Advisory Committee Chair
Arie Nakhmani
Advisory Committee Members
Abd Elmonieum Naser
Earl Wells
Karthik Lingasubramanian
Moustafa Hussein Aly
Document Type
Dissertation
Date of Award
2018
Degree Name by School
Doctor of Philosophy (PhD) School of Engineering
Abstract
In the United States, many patients pass away due to delayed detection of Cerebral Vasospasm (CV). Transcranial Doppler (TCD) is an efficient noninvasive, reliable device that can monitor the blood flow in various parts of the human brain like middle cerebral artery and basilar artery without any surgical intervention. It can help neurologists to diagnose many brain problems like edema, trauma, hemorrhage, and aneurysm. Unfortunately, continuous or even daily TCD monitoring is challenging due to the operator expertise and certification required in the form of a trained technologist, and certified physician to perform these studies. This barrier exists due to lack of automation for detection (without human intervention) of CV using TCD. To overcome this barrier, in this dissertation we present a timely, efficient novel algorithm that automates early detection of CV. We used the machine learning tools for CV detection, with the chosen time-frequency extracted features. Then we applied principal component analysis to reduce the data dimensionality. All the chosen features were used as input for training a decision-tree classifier. After that the CV detection model accuracy was evaluated with regard to noise robustness in real-time. The algorithm generated moderate accuracy. So, we enhanced the CV detection accuracy by adding continuous wavelet transform (CWT). The experimental results show 92.5% sensitivity for CV, and 95% specify for normal signals. Subsequently, it is very important to establish a model for diagnosing and analyzing the abnormal cerebral blood flow velocity associated with CV. By generating such a model, it would be possible to design machine learning mechanisms for earlier prediction of CV. In the previous studies, cerebrovascular models of a blood flow behavior for different disorders were established. Unfortunately, those models are disease-specific and have too many parameters to tune. We have established a model for the signal envelope of the cerebral blood flow velocity that was produced by a transcranial Doppler. We have applied it to simulate CV behavior, but it is general enough to be applied to other cerebrovascular disorders. The model is based on the delay differential equations as a representative modeling equation for three cases: control (no CV or hyperemia), hyperemia, or CV. The model has only 4 tunable parameters and allows switching from one case to another by changing those parameters. After the validation of the model, the generated envelope signals were compared to recorded by transcranial Doppler spectrograms and demonstrated good match between the model and real signals in all three cases. This result could be used for modeling cerebral blood velocity abnormalities, early detection of CV, and potentially other disorders. Finally, A syringe pump control system has been designed to supply an automatic process for flow control of medication. We have presented a device that injects certain amount of cc of drug into the patient in a prescribed time to handle the CV.
Recommended Citation
Elzaafarany, Khaled, "Cerebral Vasospasm Detection And Treatment Using Transcranial Doppler Signal Analysis And Adaptive Medication Pump Mechanism" (2018). All ETDs from UAB. 1597.
https://digitalcommons.library.uab.edu/etd-collection/1597
