Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm Infants
Objective: To test the potential utility of applying machine learning methods to regional cerebral (rcSO<sub>2</sub>) and peripheral oxygen saturation (SpO<sub>2</sub>) signals to detect brain injury in extremely preterm infants. Study design: A subset of infants enrolled in...
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Children |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-9067/10/6/917 |
| _version_ | 1827737988443930624 |
|---|---|
| author | Minoo Ashoori John M. O’Toole Ken D. O’Halloran Gunnar Naulaers Liesbeth Thewissen Jan Miletin Po-Yin Cheung Afif EL-Khuffash David Van Laere Zbyněk Straňák Eugene M. Dempsey Fiona B. McDonald |
| author_facet | Minoo Ashoori John M. O’Toole Ken D. O’Halloran Gunnar Naulaers Liesbeth Thewissen Jan Miletin Po-Yin Cheung Afif EL-Khuffash David Van Laere Zbyněk Straňák Eugene M. Dempsey Fiona B. McDonald |
| author_sort | Minoo Ashoori |
| collection | DOAJ |
| description | Objective: To test the potential utility of applying machine learning methods to regional cerebral (rcSO<sub>2</sub>) and peripheral oxygen saturation (SpO<sub>2</sub>) signals to detect brain injury in extremely preterm infants. Study design: A subset of infants enrolled in the Management of Hypotension in Preterm infants (HIP) trial were analysed (<i>n</i> = 46). All eligible infants were <28 weeks’ gestational age and had continuous rcSO<sub>2</sub> measurements performed over the first 72 h and cranial ultrasounds performed during the first week after birth. SpO<sub>2</sub> data were available for 32 infants. The rcSO<sub>2</sub> and SpO<sub>2</sub> signals were preprocessed, and prolonged relative desaturations (PRDs; data-driven desaturation in the 2-to-15-min range) were extracted. Numerous quantitative features were extracted from the biosignals before and after the exclusion of the PRDs within the signals. PRDs were also evaluated as a stand-alone feature. A machine learning model was used to detect brain injury (intraventricular haemorrhage-IVH grade II–IV) using a leave-one-out cross-validation approach. Results: The area under the receiver operating characteristic curve (AUC) for the PRD rcSO<sub>2</sub> was 0.846 (95% CI: 0.720–0.948), outperforming the rcSO<sub>2</sub> threshold approach (AUC 0.593 95% CI 0.399–0.775). Neither the clinical model nor any of the SpO<sub>2</sub> models were significantly associated with brain injury. Conclusion: There was a significant association between the data-driven definition of PRDs in rcSO<sub>2</sub> and brain injury. Automated analysis of PRDs of the cerebral NIRS signal in extremely preterm infants may aid in better prediction of IVH compared with a threshold-based approach. Further investigation of the definition of the extracted PRDs and an understanding of the physiology underlying these events are required. |
| first_indexed | 2024-03-11T02:37:51Z |
| format | Article |
| id | doaj.art-6050083c75d0479684c27e8c6ee825d8 |
| institution | Directory Open Access Journal |
| issn | 2227-9067 |
| language | English |
| last_indexed | 2024-03-11T02:37:51Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Children |
| spelling | doaj.art-6050083c75d0479684c27e8c6ee825d82023-11-18T09:49:02ZengMDPI AGChildren2227-90672023-05-0110691710.3390/children10060917Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm InfantsMinoo Ashoori0John M. O’Toole1Ken D. O’Halloran2Gunnar Naulaers3Liesbeth Thewissen4Jan Miletin5Po-Yin Cheung6Afif EL-Khuffash7David Van Laere8Zbyněk Straňák9Eugene M. Dempsey10Fiona B. McDonald11INFANT Research Centre, University College Cork, T12 AK54 Cork, IrelandINFANT Research Centre, University College Cork, T12 AK54 Cork, IrelandINFANT Research Centre, University College Cork, T12 AK54 Cork, IrelandDepartment of Development and Regeneration, Katholieke Universiteit Leuven, Herestraat 49, 3000 Leuven, BelgiumNeonatal Intensive Care, Katholieke Universiteit Hospital Leuven, Herestraat 49, 3000 Leuven, BelgiumPaediatric and Newborn Medicine, Coombe Women’s Hospital, D08 XW7X Dublin, IrelandDepartment of Paediatrics, University of Alberta, Edmonton, AB T6G 1C9, CanadaFaculty of Medicine and Health Sciences, Royal College of Surgeons in Ireland, D02 P796 Dublin, IrelandNeonatale Intensive Care Unit, Universitair Ziekenhuis, (UZ) Antwerp, Drie Eikenstraat 655, 2650 Antwerp, BelgiumInstitute for the Care of Mother and Child, Third Faculty of Medicine, Charles University, 100 00 Prague, Czech RepublicINFANT Research Centre, University College Cork, T12 AK54 Cork, IrelandINFANT Research Centre, University College Cork, T12 AK54 Cork, IrelandObjective: To test the potential utility of applying machine learning methods to regional cerebral (rcSO<sub>2</sub>) and peripheral oxygen saturation (SpO<sub>2</sub>) signals to detect brain injury in extremely preterm infants. Study design: A subset of infants enrolled in the Management of Hypotension in Preterm infants (HIP) trial were analysed (<i>n</i> = 46). All eligible infants were <28 weeks’ gestational age and had continuous rcSO<sub>2</sub> measurements performed over the first 72 h and cranial ultrasounds performed during the first week after birth. SpO<sub>2</sub> data were available for 32 infants. The rcSO<sub>2</sub> and SpO<sub>2</sub> signals were preprocessed, and prolonged relative desaturations (PRDs; data-driven desaturation in the 2-to-15-min range) were extracted. Numerous quantitative features were extracted from the biosignals before and after the exclusion of the PRDs within the signals. PRDs were also evaluated as a stand-alone feature. A machine learning model was used to detect brain injury (intraventricular haemorrhage-IVH grade II–IV) using a leave-one-out cross-validation approach. Results: The area under the receiver operating characteristic curve (AUC) for the PRD rcSO<sub>2</sub> was 0.846 (95% CI: 0.720–0.948), outperforming the rcSO<sub>2</sub> threshold approach (AUC 0.593 95% CI 0.399–0.775). Neither the clinical model nor any of the SpO<sub>2</sub> models were significantly associated with brain injury. Conclusion: There was a significant association between the data-driven definition of PRDs in rcSO<sub>2</sub> and brain injury. Automated analysis of PRDs of the cerebral NIRS signal in extremely preterm infants may aid in better prediction of IVH compared with a threshold-based approach. Further investigation of the definition of the extracted PRDs and an understanding of the physiology underlying these events are required.https://www.mdpi.com/2227-9067/10/6/917near-infrared spectroscopy (NIRS)regional cerebral oxygen saturation (rcSO<sub>2</sub>)peripheral oxygen saturation (SpO<sub>2</sub>)prolonged relative desaturation (PRD)extreme gradient boosting (XGBoost) |
| spellingShingle | Minoo Ashoori John M. O’Toole Ken D. O’Halloran Gunnar Naulaers Liesbeth Thewissen Jan Miletin Po-Yin Cheung Afif EL-Khuffash David Van Laere Zbyněk Straňák Eugene M. Dempsey Fiona B. McDonald Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm Infants Children near-infrared spectroscopy (NIRS) regional cerebral oxygen saturation (rcSO<sub>2</sub>) peripheral oxygen saturation (SpO<sub>2</sub>) prolonged relative desaturation (PRD) extreme gradient boosting (XGBoost) |
| title | Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm Infants |
| title_full | Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm Infants |
| title_fullStr | Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm Infants |
| title_full_unstemmed | Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm Infants |
| title_short | Machine Learning Detects Intraventricular Haemorrhage in Extremely Preterm Infants |
| title_sort | machine learning detects intraventricular haemorrhage in extremely preterm infants |
| topic | near-infrared spectroscopy (NIRS) regional cerebral oxygen saturation (rcSO<sub>2</sub>) peripheral oxygen saturation (SpO<sub>2</sub>) prolonged relative desaturation (PRD) extreme gradient boosting (XGBoost) |
| url | https://www.mdpi.com/2227-9067/10/6/917 |
| work_keys_str_mv | AT minooashoori machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT johnmotoole machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT kendohalloran machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT gunnarnaulaers machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT liesbeththewissen machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT janmiletin machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT poyincheung machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT afifelkhuffash machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT davidvanlaere machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT zbynekstranak machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT eugenemdempsey machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants AT fionabmcdonald machinelearningdetectsintraventricularhaemorrhageinextremelypreterminfants |