2026-09-24T20:37:33Zhttps://dspace.mit.edu/server/oai/requestoai:dspace.mit.edu:1721.1/1395722022-01-15T03:00:53Zcom_1721.1_7582com_1721.1_7581col_1721.1_131022
Heldt, Thomas
Jaishankar, Rohan
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
2022-01-14T15:20:44Z
2022-01-14T15:20:44Z
2021-06
2021-06-23T19:37:50.511Z
https://hdl.handle.net/1721.1/139572
https://orcid.org/0000-0003-0114-1705
Intracranial pressure (ICP) is a cranial vital sign for monitoring patients with head injuries and to guide treatment decisions. Clinical ICP measurements are highly invasive and hence, ICP measurement is limited to critically ill patients. We present a spectral approach to model-based noninvasive ICP estimation, relying on a second-order circuit model of cerebrovascular physiology. We estimate ICP in the frequency domain, from arterial blood pressure and cerebral blood flow velocity waveforms. When validating our algorithm on two clinical patient cohorts of eight and a half hours, with measured ICP ranging from 1.3 mmHg to 24.8 mmHg, we achieved an accuracy and precision of 0.1 mmHg and 5.1 mmHg, respectively. Additionally, we designed an experimental porcine model to titrate the ICP in a predetermined manner over a wide range. This experimental model resulted in a rich dataset comprising 35 hours of data from eight pigs, with measured ICP ranging from 2.1 mmHg to 78.2 mmHg. We obtained an accuracy of 1.6 mmHg and a precision of 5.2 mmHg in estimating ICP on the porcine data. To evaluate our estimates' ability to correctly classify elevated ICP (defined as ICP>22 mmHg), we obtained an area under the receiver operating characteristic curve of 0.94. Additionally, the algorithm achieved a sensitivity of 0.88 and a specificity of 0.87 in this binary classification task at a noninvasive ICP threshold of 22 mmHg. Clinically, missing an episode of elevated ICP or under-treatment can have potentially fatal consequences, and we demonstrated that with appropriate margins on the classification thresholds, the probabilities of these events are less than 1\%, using our noninvasive ICP estimates. Finally, we obtained a correlation coefficient of 0.89 between our estimates and the measured ICP, indicating a high degree of capturing underlying variations in measured ICP. Our algorithm's performance is well within the clinically acceptable range and comparable or superior to past attempts at estimating ICP noninvasively reported in literature. We believe that the work presented here takes a significant step towards realizing the clinical dream of implementing a real-time, noninvasive ICP measurement modality in a calibration-free and patient-specific manner at the bedside.
Ph.D.
Massachusetts Institute of Technology
In Copyright - Educational Use Permitted
Copyright MIT
http://rightsstatements.org/page/InC-EDU/1.0/
A Spectral Approach to Noninvasive ICP Estimation: From Modeling to Clinical and Experimental Validation
Thesis
application/pdf
Doctoral
Doctor of Philosophy
Publication
unknown
unknown
<Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="fd6c0450-6159-4d0e-b8d2-7f54a9718ba7">
<Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843</Type>
<Title>A Spectral Approach to Noninvasive ICP Estimation: From Modeling to Clinical and Experimental Validation</Title>
<PublishedIn>
<Publication>
</Publication>
</PublishedIn>
<PublicationDate>2021-06</PublicationDate>
<Authors>
<Author>
<DisplayName>Jaishankar, Rohan</DisplayName>
<Affiliation>
<OrgUnit>
</OrgUnit>
</Affiliation>
</Author>
</Authors>
<Editors>
</Editors>
<Publishers>
<Publisher>
<DisplayName>Massachusetts Institute of Technology</DisplayName>
<OrgUnit />
</Publisher>
</Publishers>
<License>http://rightsstatements.org/page/InC-EDU/1.0/</License>
<Abstract>Intracranial pressure (ICP) is a cranial vital sign for monitoring patients with head injuries and to guide treatment decisions. Clinical ICP measurements are highly invasive and hence, ICP measurement is limited to critically ill patients. We present a spectral approach to model-based noninvasive ICP estimation, relying on a second-order circuit model of cerebrovascular physiology. We estimate ICP in the frequency domain, from arterial blood pressure and cerebral blood flow velocity waveforms. When validating our algorithm on two clinical patient cohorts of eight and a half hours, with measured ICP ranging from 1.3 mmHg to 24.8 mmHg, we achieved an accuracy and precision of 0.1 mmHg and 5.1 mmHg, respectively. Additionally, we designed an experimental porcine model to titrate the ICP in a predetermined manner over a wide range. This experimental model resulted in a rich dataset comprising 35 hours of data from eight pigs, with measured ICP ranging from 2.1 mmHg to 78.2 mmHg. We obtained an accuracy of 1.6 mmHg and a precision of 5.2 mmHg in estimating ICP on the porcine data. To evaluate our estimates' ability to correctly classify elevated ICP (defined as ICP>22 mmHg), we obtained an area under the receiver operating characteristic curve of 0.94. Additionally, the algorithm achieved a sensitivity of 0.88 and a specificity of 0.87 in this binary classification task at a noninvasive ICP threshold of 22 mmHg. Clinically, missing an episode of elevated ICP or under-treatment can have potentially fatal consequences, and we demonstrated that with appropriate margins on the classification thresholds, the probabilities of these events are less than 1\%, using our noninvasive ICP estimates. Finally, we obtained a correlation coefficient of 0.89 between our estimates and the measured ICP, indicating a high degree of capturing underlying variations in measured ICP. Our algorithm's performance is well within the clinically acceptable range and comparable or superior to past attempts at estimating ICP noninvasively reported in literature. We believe that the work presented here takes a significant step towards realizing the clinical dream of implementing a real-time, noninvasive ICP measurement modality in a calibration-free and patient-specific manner at the bedside.</Abstract>
<Access xmlns="http://purl.org/coar/access_right"
>
</Access>
</Publication>