test blog

Internship Position: Scalable and Versatile Normative Modelling using Artificial Neural Networks

Recently, there has been great interest in applying machine learning methods to quantitative biological measures (biomarkers) to assist medical decision making. In psychiatry, this is very challenging because the diagnosis is typically based on clinical symptoms and the underlying biology is highly heterogeneous. Normative modeling is one promising approach for this that aims to characterize variation across a healthy cohort before making predictions so that subjects that deviate from the resulting normative model can be detected as outliers in an anomaly detection setting and the pattern underlying the deviation can be further analyzed to understand the biological underpinnings. Gaussian process regression (GPR) and its multitask variation (MT-GPR) showed significant potentials to model variation in neuroimaging data in the normative modeling framework. However, on the downside, the high computational complexities limit their applications, especially when applied to high-dimensional neuroimaging data. We, at the predictive clinical neuroscience lab, are opening the call for an internship position with the main aim to decrease these computational complexities by adopting alternative approaches such as neural networks in the normative modeling framework. This position provides a unique opportunity for the trainee to learn the state-of-the-art machine learning approaches and their applications on clinical neuroimaging data. Our traineeship is designed to prepare the student to apply the acquired knowledge to his/her next research stages for solving emerging problems in precision psychiatry with the prospect of publication in prestigious venues. Our work is highly interdisciplinary, and applicants must have strong statistical and programming (Python, Matlab) skills. Previous experience with neuroscience is very helpful. Candidates with backgrounds in statistics, computer science, engineering, and related areas are welcome to apply. The internship is based in the Donders Institute that brings together more than 600 researchers from 35 countries with the goal of understanding the mechanistic underpinnings of human cognition and behavior in health and disease. The trainee will have access to the state-of-the-art computational facilities, e.g., high-performance computing cluster and several GPUs, to accomplish the traineeship goals. Interested applicants should apply by sending their CV to s.kia@donders.ru.nl .

Relevant literature:

  • Marquand, Andre F., Iead Rezek, Jan Buitelaar, and Christian F. Beckmann. “Understanding heterogeneity in clinical cohorts using normative models: beyond case-control studies.” Biological psychiatry 80, no. 7 (2016): 552-561.
  • Wolfers, Thomas, Nhat Trung Doan, Tobias Kaufmann, Dag Alnæs, Torgeir Moberget, Ingrid Agartz, Jan K. Buitelaar et al. “Mapping the heterogeneous phenotype of schizophrenia and bipolar disorder using normative models.” JAMA psychiatry (2018).
  • Kia, Seyed Mostafa, Christian F. Beckmann, and Andre F. Marquand. “Scalable Multi-Task Gaussian Process Tensor Regression for Normative Modeling of Structured Variation in Neuroimaging Data.” arXiv preprint arXiv:1808.00036 (2018).

Internship Position: Bayesian Neural Networks for Normative Modeling on Multi-Site Neuroimaging Data

Recently, there has been great interest in applying machine learning methods to quantitative biological measures (biomarkers) to assist medical decision making. In psychiatry, this is very challenging because the diagnosis is typically based on clinical symptoms and the underlying biology is highly heterogeneous. Normative modeling is one promising approach for this that aims to characterize variation across a healthy cohort before making predictions so that subjects that deviate from the resulting normative model can be detected as outliers in an anomaly detection setting and the pattern underlying the deviation can be further analyzed to understand the biological underpinnings. Due to its design, normative modeling is generally applied to large clinical cohorts, thus it should be able to cope with multi-site neuroimaging data. Batch effects in multi-site neuroimaging data, due to different scanner types and calibrations, can have a catastrophic effect on the quality of the normative model. We, at the predictive clinical neuroscience lab, are opening the call for an internship position with the goal to address this problem using Bayesian neural network theory in the normative modeling framework. This position provides a unique opportunity for the trainee in order to learn the state-of-the-art machine learning approaches and their applications on clinical neuroimaging data. Our traineeship is designed to prepare the student to apply the acquired knowledge to his/her next research stages for solving emerging problems in precision psychiatry with the prospect of publication in prestigious venues. Our work is highly interdisciplinary, and applicants must have strong statistical and programming (Python, Matlab) skills. Previous experience with neuroscience is very helpful. Candidates with backgrounds in statistics, computer science, engineering, and related areas are welcome to apply. The internship is based in the Donders Institute that brings together more than 600 researchers from 35 countries with the goal of understanding the mechanistic underpinnings of human cognition and behavior in health and disease. The trainee will have access to the state-of-the-art computational facilities, e.g., high-performance computing cluster and several GPUs, to accomplish the traineeship goals. Interested applicants should apply by sending their CV to s.kia@donders.ru.nl .

Relevant literature:

  • Marquand, Andre F., Iead Rezek, Jan Buitelaar, and Christian F. Beckmann. “Understanding heterogeneity in clinical cohorts using normative models: beyond case-control studies.” Biological Psychiatry 80, no. 7 (2016): 552-561.
  • Nielson, Dylan M., Francisco Pereira, Charles Y. Zheng, Nino Migineishvili, John A. Lee, Adam G. Thomas, and Peter A. Bandettini. “Detecting and harmonizing scanner differences in the ABCD study-annual release 1.0.” bioRxiv (2018): 309260.
  • Lacoste, Alexandre, Boris Oreshkin, Wonchang Chung, Thomas Boquet, Negar Rostamzadeh, and David Krueger. “Uncertainty in Multitask Transfer Learning.” arXiv preprint arXiv:1806.07528 (2018).