Object oriented • 3 modes of operation: ◦ Realtime display ◦ Recording ◦ Serial port output
82. Libraries - Display: U8g2 • Large hardware support • Support fonts • Graphics drawing • Low memory display mode (tiled) • Custom icons • GUI widgets like menus and dialogs • Physical buttons interface and debouncing
83. Libraries - Storage: Bill Greiman’ SDfatlib • Non monolithic • Multiple classes: SdFat, SdBaseFile, SdFile, File, StdioStream, ifstream, ofstream • Mature • Fast • Economic memory usage
84. Libraries - RTC: Adafruit’s RTClib • Supports several Arduino MCUs • Provides access to the square wave pin • Cons: High memory consumption
85. Challenges and Tricks • Sampling rate reliability • Sample jitter • Fast OLED display updates • Heartbeat detection • BPM calculation • Interference filtering • Memory footprint reduction
86. Challenges and Tricks • Sampling rate reliability • Sample jitter • Fast OLED display updates • Heartbeat detection • BPM calculation • Interference filtering • Memory footprint reduction
87. Sampling rate reliability • Arduino’s crystal is subject to drift • Interrupt driven • Improve periodicity of readings • Increase sample rate
88. Sampling rate reliability - Solution Use the DS3231 square wave output pin to trigger Arduino’s hardware interrupt and do sampling in an ISR.
89. Sampling rate reliability - Solution The enable_interrupt function turns on the square wave pin and enables the Arduino hardware interrupt pin to call a callback function on the falling edge of the wave.
90. Sampling rate reliability - Solution By default, a 4 KHz square wave is used, triggering the ISR 4,000 times a second or once every 0.000244141 seconds. This, we’ll call RTC ticks.
91. Sampling rate reliability - Solution To get 100 samples a seconds, how many ticks between each? 1 / 100 / 0.000244141 = every 40.959937086 ticks
92. Sampling rate reliability - Solution
93. Sampling rate reliability - Solution
94. Sampling rate reliability - Solution
95. Sampling rate reliability - Solution MCU Main loop MCU RTC driven ISR ad8232 ad8232_buffer ad8232 interrupt_data_ready RTC
96. Sampling rate reliability - Solution Consumer Hardware timer Generator Buffer (FIFO Stack) Lock Realtime sampling and multithreading on the Arduino.
97. Recordings format
98. Recordings format FAT only allows 8.3 filenames.
99. Recordings format Unique non sequential filenames?
100. Recordings format Hexadecimal unixtime!
101. Recordings format
102. Recordings format Decimal to time_struct, to epoch, to datetime
103. Recordings format
104. Recordings format Unique up to a second of difference.
105. Heartbeat detection A heartbeat is not a spike but a wave
106. Heartbeat detection
107. Heartbeat detection
108. Heartbeat detection
109. Heartbeat detection We are interested in #2
110. Heartbeat detection
111. Heartbeat detection
112. Heartbeat detection
113. Heartbeat detection
114. Heartbeat detection
115. Heartbeat detection
116. Heartbeat detection
117. Heartbeat detection
118. Heartbeat detection
119. The build
120. First prototype
121. First prototype
122. First prototype
123. First prototype
124. First prototype
125. It’s ALIVE!
127. It’s ALIVE!
128. First prototype • 3 modes of operation • Pocket size • Total cost ~ $30 USD • 4 GB SD Card (vs. 128 MB commercial • 12 days battery life (vs. 1 day commercial) • Rechargeable via USB • Upgradeable via USB
129. First prototype • Total flash usage: 29KB • Total RAM usage: 500 bytes
130. The App
131. Creating an EKG journal A chart is good. A journal is better.
132. Creating an EKG journal Recording downloader Packaged program Historic recordings Metadata Graphical Tags Share
134. Not just a DMS, but also a platform
135. Features of interest Mayan Apps Generic templates Metadata Graphical previews Tags Email sharing Access Control
136. App - Requirements • Upload recordings • Preview/render recordings • Selection and trim • Conversion to archival version • Metadata, tagging, etc. • Search • Sharing
137. App
138. App - Registration
139. App - Registration
140. App - Registration
141. App - Models
142. App - Models
143. App - Models
144. App - Model views
145. App - Model views
146. App - Model views
147. App - Model views
148. App - Model views
149. App - Model views
150. App - Widget, dygraphs
151. App - Widget, dygraphs
152. App - API views
153. App - Capture form
154. App - Capture form
155. App - Capture form
156. App - Capture form
157. App - Capture form
158. App - Capture form
159. App - Capture form
160. App - Recording slices list
161. App - Recording slices previews
162. App - Recording slices comments
163. App - Recording slices categorization
164. App - Recording slices categorization
165. App - Recording slices categorization
166. App - Recording slices searching
167. App - Recording slices searching
168. App - Recording slices searching
169. Demo
170. What’s next? Future enhancements, ideas, wishes.
171. Improvements, wishlist • Smaller • Single board • More leads • Panic button • Remove RTC -> Interrupt • Wireless • Battery level & low battery indicator • Audible “beeps” • Power save mode, smart power on & off
172. Improvements, wishlist
173. Improvements, wishlist Mayan Box (Mayan on a battery powered, wireless, single board computer) + OpenHolter = ??
174. Improvements, wishlist Solar powered, last mile, electrophysiologist office.
175. Improvements, wishlist Solar powered, on-the-field, emergency electrophysiologist station.
176. Improvements, wishlist • Mayan Box combo • SBC and Mayan inside
177. Next prototype
178. Questions? Critiques Comments Suggestions
179. Thank you!
180. The OpenHolter Project: D.I.Y. Cardiometry with Arduino and Django. Roberto Rosario robertorosario.com