MRI, machine learning may help predict need for postnatal CSF diversion in fetal ventriculomegaly

MRI image analysis and machine learning may be a more accurate and efficient when applied to fetal MRI findings to predict the need for postnatal cerebrospinal fluid (CSF) diversion, according to a retrospective analysis study recently published by JAMA. According the study, researchers were able to develop a prognostic information model that can guide a more efficient candidate selection for potential fetal surgery.  

According to the study, fetal ventriculomegaly occurs in every one in 1,000 births, usually diagnosed through mid-trimester ultrasounds and fetal MRI. The causes of fetal ventriculomegaly are predicted by experts to be caused by the loss of cerebral tissue, obstruction of the ventricular system or excessive CSF production. Furthermore, the condition progresses in 15 percent of cases, with the fetus showing an atrial diameter of 10 millimeters or larger and causing increased intracranial pressure after birth.   

"With a renewing interest in fetal surgery, the predictive model may be used to select patients with the highest likelihood of requiring postnatal CSF diversion as candidates for in utero shunting," said lead author of the study Jared Pisapia, MD, from the department of neurosurgery from the University of Pennsylvania. "Work is currently under way to incorporate the fetal MRI-based predictive model into user-friendly software, and future studies will investigate the integration of the software into clinical workflow and prospectively assess model accuracy."   

Pisapia and his colleagues determined whether the extraction of various imaging features from fetal MRIs using machine learning techniques can predict which patients need postnatal CSF diversion after birth.  

"The ability to prospectively identify which patients with fetal ventriculomegaly will develop postnatal hydrocephalus and require subsequent CSF diversion would provide critical prognostic in-formation to parents and aid in the clinical management in the postnatal period," Pisapia said.  

A total of 253 patients diagnosed with fetal ventriculomegaly from 2008 through 2014 were identified and recruited from an institutional database from the Children's Hospital of Philadelphia (CHOP) and analyzed by researchers from 2008 through 2015.  

Of the 253 patients recruited for the study, 25 of whom required postnatal CSF diversion were selected and matched by gestational age with 25 patients with fetal ventriculomegaly who did not require CSF diversion.  

Pisapia and his colleagues generated a fetal MRI based model stemmed from four notable measurements extracted from the fetal MRI images: linear measurements, area, volume, and morphologic features. Measurements were then inputted into a machine learning algorithm that determined the combination of features to predict whether each patient required postnatal CSF diversion surgery. Additionally, all MRI images underwent smoothing and bias correction processing. 

Pisapia and his colleagues concluded that of the 50 patients found with fetal ventriculomegaly extracted multiple imaging features enough to correctly classify postnatal cerebrospinal fluid diversion status with an 82 percent accuracy. Additionally, the fetal MRI based model achieved a 91 percent accuracy when classifying postnatal CSF diversion status.  

"Our findings suggest that image analysis and machine learning techniques can be applied to fetal MRI to predict the need for postnatal CSF diversion in patients with fetal ventriculomegaly with high accuracy and generalizability," Pisapia said. "A significant strength of the computational approach is the ability to simultaneously assess multiple imaging features, most of which are not appreciable by visual inspection alone."  

The image based predictive model characterized fetal ventriculomegaly and the development of postnatal hydrocephalus the best through the integration of patterns of ventricle size, shape, and configuration, according to the study discussion, according to study findings.   

"An image-based predictive model with high accuracy and generalizability may provide prenatal prognostic information and help guide postnatal clinical management in fetal ventriculomegaly," Pisapia concluded from study findings. 

""

A recent graduate from Dominican University (IL) with a bachelor’s in journalism, Melissa joined TriMed’s Chicago team in 2017 covering all aspects of health imaging. She’s a fan of singing and playing guitar, elephants, a good cup of tea, and her golden retriever Cooper.

Trimed Popup
Trimed Popup