Periventricular Leukomalacia (PVL) Prediction

What is PVL and why it is Important

Birth defects affect approximately 3% of all births and are a leading cause of infant mortality in the United States. Congenital heart defects (CHD) are among the most common birth defects in neonates. Newborns with congenital heart disease are at high risk for different types of brain injury. Underlying this neuro-behavioral dysfunction is a high prevalence of white matter injury, termed periventricular leukomalacia (PVL) presented in Fig. (1), the physiological cause of which is currently a mystery.

Since clinical examination of neonates is not a good predictive of brain injury, clinicians are reliant on magnetic resonance imaging (MRI) for injury classification and quantification. The MRI usually is done on the patient a week after cardiac surgery. While MRI provides excellent diagnostic information, it yields no information on the timing or cause of injury and is not ideally suited to be used as a predictive tool as Fig. (2)

This figure represents the difference between the current diagnostic methods based on MRI and developed prognostic method for clinical decision support.

The Need for Clinical Decision Support System

Since the current practice methods in medicine can’t provide predictive information regarding when and why PVL is happening, it is of great interest to develop a reliable algorithm to provide a decision support for PVL prediction. In this work, we have used the physiological measurement data collected from patients at CHOP for a period of up to 12 hours after cardiac surgery to predict occurrence of PVL. The collected data is heterogeneous due to fact that it included high resolution vital signs data, scattered blood gas data, and one time demographic recordings. Samples of data for PVL and healthy patients have been provided in Fig. (3)

A comparative plot of healthy and PVL vital data has been shown in Fig. (4). As seen in the plot there no significant difference between healthy and PVL patients in terms of means and standard deviation of the data. Thus, statistical methods are not able to provide an accurate and timely decision support for PVL occurrence prediction. As a result, we need to pursue more sophisticated machine learning algorithm for higher prediction accuracy.

The other problems that needed to be addressed in this work include: very limited physiological understanding of the PVL pathophysiology, which made application of model based methods unfeasible, change in the dynamic of patients population due to the fact that patients were diagnosed with different types of CHD. We designed an algorithm for the task of data classification and rule extraction.

Since the PVL pathophysiology knowledge is very limited, we need to design an algorithm to discover hidden relationship between the data. The developed algorithm consists of three main steps: feature extraction, feature ranking and classifier design. First, the patient data was collected at the hospital and was used to form the pool of features. A modified version of the mutual information method that takes into account the reliability of the collected data was then used for ranking the extracted features in the feature pool. After forming the ordered feature set the optimal feature subset that encapsulates the most critical features was selected by maximizing the class separability measure. The optimal feature subset was reduced in size compared to the original feature set; however, by maximizing the class separability measure, this subset can be expected to result in a higher accuracy in the final prediction. The selected features were then fed to the classifier. We then improved the accuracy of the developed decision support system over time as we collect more data by implementation of online learning algorithm. The schematic of the designed algorithm has been presented in Fig (5)

Results

The most notable findings of the PVL research are:

Sample Publications

Book Chapter

  1. C. Nataraj, A. Jalali, P. Ghorbanian, Application of Computational Intelligence Techniques for Cardiovascu¬lar Diagnostics, In Book: The Cardiovascular System -Physiology, Diagnostics and Clinical Implications, InTech, 2012.

Journal Publications

  1. B Samanta, GL Bird, M Kuijpers, RA Zimmerman, GP Jarvik, G Wernovsky, RR Clancy, DJ Licht, J William Gaynor, C Nataraj, Prediction of periventricular leukomalacia. Part I. Selection of hemodynamic features using logistic regression and decision tree algorithms, (46), 201-215, 2009.
  2. B Samanta, GL Bird, M Kuijpers, RA Zimmerman, GP Jarvik, G Wernovsky, RR Clancy, DJ Licht, J William Gaynor, C Nataraj, Prediction of periventricular leukomalacia. Part II: Selection of hemodynamic features using computational intelligence, (46), 217-231, 2009.
  3. A Jalali, DJ Licht, C Nataraj, Prediction of Periventricular Leukomalacia (PVL) Occurrence in Neonates After Neonatal Heart Surgery, IEEE Journal of Health and Biomedical Informatics, (18) 1453-1460, 2014.
  4. A Jalali, P Ghorbanian, A Ghaffari, C Nataraj, A novel technique for identifying patients with ICU needs using hemodynamic features, Advances in Fuzzy Systems, 2012.
  5. A Jalali, A Ghaffari, P Ghorbanian, C Nataraj, Identification of sympathetic and parasympathetic nerves function in cardiovascular regulation using ANFIS approximation, Artificial Intelligence in Medicine, (52) 27-32, 2011.
  6. P Ghorbanian, A Jalali, A Ghaffari, C Nataraj, An improved procedure for detection of heart arrhythmias with novel pre-processing techniques, Expert Systems, (29) 478-491, 2012.

Conference Publications

  1. A Jalali, RA Berg, V Nadkarni, C Nataraj, Improving Cardiopulmonary Resuscitation (CPR) by Dynamic Variation of CPR Parameters, Dynamic Systems and Controls Conference (DSCC), 2013.
  2. A Jalali, DJ Licht, C Nataraj, Discovering Hidden Relationships in Physiological Signals For Prediction of Periventricular Leukomalacia (PVL), Engineering in Medicine and Biology Society (EMBC), 2013 Annual International Conference of the IEEE.
  3. A Jalali, DJ Licht, C Nataraj, Time-Frequency analysis of hemodynamic waveforms to predict the oc-currence and severity of Periventricular Leukomalacia, Dynamic Systems and Controls Conference (DSCC), 2012.
  4. A Jalali, DJ Licht, C Nataraj, Application of decision tree in the prediction of periventricular leuko-malacia (PVL) occurrence in neonates after heart surgery, Engineering in Medicine and Biology Society (EMBC), 2012 Annual International Conference of the IEEE Pages 5931-5934.(link)
  5. A Jalali, RA Berg, V Nadkarni, C Nataraj, Model based optimization of the cardiopulmonary resus-citation (CPR), procedure Engineering in Medicine and Biology Society (EMBC), 2012 Annual International Conference of the IEEE Pages 715-718.
  6. A Jalali, DJ Licht, C Nataraj, Prediction of occurrence of Periventricular Leukomalacia (PVL) in Neonates after Heart Surgery Using a Decision Tree Algorithm, International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (IDETC/CIE2012).
  7. A Jalali, GF Jones, DJ Licht, C Nataraj, Computational Modeling of Hypoplastic Left Heart Syndrome (HLHS) in Newborn Babies, Engineering in Medicine and Biology Society (EMBC), 2011 Annual International Conference of the IEEE Pages 185-189.
  8. A Jalali, DJ Licht, C Nataraj, Computational Modeling of Hypoplastic Left Heart Syndrome (HLHS) in Newborn Babies, International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (IDETC/CIE2011) Pages 185-189.
  9. A Jalali, M Butchy and A Ghaffari, C Nataraj, Feature Extraction and Abnormality Detection in Au-tonomic Regulation of Cardiovascular System, International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (IDETC/CIE2011) Pages 201-205.
  10. P Ghorbanian, A Ghaffari, A Jalali, C Nataraj, Heart arrhythmia detection using continuous wavelet transform and principal component analysis with neural network classifier, Computing in Cardiology (CinC2010) Pages 669-672.
  11. A Jalali, A Ghaffari, P Ghorbanian, F Jalali, C Nataraj, Quantitative analysis of heart rate baroreflex in healthy subjects using adaptive neuro fuzzy inference system approximation, Computing in Cardiology (CinC2010) Pages 951-954.