PhD Thesis: Application of Pattern Recognition in Clinical Ex-Vivo 1H-Magnetic Resonance Spectroscopy

As I was preparing my slides for the ISI CODATA International Training Workshop on Big Data, revelling on polemic about the need for Open Access, Open Data, Open Source etc., it occurred to me that my PhD thesis is languishing in the library of the Institute of Cancer Research in Sutton.

A quick search for it online suggests that it’s probably available if you a) have access to certain institutional memberships b) go to Sutton.

As there have never been any ‘patents pending’ on the content of my thesis, I feel that it isn’t breaching any copyright worth suing over to post the thing here. It’s a 4.5mb .pdf document (see link below). This includes all the theory and results etc but also the Matlab code that was used.

PhD Thesis: Applications of Pattern Recognition in Clinical Ex-Vivo 1H-Magnetic Resonance Spectroscopy (.pdf)

In the interests of being more modern, open and accessible, I’ll put up a GitHub with the code too (will update this post with that later).

Although the results were not overly interesting in themselves, some of the theory sections may prove interesting to the novice researcher and there were some uses of Statistical Total Correlation Spectroscopy (STOCSY) that may yet be worth publishing – if you feel inclined, please let me know.

In the interests of making this post more accessible/findable, here is the table of contents from the thesis:

  1. Introduction

    1. Nuclear magnetic resonance in medicine
    2. Analysis of metabolites using NMR spectra
      1. Peak analysis
      2. Spectral analysis and pattern recognition
    3. Metabolomics in practice
      1. Functional genomics
      2. Toxicology and drug development
      3. Cancer diagnosis and prognosis
    4. Aims and structure of thesis
    5. Provenance of data
  2. Pattern recognition theory
    1. Matrix decomposition methods
      1. Principal component analysis
      2. Independent component analysis
      3. Non-negative matrix factorisation
    2. Classification
      1. Linear discriminant analysis
      2. Partial least squares discriminant analysis
    3. Validation and performance
      1. Cross-validation
      2. ROC curves
    4. Chapter summary
  3. Processing theory and implementation
    1. Introduction
      1. Aims and layout
    2. NMR processing methods
      1. Two domains and the Fourier transform
      2. Apodisation and zero-filling
      3. Phase
      4. Alignment
      5. Removal of water
      6. Normalisation
    3. Processing software developed during thesis
      1. Choice of processing system
      2. Design
      3. Input and output
      4. Provenance of Code
    4. Additional applications
      1. Liver and tumour response to 90Y-DOTATOC
      2. Apparent diffusion coefficient histogram parameters used for treatment response indicators
    5. Chapter summary
  4. Processing studies
    1. Introduction
      1. Aims and layout
    2. Removal of contaminants
      1. Background
      2. Method
      3. Manual modelling
      4. Investigation of ‘knock on effect’ of lignocaine
      5. Results
      6. Conclusion
    3. Normalisation
      1. Background
      2. Method
      3. Results
      4. Conclusion
    4. Chapter summary
  5. Application to 1H HR MAS of cervical tissue samples
    1. Introduction
      1. Motivation
      2. Pathology
      3. Framework in literature
      4. Aims and layout
    2. Comparison of separation achieved by peak areas and spectral analysis
      1. Background
      2. Data
      3. Methods – peak areas
      4. Methods – spectral analysis
      5. Results and discussion
      6. Conclusion
    3. Separation of cancer and ‘non-cancer’ using mobile lipid resonance peaks
      1. Data
      2. Methods
      3. Results and discussion
      4. Conclusion
    4. Chapter summary
  6. Application to prostate tissue samples and cell lines
    1. Introduction
      1. Background
      2. Motivation
      3. Aims and layout of chapter
    2. Literature
      1. Magnetic resonance spectroscopy imaging
      2. Metabolomics and high resolution spectroscopy
      3. Pathology and biology
      4. Androgen independence and cancer recurrence
    3. Tissue samples near and far from tumour
      1. Introduction
      2. Data
      3. Spectroscopy
      4. Analysis
      5. Results
      6. Conclusion
    4. Cell lines grown with and without androgens
      1. Introduction
      2. Data
      3. Spectroscopy
      4. Analysis
      5. Results
      6. Conclusion
    5. Chapter summary
  7. Application to drug treatment studies using xenograft tissue extracts
    1. Introduction
    2. Aims and Layout
    3. Data collection
      1. Issues
    4. Preliminary check: can the datasets be merged?
      1. Methods
      2. Results
      3. Conclusion
    5. Processing methods common to studying SJG136 and MN58b datasets
      1. Preprocessing
      2. Pattern recognition
    6. SJG136, DNA binding-agent and tumour growth inhibitor applied to human colon xenografts
      1. Background
      2. Prior findings
      3. Data
      4. Methods particular to SJG136 data
      5. Results
      6. Investigation of components
      7. Discussion and conclusion
    7. MN58b, choline kinase inhibitor
      1. Background
      2. Prior findings
      3. Data
      4. Methods specific to MN58b data
      5. Results
      6. Discussion and conclusion
    8. Chapter summary
  8. Discussion
    1. Experimental design and data collection
    2. Data processing
    3. Metabolomics methods
    4. Conclusion
  9. Code

 

 

PhD Thesis: Applications of Pattern Recognition in Clinical Ex-Vivo 1H-Magnetic Resonance Spectroscopy (.pdf)

Leave a Reply

Your email address will not be published. Required fields are marked *