Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control Analysis

Cerebral near-infrared spectroscopy (NIRS) systems have been demonstrated to continuously measure aspects of oxygen delivery and cerebrovascular reactivity. However, it remains unknown whether the prediction of these cerebral physiologic signals into the future is feasible. Leveraging existing archi...

Full description

Saved in:
Bibliographic Details
Main Authors: Amanjyot Singh Sainbhi, Logan Froese, Kevin Y. Stein, Nuray Vakitbilir, Rakibul Hasan, Alwyn Gomez, Tobias Bergmann, Noah Silvaggio, Mansoor Hayat, Jaewoong Moon, Frederick A. Zeiler
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:Bioengineering
Subjects:
Online Access:https://www.mdpi.com/2306-5354/12/7/682
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850077026950578176
author Amanjyot Singh Sainbhi
Logan Froese
Kevin Y. Stein
Nuray Vakitbilir
Rakibul Hasan
Alwyn Gomez
Tobias Bergmann
Noah Silvaggio
Mansoor Hayat
Jaewoong Moon
Frederick A. Zeiler
author_facet Amanjyot Singh Sainbhi
Logan Froese
Kevin Y. Stein
Nuray Vakitbilir
Rakibul Hasan
Alwyn Gomez
Tobias Bergmann
Noah Silvaggio
Mansoor Hayat
Jaewoong Moon
Frederick A. Zeiler
author_sort Amanjyot Singh Sainbhi
collection DOAJ
description Cerebral near-infrared spectroscopy (NIRS) systems have been demonstrated to continuously measure aspects of oxygen delivery and cerebrovascular reactivity. However, it remains unknown whether the prediction of these cerebral physiologic signals into the future is feasible. Leveraging existing archived data sources, four point and interval-forecasting methods using autoregressive integrative moving average (ARIMA) models were evaluated to assess their ability to predict NIRS cerebral physiologic signals. NIRS-based regional cerebral oxygen saturation (rSO<sub>2</sub>) and cerebral oximetry index signals were derived in three temporal resolutions (10 s, 1 min, and 5 min). Anchored- and sliding-window forecasting, with varying model memory, using point and interval approaches were used to forecast signals using fitted optimal ARIMA models. The absolute difference in the forecasted and measured data was evaluated with median absolute deviation, along with root mean squared error analysis. Further, Pearson correlation and Bland–Altman statistical analyses were performed. Data from 102 healthy controls, 27 spinal surgery patients, and 101 traumatic brain injury patients were retrospectively analyzed. All ARIMA-based point and interval prediction models demonstrated small residuals, while correlation and agreement varied based on model memory. The ARIMA-based sliding-window approach performed superior to the anchored approach due to data partitioning and model memory. ARIMA-based sliding-window forecasting using point and interval approaches can forecast rSO<sub>2</sub> and the cerebral oximetry index with reasonably small residuals across all populations. Correlation and agreement between the predicted versus actual values varies substantially based on data-partitioning methods and model memory. Further work is required to assess the ability to forecast high-frequency NIRS signals using ARIMA and ARIMA-variant models in healthy and cranial trauma populations.
format Article
id doaj-art-a7db96bf5e594b6d97bd14b57b3677b8
institution DOAJ
issn 2306-5354
language English
publishDate 2025-06-01
publisher MDPI AG
record_format Article
series Bioengineering
spelling doaj-art-a7db96bf5e594b6d97bd14b57b3677b82025-08-20T02:45:53ZengMDPI AGBioengineering2306-53542025-06-0112768210.3390/bioengineering12070682Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control AnalysisAmanjyot Singh Sainbhi0Logan Froese1Kevin Y. Stein2Nuray Vakitbilir3Rakibul Hasan4Alwyn Gomez5Tobias Bergmann6Noah Silvaggio7Mansoor Hayat8Jaewoong Moon9Frederick A. Zeiler10Department of Biomedical Engineering, Price Faculty of Engineering, University of Manitoba, Winnipeg, MB R3T 5V6, CanadaDepartment of Clinical Neurosciences, Karolinska Institutet, 171 77 Stockholm, SwedenDepartment of Biomedical Engineering, Price Faculty of Engineering, University of Manitoba, Winnipeg, MB R3T 5V6, CanadaDepartment of Biomedical Engineering, Price Faculty of Engineering, University of Manitoba, Winnipeg, MB R3T 5V6, CanadaDepartment of Biomedical Engineering, Price Faculty of Engineering, University of Manitoba, Winnipeg, MB R3T 5V6, CanadaSection of Neurosurgery, Department of Surgery, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB R3T 1R9, CanadaDepartment of Biomedical Engineering, Price Faculty of Engineering, University of Manitoba, Winnipeg, MB R3T 5V6, CanadaDepartment of Human Anatomy and Cell Science, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB R3T 0J9, CanadaSection of Neurosurgery, Department of Surgery, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB R3T 1R9, CanadaSection of Neurosurgery, Department of Surgery, Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB R3T 1R9, CanadaDepartment of Biomedical Engineering, Price Faculty of Engineering, University of Manitoba, Winnipeg, MB R3T 5V6, CanadaCerebral near-infrared spectroscopy (NIRS) systems have been demonstrated to continuously measure aspects of oxygen delivery and cerebrovascular reactivity. However, it remains unknown whether the prediction of these cerebral physiologic signals into the future is feasible. Leveraging existing archived data sources, four point and interval-forecasting methods using autoregressive integrative moving average (ARIMA) models were evaluated to assess their ability to predict NIRS cerebral physiologic signals. NIRS-based regional cerebral oxygen saturation (rSO<sub>2</sub>) and cerebral oximetry index signals were derived in three temporal resolutions (10 s, 1 min, and 5 min). Anchored- and sliding-window forecasting, with varying model memory, using point and interval approaches were used to forecast signals using fitted optimal ARIMA models. The absolute difference in the forecasted and measured data was evaluated with median absolute deviation, along with root mean squared error analysis. Further, Pearson correlation and Bland–Altman statistical analyses were performed. Data from 102 healthy controls, 27 spinal surgery patients, and 101 traumatic brain injury patients were retrospectively analyzed. All ARIMA-based point and interval prediction models demonstrated small residuals, while correlation and agreement varied based on model memory. The ARIMA-based sliding-window approach performed superior to the anchored approach due to data partitioning and model memory. ARIMA-based sliding-window forecasting using point and interval approaches can forecast rSO<sub>2</sub> and the cerebral oximetry index with reasonably small residuals across all populations. Correlation and agreement between the predicted versus actual values varies substantially based on data-partitioning methods and model memory. Further work is required to assess the ability to forecast high-frequency NIRS signals using ARIMA and ARIMA-variant models in healthy and cranial trauma populations.https://www.mdpi.com/2306-5354/12/7/682autoregressive integrative moving average structurecerebral oximetry indexinterval forecastnear-infrared spectroscopypoint forecastregional cerebral oxygen saturation
spellingShingle Amanjyot Singh Sainbhi
Logan Froese
Kevin Y. Stein
Nuray Vakitbilir
Rakibul Hasan
Alwyn Gomez
Tobias Bergmann
Noah Silvaggio
Mansoor Hayat
Jaewoong Moon
Frederick A. Zeiler
Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control Analysis
Bioengineering
autoregressive integrative moving average structure
cerebral oximetry index
interval forecast
near-infrared spectroscopy
point forecast
regional cerebral oxygen saturation
title Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control Analysis
title_full Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control Analysis
title_fullStr Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control Analysis
title_full_unstemmed Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control Analysis
title_short Time-Series Autoregressive Models for Point and Interval Forecasting of Raw and Derived Commercial Near-Infrared Spectroscopy Measures: An Exploratory Cranial Trauma and Healthy Control Analysis
title_sort time series autoregressive models for point and interval forecasting of raw and derived commercial near infrared spectroscopy measures an exploratory cranial trauma and healthy control analysis
topic autoregressive integrative moving average structure
cerebral oximetry index
interval forecast
near-infrared spectroscopy
point forecast
regional cerebral oxygen saturation
url https://www.mdpi.com/2306-5354/12/7/682
work_keys_str_mv AT amanjyotsinghsainbhi timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT loganfroese timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT kevinystein timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT nurayvakitbilir timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT rakibulhasan timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT alwyngomez timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT tobiasbergmann timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT noahsilvaggio timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT mansoorhayat timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT jaewoongmoon timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis
AT frederickazeiler timeseriesautoregressivemodelsforpointandintervalforecastingofrawandderivedcommercialnearinfraredspectroscopymeasuresanexploratorycranialtraumaandhealthycontrolanalysis