Blood Oxygen Sensor - Revision history

Yintong: /* Results, Comparison and Improvement */

2025-04-29T07:00:47Z

Results, Comparison and Improvement

← Older revision		Revision as of 15:00, 29 April 2025
Line 239:		Line 239:
	'''Aggregate performance'''		'''Aggregate performance'''

	; Group A (n = 2, after cleaning)		''Group A (n = 2, after cleaning)''
	: Mean SpO₂ (Exp): 97.024 %　Reference: 98 %<br>		: Mean SpO₂ (Exp): 97.024 %　Reference: 98 %<br>
	: Mean deviation: −0.976 %<br>		: Mean deviation: −0.976 %<br>
	: Variance / Std-dev: 1.385 / 1.178		: Variance / Std-dev: 1.385 / 1.178

	; Group B (n = 2, after cleaning)		''Group B (n = 2, after cleaning)''
	: Mean SpO₂ (Exp): 98.043 %　Reference: 99 %<br>		: Mean SpO₂ (Exp): 98.043 %　Reference: 99 %<br>
	: Mean deviation: −0.957 %<br>		: Mean deviation: −0.957 %<br>

Yintong: /* Results, Comparison and Improvement */

2025-04-29T06:58:28Z

Results, Comparison and Improvement

← Older revision		Revision as of 14:58, 29 April 2025
Line 289:		Line 289:
	: Fuse an off-the-shelf accelerometer with the PPG waveform and apply adaptive noise cancellation (ANC) to filter motion-induced spikes.		: Fuse an off-the-shelf accelerometer with the PPG waveform and apply adaptive noise cancellation (ANC) to filter motion-induced spikes.

	~~'''~~Conclusion~~'''~~		=== Conclusion ===

	With minimal hardware cost (less than SGD 20 for core components), our prototype achieved a mean absolute error of '''0.97%''' compared with a certified pulse oximeter, demonstrating solid proof-of-concept performance.		With minimal hardware cost (less than SGD 20 for core components), our prototype achieved a mean absolute error of '''0.97%''' compared with a certified pulse oximeter, demonstrating solid proof-of-concept performance.

Yintong: /* Results, Comparison and Improvement */

2025-04-29T06:58:00Z

Results, Comparison and Improvement

← Older revision		Revision as of 14:58, 29 April 2025
Line 216:		Line 216:
	== Results, Comparison and Improvement ==		== Results, Comparison and Improvement ==

	~~'''~~Results~~'''~~		=== Results ===

	After processing six datasets (three from '''Group A''', three from '''Group B''') and removing one outlier from each group, the remaining four high-quality datasets were analysed.		After processing six datasets (three from '''Group A''', three from '''Group B''') and removing one outlier from each group, the remaining four high-quality datasets were analysed.
Line 237:		Line 237:
	''Outlier datasets'' (A-2 and B-3) were discarded because their variance exceeded 3 × the group median, indicating motion artefacts and baseline drift.		''Outlier datasets'' (A-2 and B-3) were discarded because their variance exceeded 3 × the group median, indicating motion artefacts and baseline drift.

	~~====~~ Aggregate performance ~~====~~		'''Aggregate performance'''

	; Group A (n = 2, after cleaning)		; Group A (n = 2, after cleaning)
Line 251:		Line 251:
	Across all retained datasets, the absolute error never exceeded '''1.6%''', and the pooled root-mean-square error (RMSE) was '''1.27%''', confirming that the prototype can reproduce commercial oximeter readings within clinically acceptable limits (±2% for SpO₂ > 90%).		Across all retained datasets, the absolute error never exceeded '''1.6%''', and the pooled root-mean-square error (RMSE) was '''1.27%''', confirming that the prototype can reproduce commercial oximeter readings within clinically acceptable limits (±2% for SpO₂ > 90%).

	~~'''~~Comparison~~'''~~		=== Comparison ===

	* Accuracy versus commercial device:		* Accuracy versus commercial device:
Line 265:		Line 265:
	: The two rejected trials showed >5% deviation and visibly distorted PPG waveforms (flattened peaks, baseline wander), highlighting the need for better motion suppression.		: The two rejected trials showed >5% deviation and visibly distorted PPG waveforms (flattened peaks, baseline wander), highlighting the need for better motion suppression.

	~~'''~~Improvement~~'''~~		=== Improvement ===

	* Mechanical stabilisation:		* Mechanical stabilisation:

Yintong: /* Results, Comparison and Improvement */

2025-04-29T06:57:12Z

Results, Comparison and Improvement

← Older revision		Revision as of 14:57, 29 April 2025
Line 237:		Line 237:
	''Outlier datasets'' (A-2 and B-3) were discarded because their variance exceeded 3 × the group median, indicating motion artefacts and baseline drift.		''Outlier datasets'' (A-2 and B-3) were discarded because their variance exceeded 3 × the group median, indicating motion artefacts and baseline drift.

	~~'''~~Aggregate performance~~'''~~		==== Aggregate performance ====

	; Group A (n = 2, after cleaning)		; Group A (n = 2, after cleaning)

Yintong: /* Improvement */

2025-04-29T06:45:28Z

Improvement

← Older revision		Revision as of 14:45, 29 April 2025
Line 266:		Line 266:

	'''Improvement'''		'''Improvement'''

	* Mechanical stabilisation:
	~~: Add a 3-D printed finger clip with an opaque shroud to maintain constant pressure and block ambient light.~~

	* Optical shielding and alignment:
	~~: Incorporate an IR-blocking gasket around the photodiode window and use collimating lenses to reduce scattering, boosting signal-to-noise ratio (SNR).~~

	* Analog front-end upgrades:
	~~: Replace the single-stage OPA177 with a two-stage transimpedance plus active low-pass filter (cut-off ≈ 30 Hz) to suppress 50 Hz mains hum.~~
	~~: Use shielded twisted-pair cables for LED drive and photodiode lines; add a driven-shield guard ring on the PCB.~~

	* Digital signal processing:
	~~: Implement adaptive band-pass filtering (0.5–5 Hz) and moving-average demodulation on a microcontroller (e.g., STM32) to enable real-time SpO₂ display.~~

	* Individual calibration:
	~~: Build a lookup table relating R-value to SpO₂ from 90%–100% using a series of controlled breath-hold exercises (ethics permitting).~~

	* Power and portability:
	~~: Replace the bench power supply with a 3.7 V Li-ion cell plus DC/DC converter (±5 V), and add Bluetooth Low-Energy (BLE) to stream data to a mobile app.~~

	* Motion artefact rejection:
	~~: Fuse an off-the-shelf accelerometer with the PPG waveform and apply adaptive noise cancellation (ANC) to filter motion-induced spikes.~~

	~~'''Conclusion'''~~

	With minimal hardware cost (less than SGD 20 for core components), our prototype achieved a mean absolute error of '''0.97%''' compared with a certified pulse oximeter, demonstrating solid proof-of-concept performance.
	By implementing the improvements above—especially mechanical stabilisation and on-board DSP—we expect to cut variance by half and unlock continuous, real-time SpO₂ monitoring suitable for everyday health tracking.

	~~== '''Improvement''' ==~~

	* Mechanical stabilisation:		* Mechanical stabilisation:

Yintong: /* Comparison */

2025-04-29T06:45:14Z

Comparison

@@ Line 293: / Line 293: @@
 With minimal hardware cost (less than SGD 20 for core components), our prototype achieved a mean absolute error of '''0.97%''' compared with a certified pulse oximeter, demonstrating solid proof-of-concept performance.
 By implementing the improvements above—especially mechanical stabilisation and on-board DSP—we expect to cut variance by half and unlock continuous, real-time SpO₂ monitoring suitable for everyday health tracking.
 == '''Improvement''' ==

Yintong: /* Results, Comparison and Improvement */

2025-04-29T06:44:52Z

Results, Comparison and Improvement

@@ Line 224: / Line 224: @@
 ! Group !! Trial !! R-value !! Calculated SpO₂ (%) !! Reference SpO₂ (%) !! Δ (Exp − Ref, %)
 |-
-| A || 1 || 0.562 || 97.67 || 98 || −0.33
+| A || 1 || 0.562 || 97.669 || 98 || −0.331
 |-
-| A || 3 || 0.458 || 99.29 || 98 || +1.29
+| A || 3 || 0.458 || 99.292 || 98 || +1.292
 |-
-| B || 1 || 0.509 || 98.62 || 99 || −0.38
+| B || 1 || 0.509 || 98.617 || 99 || −0.383
 |-
-| B || 2 || 0.572 || 97.47 || 99 || −1.53
+| B || 2 || 0.572 || 97.469 || 99 || −1.531
 |}
@@ Line 250: / Line 250: @@
 Across all retained datasets, the absolute error never exceeded '''1.6%''', and the pooled root-mean-square error (RMSE) was '''1.27%''', confirming that the prototype can reproduce commercial oximeter readings within clinically acceptable limits (±2% for SpO₂ > 90%).
 == '''Comparison''' ==

Yintong: /* Results, Comparison and Improvement */

2025-04-29T06:43:03Z

Results, Comparison and Improvement

← Older revision		Revision as of 14:43, 29 April 2025
Line 214:		Line 214:


	=== Results, Comparison and Improvement ===		== Results, Comparison and Improvement ==

	'''Results'''		'''Results'''
Line 251:		Line 251:
	Across all retained datasets, the absolute error never exceeded '''1.6%''', and the pooled root-mean-square error (RMSE) was '''1.27%''', confirming that the prototype can reproduce commercial oximeter readings within clinically acceptable limits (±2% for SpO₂ > 90%).		Across all retained datasets, the absolute error never exceeded '''1.6%''', and the pooled root-mean-square error (RMSE) was '''1.27%''', confirming that the prototype can reproduce commercial oximeter readings within clinically acceptable limits (±2% for SpO₂ > 90%).

	'''Comparison'''		== '''Comparison''' ==

	* ~~'''~~Accuracy versus commercial device:~~'''~~		* Accuracy versus commercial device:
	: The overall bias (prototype – reference) is −0.97%, which is comparable to the ±2% accuracy range quoted for low-cost fingertip sensors.		: The overall bias (prototype – reference) is −0.97%, which is comparable to the ±2% accuracy range quoted for low-cost fingertip sensors.

	* ~~'''~~Inter-group consistency~~'''~~		* Inter-group consistency:
	: Group A exhibited lower variance (σ ≈ 1.18%) than Group B (σ ≈ 2.03%), suggesting that finger positioning and contact pressure were more stable for Group A.		: Group A exhibited lower variance (σ ≈ 1.18%) than Group B (σ ≈ 2.03%), suggesting that finger positioning and contact pressure were more stable for Group A.

	* ~~'''~~Trial-to-trial repeatability~~'''~~		* Trial-to-trial repeatability:
	: Coefficient of variation (CV) across the four retained trials is 1.3%, indicating good repeatability after basic data cleaning.		: Coefficient of variation (CV) across the four retained trials is 1.3%, indicating good repeatability after basic data cleaning.

	* ~~'''~~Outliers~~'''~~		* Outliers:
	: The two rejected trials showed >5% deviation and visibly distorted PPG waveforms (flattened peaks, baseline wander), highlighting the need for better motion suppression.		: The two rejected trials showed >5% deviation and visibly distorted PPG waveforms (flattened peaks, baseline wander), highlighting the need for better motion suppression.

	'''Improvement'''		== '''Improvement''' ==

	* ~~'''~~Mechanical stabilisation~~'''~~		* Mechanical stabilisation:
	: Add a 3-D printed finger clip with an opaque shroud to maintain constant pressure and block ambient light.		: Add a 3-D printed finger clip with an opaque shroud to maintain constant pressure and block ambient light.

	* ~~'''~~Optical shielding and alignment~~'''~~		* Optical shielding and alignment:
	: Incorporate an IR-blocking gasket around the photodiode window and use collimating lenses to reduce scattering, boosting signal-to-noise ratio (SNR).		: Incorporate an IR-blocking gasket around the photodiode window and use collimating lenses to reduce scattering, boosting signal-to-noise ratio (SNR).

	* ~~'''~~Analog front-end upgrades~~'''~~:		* Analog front-end upgrades:
	** Replace the single-stage OPA177 with a two-stage transimpedance plus active low-pass filter (cut-off ≈ 30 Hz) to suppress 50 Hz mains hum.		: Replace the single-stage OPA177 with a two-stage transimpedance plus active low-pass filter (cut-off ≈ 30 Hz) to suppress 50 Hz mains hum.
	** Use shielded twisted-pair cables for LED drive and photodiode lines; add a driven-shield guard ring on the PCB.		: Use shielded twisted-pair cables for LED drive and photodiode lines; add a driven-shield guard ring on the PCB.

	* ~~'''~~Digital signal processing~~'''~~		* Digital signal processing:
	: Implement adaptive band-pass filtering (0.5–5 Hz) and moving-average demodulation on a microcontroller (e.g., STM32) to enable real-time SpO₂ display.		: Implement adaptive band-pass filtering (0.5–5 Hz) and moving-average demodulation on a microcontroller (e.g., STM32) to enable real-time SpO₂ display.

	* ~~'''~~Individual calibration~~'''~~		* Individual calibration:
	: Build a lookup table relating R-value to SpO₂ from 90%–100% using a series of controlled breath-hold exercises (ethics permitting).		: Build a lookup table relating R-value to SpO₂ from 90%–100% using a series of controlled breath-hold exercises (ethics permitting).

	* ~~'''~~Power and portability~~'''~~		* Power and portability:
	: Replace the bench power supply with a 3.7 V Li-ion cell plus DC/DC converter (±5 V), and add Bluetooth Low-Energy (BLE) to stream data to a mobile app.		: Replace the bench power supply with a 3.7 V Li-ion cell plus DC/DC converter (±5 V), and add Bluetooth Low-Energy (BLE) to stream data to a mobile app.

	* ~~'''~~Motion artefact rejection~~'''~~		* Motion artefact rejection:
	: Fuse an off-the-shelf accelerometer with the PPG waveform and apply adaptive noise cancellation (ANC) to filter motion-induced spikes.		: Fuse an off-the-shelf accelerometer with the PPG waveform and apply adaptive noise cancellation (ANC) to filter motion-induced spikes.

Yintong: /* Results, Comparison and Improvement */

2025-04-29T06:39:52Z

Results, Comparison and Improvement

← Older revision		Revision as of 14:39, 29 April 2025
Line 277:		Line 277:
	** Use shielded twisted-pair cables for LED drive and photodiode lines; add a driven-shield guard ring on the PCB.		** Use shielded twisted-pair cables for LED drive and photodiode lines; add a driven-shield guard ring on the PCB.

	* '''Digital signal processing''':		* '''Digital signal processing'''
	Implement adaptive band-pass filtering (0.5–5 Hz) and moving-average demodulation on a microcontroller (e.g., STM32) to enable real-time SpO₂ display.		: Implement adaptive band-pass filtering (0.5–5 Hz) and moving-average demodulation on a microcontroller (e.g., STM32) to enable real-time SpO₂ display.

	* '''Individual calibration''':		* '''Individual calibration'''
	Build a lookup table relating R-value to SpO₂ from 90%–100% using a series of controlled breath-hold exercises (ethics permitting).		: Build a lookup table relating R-value to SpO₂ from 90%–100% using a series of controlled breath-hold exercises (ethics permitting).

	* '''Power and portability''':		* '''Power and portability'''
	Replace the bench power supply with a 3.7 V Li-ion cell plus DC/DC converter (±5 V), and add Bluetooth Low-Energy (BLE) to stream data to a mobile app.		: Replace the bench power supply with a 3.7 V Li-ion cell plus DC/DC converter (±5 V), and add Bluetooth Low-Energy (BLE) to stream data to a mobile app.

	* '''Motion artefact rejection''':		* '''Motion artefact rejection'''
	Fuse an off-the-shelf accelerometer with the PPG waveform and apply adaptive noise cancellation (ANC) to filter motion-induced spikes.		: Fuse an off-the-shelf accelerometer with the PPG waveform and apply adaptive noise cancellation (ANC) to filter motion-induced spikes.

	'''Conclusion'''		'''Conclusion'''

Yintong: /* Results, Comparison and Improvement */

2025-04-29T06:39:08Z

Results, Comparison and Improvement

← Older revision		Revision as of 14:39, 29 April 2025
Line 256:		Line 256:
	: The overall bias (prototype – reference) is −0.97%, which is comparable to the ±2% accuracy range quoted for low-cost fingertip sensors.		: The overall bias (prototype – reference) is −0.97%, which is comparable to the ±2% accuracy range quoted for low-cost fingertip sensors.

	* '''Inter-group consistency''':		* '''Inter-group consistency'''
	Group A exhibited lower variance (σ ≈ 1.18%) than Group B (σ ≈ 2.03%), suggesting that finger positioning and contact pressure were more stable for Group A.		: Group A exhibited lower variance (σ ≈ 1.18%) than Group B (σ ≈ 2.03%), suggesting that finger positioning and contact pressure were more stable for Group A.

	* '''Trial-to-trial repeatability''':		* '''Trial-to-trial repeatability'''
	Coefficient of variation (CV) across the four retained trials is 1.3%, indicating good repeatability after basic data cleaning.		: Coefficient of variation (CV) across the four retained trials is 1.3%, indicating good repeatability after basic data cleaning.

	* '''Outliers''':		* '''Outliers'''
	The two rejected trials showed >5% deviation and visibly distorted PPG waveforms (flattened peaks, baseline wander), highlighting the need for better motion suppression.		: The two rejected trials showed >5% deviation and visibly distorted PPG waveforms (flattened peaks, baseline wander), highlighting the need for better motion suppression.

	'''Improvement'''		'''Improvement'''

	* '''Mechanical stabilisation''':		* '''Mechanical stabilisation'''
	Add a 3-D printed finger clip with an opaque shroud to maintain constant pressure and block ambient light.		: Add a 3-D printed finger clip with an opaque shroud to maintain constant pressure and block ambient light.

	* '''Optical shielding and alignment''':		* '''Optical shielding and alignment'''
	Incorporate an IR-blocking gasket around the photodiode window and use collimating lenses to reduce scattering, boosting signal-to-noise ratio (SNR).		: Incorporate an IR-blocking gasket around the photodiode window and use collimating lenses to reduce scattering, boosting signal-to-noise ratio (SNR).

	* '''Analog front-end upgrades''':		* '''Analog front-end upgrades''':