Prediction of pH using acidoCEST MRI and machine learning

Introduction

• Brief description of current acidoCEST method and its limitations
• Previous work in ML and why it could solve some of these issues. Previous work by our lab
• Objective of this paper.

Theory

• Short description of ML approach (learning and testing)
• Describe that our current methodology does not account for variable experimental conditions, but it could do if we use an ML approach.
• Description of using the entire Z spectra.

Methods

-ML and analysis

• Short description of ML approach (learning and testing)
• Lorentzian fitting
• Describe that our current methodology does not account for variable experimental conditions, but it could do if we use an ML approach.
• Description of using the entire Z spectra.

-Experimental

• Samples
• MRI methods

Results

• Regression
• standard method when applied to out-of-training data
• standard method when built using a ML approach
• Linear regression using entire spectra
• Classification
• standard method when applied to out-of-training data
• standard method when built using a ML approach
• Linear regression using entire spectra

Discussion

• how does the method building method affect the accessible pH range
• limitation of current study
• conclusions

Figures

• Figure 01: Principles of ML. Simlar to Joey's thesis
• Figure 02: Effect of pH on the "shape" of a Z spectra for Iopamidol - Include other molecules in supplemental info.
• Figure 03: Iopamidol pH regression
• Figure 04: polymer pH regression
• Figure 05: monomer-2 pH regression
• Figure 06: Iopamidol pH classification (two cut offs)
• Figure 07: polymer pH classification (two cut offs)
• Figure 08: monomer-2 pH classification (two cut offs)