Skip to main content

Anomaly Detection in MRI Scans through Deep Learning: A Healthy Cohort Training Approach

Primary supervisor

Zhaolin Chen


The early detection of neurological abnormalities through Magnetic Resonance Imaging (MRI) is crucial in the medical field, potentially leading to timely interventions and better patient outcomes. However, the traditional diagnostic process is often time-consuming and subject to human error. This project seeks to improve this aspect by employing deep learning for anomaly detection in MRI scans, exclusively using images from healthy participants for model training [1].

The project aims to develop the algorithm to understand the parameters of a 'healthy' MRI scan [2], subsequently identifying deviations indicative of potential anomalies or pathological changes. This strategy significantly minimizes the bias introduced by various disease-specific patterns, enabling the model to identify unknown or unexpected anomalies efficiently.

The goal is to develop a robust, sensitive, and unbiased anomaly detection system that can assist healthcare professionals in early diagnosis and treatment planning, thereby improving patient care and management.

anomaly detection
Autoencoders are trained to replicate inputs. The differences between the input and its reconstruction can be used for anomaly detection in signal or image data

What will students learn in this project:

Participants in this project will submerge themselves in various learning opportunities:

  • Deep Learning Techniques: Understand the nuances of deep learning algorithms and their application in anomaly detection, specifically within the realm of medical imaging.
  • MRI Imaging Insights: Gain deep insights into MRI technology, focusing on the characteristics of scans from healthy individuals, understanding the subtle differences that indicate potential issues.
  • Data Management and Ethics: Learn the principles of managing sensitive health data, ensuring the highest standards of privacy, confidentiality, and ethical research practices.
  • Algorithm Training: Acquire hands-on experience in training deep learning models, understanding the importance of data, and nuances of 'healthy' baselines in training phases.
  • Validation and Testing: Engage in the validation and testing of trained models, using unseen data to assess algorithm performance, and understand the practical challenges of real-world application.
  • Research Documentation: Develop skills in documenting research findings, preparing reports, and contributing to scientific literature.
  • Interdisciplinary Collaboration: Work in an interdisciplinary environment, collaborating with professionals in healthcare, machine learning, and data science.

Student cohort

Double Semester


[1] Emergency triage of brain computed tomography via anomaly detection with a deep generative model,

[2] Improving portable low-field MRI image quality through image-to-image translation using paired low-and high-field images,

Required knowledge

Ideal candidates for this project should demonstrate:

  • Background in Deep Learning: Experience with deep learning methodologies, particularly in the context of image recognition or anomaly detection.
  • Programming Proficiency: Strong skills in programming languages used in machine learning environments, such as Python, and familiarity with relevant libraries and frameworks (e.g., TensorFlow, PyTorch, MATLAB).
  • Research Orientation: A strong research background with a focus on producing publishable quality work and a passion for contributing to new discoveries in medical science.


Application Instructions:​​​​​​​

Interested candidates are invited to prepare a comprehensive application, including:

  • A motivation letter expressing your interest in the project, detailing relevant experience, and explaining your aspirations for participating in this research.
  • Academic transcripts and any other documents demonstrating your credentials and achievements in related fields.
  • If applicable, a list of publications or previous research contributions.
  • Please consolidate your documents into a single PDF file for submission. For further information or inquiries about the project, feel free to reach out through the contact information provided below.