Prashant Hariharan

Bio: Prashant Hariharan is an academic researcher from Wayne State University. The author has contributed to research in topics: Shunt (electrical) & Hydrocephalus. The author has an hindex of 2, co-authored 4 publications receiving 6 citations.

Machine Learning Applications in the Neuro ICU: A Solution to Big Data Mayhem?

Abstract: The neurological ICU (neuro ICU) often suffers from significant limitations due to scarce resource availability for their neurocritical care patients. Neuro ICU patients require frequent neurological evaluations, continuous monitoring of various physiological parameters, frequent imaging, and routine lab testing. This amasses large amounts of data specific to each patient. Neuro ICU teams are often overburdened by the resulting complexity of data for each patient. Machine Learning algorithms (ML), are uniquely capable of interpreting high-dimensional datasets that are too difficult for humans to comprehend. Therefore, the application of ML in the neuro ICU could alleviate the burden of analyzing big datasets for each patient. This review serves to (1) briefly summarize ML and compare the different types of MLs, (2) review recent ML applications to improve neuro ICU management and (3) describe the future implications of ML to neuro ICU management.

A multicenter retrospective study of heterogeneous tissue aggregates obstructing ventricular catheters explanted from patients with hydrocephalus.

Abstract: Implantation of ventricular catheters (VCs) to drain cerebrospinal fluid (CSF) is a standard approach to treat hydrocephalus. VCs fail frequently due to tissue obstructing the lumen via the drainage holes. Mechanisms driving obstruction are poorly understood. This study aimed to characterize the histological features of VC obstructions and identify links to clinical factors. 343 VCs with relevant clinical data were collected from five centers. Each hole on the VCs was classified by degree of tissue obstruction after macroscopic analysis. A subgroup of 54 samples was analyzed using immunofluorescent labelling, histology and immunohistochemistry. 61.5% of the 343 VCs analyzed had tissue aggregates occluding at least one hole (n = 211) however the vast majority of the holes (70%) showed no tissue aggregates. Mean age at which patients with occluded VCs had their first surgeries (3.25 yrs) was lower than in patients with non-occluded VCs (5.29 yrs, p < 0.02). Mean length of time of implantation of occluded VCs, 33.22 months was greater than for non-occluded VCs, 23.8 months (p = 0.02). Patients with myelomeningocele had a greater probability of having an occluded VC (p = 0.0426). VCs with occlusions had greater numbers of macrophages and astrocytes in comparison to non-occluded VCs (p < 0.01). Microglia comprised only 2–6% of the VC-obstructing tissue aggregates. Histologic analysis showed choroid plexus occlusion in 24%, vascularized glial tissue occlusion in 24%, prevalent lymphocytic inflammation in 29%, and foreign body giant cell reactions in 5% and no ependyma. Our data show that age of the first surgery and length of time a VC is implanted are factors that influence the degree of VC obstruction. The tissue aggregates obstructing VCs are composed predominantly of astrocytes and macrophages; microglia have a relatively small presence.

Characterization of a multicenter pediatric-hydrocephalus shunt biobank

Abstract: Pediatric hydrocephalus is a devastating and costly disease. The mainstay of treatment is still surgical shunting of cerebrospinal fluid (CSF). These shunts fail at a high rate and impose a significant burden on patients, their families and society. The relationship between clinical decision making and shunt failure is poorly understood and multifaceted, but catheter occlusion remains the most frequent cause of shunt complications. In order to investigate factors that affect shunt failure, we have established the Wayne State University (WSU) shunt biobank. To date, four hospital centers have contributed various components of failed shunts and CSF from patients diagnosed with hydrocephalus before adulthood. The hardware samples are transported in paraformaldehyde and transferred to phosphate-buffered saline with sodium azide upon deposit into the biobank. Once in the bank, they are then available for study. Informed consent is obtained by the local center before corresponding clinical data are entered into a REDCap database. Data such as hydrocephalus etiology and details of shunt revision history. All data are entered under a coded identifier. 293 shunt samples were collected from 228 pediatric patients starting from May 2015 to September 2019. We saw a significant difference in the number of revisions per patient between centers (Kruskal–Wallis H test, p value < 0.001). The leading etiology at all centers was post-hemorrhagic hydrocephalus, a fisher’s exact test showed there to be statistically significant differences in etiology between center (p = 0.01). Regression showed age (p < 0.01), race (p = 0.038) and hospital-center (p < 0.001) to explain significant variance in the number of revisions. Our model accounted for 31.9% of the variance in revisions. Generalized linear modeling showed hydrocephalus etiology (p < 0.001), age (p < 0.001), weight and physician (p < 0.001) to impact the number of ventricular obstructions. The retrospective analysis identified that differences exist between currently enrolled centers, although further work is needed before clinically actionable recommendations can be made. Moreover, the variables collected from this chart review explain a meaningful amount of variance in the number of revision surgeries. Future work will expand on the contribution of different site-specific and patient-specific factors to identify potential cause and effect relationships.

Shunts and Shunt Malfunction

Abstract: Perhaps one of the most intriguing events in neurosurgical history is the invention and use of the shunt system to treat hydrocephalus, despite the shunt’s high failure rates. The shunt system, classically composed of a proximal catheter, valve, and distal catheter, was developed in the 1950s. The system was envisioned by John Holter, a toolmaker, whose son, Casey, had hydrocephalus. The concept, modeled after the nipple of a baby bottle, allowed for one-way flow through a pressure-regulated valve. The shunt system could take advantage of surgical aseptic technique (1860s), and followed historical predecessors: the external ventricular drain (1881), and ventricular-subdural shunts made of glass wool, gold tubes, bundled catgut (1890s), rubber (1903), glass, silver, and linen threads (1908–1926). Catheters from the ventricles to the cisterna magna and Nulsen-Spitz’s ventriculo-jugular shunt were revolutionary precursors made of rubber or polyethylene (1940–1950) [1]. The Spitz–Holter valve is the result of these efforts and that of John Holter. The modern-day shunt remains similar now as when it was created (1955, implanted in 1957/58): slight modifications to the catheter, major iterations to the shunt valve, and the addition of compensators for gravity and siphoning are key components to treatment.

Demystifying the Black Box: The Importance of Interpretability of Predictive Models in Neurocritical Care

Abstract: Abstract Neurocritical care patients are a complex patient population, and to aid clinical decision-making, many models and scoring systems have previously been developed. More recently, techniques from the field of machine learning have been applied to neurocritical care patient data to develop models with high levels of predictive accuracy. However, although these recent models appear clinically promising, their interpretability has often not been considered and they tend to be black box models, making it extremely difficult to understand how the model came to its conclusion. Interpretable machine learning methods have the potential to provide the means to overcome some of these issues but are largely unexplored within the neurocritical care domain. This article examines existing models used in neurocritical care from the perspective of interpretability. Further, the use of interpretable machine learning will be explored, in particular the potential benefits and drawbacks that the techniques may have when applied to neurocritical care data. Finding a solution to the lack of model explanation, transparency, and accountability is important because these issues have the potential to contribute to model trust and clinical acceptance, and, increasingly, regulation is stipulating a right to explanation for decisions made by models and algorithms. To ensure that the prospective gains from sophisticated predictive models to neurocritical care provision can be realized, it is imperative that interpretability of these models is fully considered.

Neocortical tissue recovery in severe congenital obstructive hydrocephalus after intraventricular administration of bone marrow-derived mesenchymal stem cells.

Abstract: In obstructive congenital hydrocephalus, cerebrospinal fluid accumulation is associated with high intracranial pressure and the presence of periventricular edema, ischemia/hypoxia, damage of the white matter, and glial reactions in the neocortex. The viability and short time effects of a therapy based on bone marrow-derived mesenchymal stem cells (BM-MSC) have been evaluated in such pathological conditions in the hyh mouse model. BM-MSC obtained from mice expressing fluorescent mRFP1 protein were injected into the lateral ventricle of hydrocephalic hyh mice at the moment they present a very severe form of the disease. The effect of transplantation in the neocortex was compared with hydrocephalic hyh mice injected with the vehicle and non-hydrocephalic littermates. Neural cell populations and the possibility of transdifferentiation were analyzed. The possibility of a tissue recovering was investigated using 1H High-Resolution Magic Angle Spinning Nuclear Magnetic Resonance (1H HR-MAS NMR) spectroscopy, thus allowing the detection of metabolites/osmolytes related with hydrocephalus severity and outcome in the neocortex. An in vitro assay to simulate the periventricular astrocyte reaction conditions was performed using BM-MSC under high TNFα level condition. The secretome in the culture medium was analyzed in this assay. Four days after transplantation, BM-MSC were found undifferentiated and scattered into the astrocyte reaction present in the damaged neocortex white matter. Tissue rejection to the integrated BM-MSC was not detected 4 days after transplantation. Hyh mice transplanted with BM-MSC showed a reduction in the apoptosis in the periventricular neocortex walls, suggesting a neuroprotector effect of the BM-MSC in these conditions. A decrease in the levels of metabolites/osmolytes in the neocortex, such as taurine and neuroexcytotoxic glutamate, also indicated a tissue recovering. Under high TNFα level condition in vitro, BM-MSC showed an upregulation of cytokine and protein secretion that may explain homing, immunomodulation, and vascular permeability, and therefore the tissue recovering. BM-MSC treatment in severe congenital hydrocephalus is viable and leads to the recovery of the severe neurodegenerative conditions in the neocortex. NMR spectroscopy allows to follow-up the effects of stem cell therapy in hydrocephalus.

A multicenter retrospective study of heterogeneous tissue aggregates obstructing ventricular catheters explanted from patients with hydrocephalus.

Abstract: Implantation of ventricular catheters (VCs) to drain cerebrospinal fluid (CSF) is a standard approach to treat hydrocephalus. VCs fail frequently due to tissue obstructing the lumen via the drainage holes. Mechanisms driving obstruction are poorly understood. This study aimed to characterize the histological features of VC obstructions and identify links to clinical factors. 343 VCs with relevant clinical data were collected from five centers. Each hole on the VCs was classified by degree of tissue obstruction after macroscopic analysis. A subgroup of 54 samples was analyzed using immunofluorescent labelling, histology and immunohistochemistry. 61.5% of the 343 VCs analyzed had tissue aggregates occluding at least one hole (n = 211) however the vast majority of the holes (70%) showed no tissue aggregates. Mean age at which patients with occluded VCs had their first surgeries (3.25 yrs) was lower than in patients with non-occluded VCs (5.29 yrs, p < 0.02). Mean length of time of implantation of occluded VCs, 33.22 months was greater than for non-occluded VCs, 23.8 months (p = 0.02). Patients with myelomeningocele had a greater probability of having an occluded VC (p = 0.0426). VCs with occlusions had greater numbers of macrophages and astrocytes in comparison to non-occluded VCs (p < 0.01). Microglia comprised only 2–6% of the VC-obstructing tissue aggregates. Histologic analysis showed choroid plexus occlusion in 24%, vascularized glial tissue occlusion in 24%, prevalent lymphocytic inflammation in 29%, and foreign body giant cell reactions in 5% and no ependyma. Our data show that age of the first surgery and length of time a VC is implanted are factors that influence the degree of VC obstruction. The tissue aggregates obstructing VCs are composed predominantly of astrocytes and macrophages; microglia have a relatively small presence.

Characterization of a multicenter pediatric-hydrocephalus shunt biobank

Abstract: Pediatric hydrocephalus is a devastating and costly disease. The mainstay of treatment is still surgical shunting of cerebrospinal fluid (CSF). These shunts fail at a high rate and impose a significant burden on patients, their families and society. The relationship between clinical decision making and shunt failure is poorly understood and multifaceted, but catheter occlusion remains the most frequent cause of shunt complications. In order to investigate factors that affect shunt failure, we have established the Wayne State University (WSU) shunt biobank. To date, four hospital centers have contributed various components of failed shunts and CSF from patients diagnosed with hydrocephalus before adulthood. The hardware samples are transported in paraformaldehyde and transferred to phosphate-buffered saline with sodium azide upon deposit into the biobank. Once in the bank, they are then available for study. Informed consent is obtained by the local center before corresponding clinical data are entered into a REDCap database. Data such as hydrocephalus etiology and details of shunt revision history. All data are entered under a coded identifier. 293 shunt samples were collected from 228 pediatric patients starting from May 2015 to September 2019. We saw a significant difference in the number of revisions per patient between centers (Kruskal–Wallis H test, p value < 0.001). The leading etiology at all centers was post-hemorrhagic hydrocephalus, a fisher’s exact test showed there to be statistically significant differences in etiology between center (p = 0.01). Regression showed age (p < 0.01), race (p = 0.038) and hospital-center (p < 0.001) to explain significant variance in the number of revisions. Our model accounted for 31.9% of the variance in revisions. Generalized linear modeling showed hydrocephalus etiology (p < 0.001), age (p < 0.001), weight and physician (p < 0.001) to impact the number of ventricular obstructions. The retrospective analysis identified that differences exist between currently enrolled centers, although further work is needed before clinically actionable recommendations can be made. Moreover, the variables collected from this chart review explain a meaningful amount of variance in the number of revision surgeries. Future work will expand on the contribution of different site-specific and patient-specific factors to identify potential cause and effect relationships.