Dr Mintu Nath

Biography

Before joining the University of Aberdeen, I worked as a Senior Statistician with Biomathematics and Statistics Scotland, and later as a Senior Biomedical Statistician with the University of Leicester. Broadly, my research, consultancy and teaching focussed on the areas of statistical genetics, genetic epidemiology and applied statistics, particularly those involving high-volume and high-dimensional datasets.

Research Overview

My research interests include exploring and integrating diverse resources of large patient cohorts - comprising of anthropometric, clinical, imaging, genetic and multi-omics data - to develop predictive models to evaluate the risk of important diseases.

Current Research

Finding Endometriosis using Machine Learning (FEMaLE)

I am collaborating with researchers from the University of Aberdeen, the University of Edinburgh and Aarhus University (Denmark) to explore the prevalence of endometriosis-like symptoms in a population of general practice patients and identify factors that can aid earlier identification of women most at risk of a diagnosis of endometriosis. We will also develop and validate a predictive model to estimate the probability of a laparoscopic diagnosis of endometriosis from the enumerated symptom trajectory, medication, lifestyle factors etc. 

A randomised, double-blind placebo controlled trial of the effectiveness of the beta-blocker bisoprolol in preventing exacerbations of chronic obstructive pulmonary disease.

I am leading statistical analyses of BICS (Bisoprolol in COPD Study) clinical trial. The study is funded by NIHR Health Technology Assessment. The primary objective of this trial is to determine the clinical and cost-effectiveness (in terms of number of exacerbations requiring change in management) of adding bisoprolol (maximal dose 5mg once a day, or maximum tolerated dose) to usual COPD therapies in patients with COPD at high risk of exacerbation (history of at least two COPD exacerbations in the previous year) over a one-year follow-up period. 

Using multi-dimensional eye-tracking data (saccade) to develop predictive models for diagnosing and managing major mental illness

Funded by the NHS Grampian Endowment award and Millar Mackenzie Trust, we are implementing machine learning tools to develop predictive models of mental health conditions using multi-dimensional eye-tracking data. 

Mathematical and statistical evaluation of Positron Emission Tomography (PET) data for the management of cancer

I am collaborating with Prof Professor Andy Welch at the University of Aberdeen on the application of mathematical and statistical tools to analyse PET data for detecting pathophysiological changes in different tissues. We are currently supervising one PhD student. 

Understanding and improving frailty pathways in the hospital setting

In collaboration with NHS Grampian researchers, I am developing a mathematical and statistical framework of patient flow in hospital settings using historical data on hospital episodes. The objective is to translate patients' movement in the hospital in a multistate modelling framework where patients stay in one state at a given time point but move between states with time.

BIOSTAT-CHF - BIOlogy Study to TAilored Treatment in Chronic Heart Failure

BIOSTAT-CHF is a Consortium of 11 European countries, funded by an FP7 Health grant from the European Union. I am leading the work to develop a comprehensive bioinformatics and statistical analysis workflow for analysing high-resolution array GeneChip® Human Transcriptomic Array 2.0 data in collaboration with the team at the University of Leicester and the EU partner organisations. 

Past Research

SHERLOCK - An observational Study on HEalthcare Resource utiLisation related to exacerbatiOns in patients with COPD

Using multiple sources and high-volume population data of primary care patients affected with Chronic Obstructive Pulmonary Disease (COPD), the project investigates the associations between the history of exacerbations with the frequency and rate of future exacerbations as well as healthcare resource utilisation (HCRU) cost. There is a PhD opportunity to develop predictive models using linked electronic health records in a machine-learning framework to assess the risk of moderate and severe exacerbations in Chronic Obstructive Pulmonary Disease (COPD). Contact me if you are interested.

The Optical Coherence Tomography (OCT) of the retinal layers and its association with the retinal pathophysiology in infants

I have collaborated with researchers at the University of Leicester to develop comprehensive statistical models to analyse foveal and retinal morphological data in preterm infants. We are developing a data-driven approach to understand the dynamics of retinal development in neonatal babies and its association with the retinopathy of prematurity using OCT data.

Data-driven design, monitoring, and adaptation of COVID and non-COVID clinical care pathways

We are using real-time as well as historical data to understand the healthcare utilisation of people shielding during COVID-19 as well as affected by the disease. The project is funded by NHS Endowment Research Grant and the Scottish Government Chief Scientist Office. 

The UK-REBOA (Resuscitative Endovascular Balloon Occlusion of the Aorta) Trial

Funded by NIHR, I am collaborating with clinicians and scientists to develop methodologies for the elicitation of experts' probabilities in the context of randomised control trial. 

Collaborations

Can deep learning help improve automated retinal image analysis algorithms for diabetic retinopathy screening?

I am collaborating with clinicians and scientists at the NHS and the University of Aberdeen for developing and optimising deep learning algorithms to automate the assessment of retinal features associated with diabetic retinopathy. 

Investigating socioeconomic disparities as a cause and consequence of deteriorating kidney health among people admitted to hospital with an acute illness

The project aims to establish the relationship between socioeconomic status and health outcomes among people leaving hospitals with and without acute kidney injury (AKI). The project will also explore the feasibility and added value of including individual-level (census) socioeconomic measures in clinical risk tools for predicting adverse health outcomes. 

Teaching Responsibilities

I undertake teaching and training programs in Statistics for MSc and PhD students. I also conduct statistical consultancies for students as well as the University and NHS staff. I am interested to develop user-friendly and interactive applications to enhance the learning of statistical principles. I provide a range of R courses for the University staff and Doctoral students.

Applied Statistics Course

I teach different statistical courses organised by the Medical Statistics team:

• MSc Course: Applied Statistics (PU5017)
• MSc Online course: Applied Statistics (PU5522)
• MSc Course: Machine Learning (PX5509, PX5517)
• Intermediate Medical Statistics for University staff and postgraduate students
• Other statistics courses as part of the Medical Statistics Team

ABACUS - Apps Based Activities for Communicating and Understanding Statistics

ABACUS is a set of R Shiny apps for effective communication and understanding of statistics. The current version includes explaining properties of the normal distribution, properties of the sampling distribution, one-sample z and t-tests, two samples independent (unpaired) t-test and analysis of variance. Check the ABACUS website and explore it. We will be adding new modules in future. Please visit again for more updates.

ABACUS on CRAN

ABACUS is also available as an R package on the Comprehensive R Archive Network (CRAN). Please download ABACUS from the CRAN package repository.

iDRUG: interactive Dose-Response User Guide

With fundings from the British Pharmacological Society, I am collaborating with Prof Steven Tucker at the University of Aberdeen to develop a simple, interactive and user-friendly application to conduct dose-response analysis in Pharmacology settings. The iDRUG website is live now. Watch a demonstration of iDRUG here. Further development of iDRUG is in progress. Watch this space for further details!

R Courses

I developed and presented an extensive array of introductory and advanced R courses as well as statistical analysis in the R software environment. If you are interested to organise bespoke R courses for your staff, I will be happy to help. Please contact me with the details of your request.

Non-course Teaching Responsibilities

I presented a CPD workshop R You Ready for Python at the Royal Statistical Science Conference 2022 in Aberdeen. Both R and Python users may embrace both languages and enhance their productivity by leveraging the best of both worlds. Using bite-sized simple R and Python scripts, the workshop highlights similarities and differences between these two programming environments – their syntax, semantics, and implementation framework. It demonstrates how understanding these basic and subtle concepts could benefit the efficient usage of both programming languages.

I presented the second CPD workshop Debugging Deciphered at the Royal Statistical Science Conference 2023 in Harrogate. The workshop explores the approach to debugging R and Python scripts, primarily focussing on semantic errors. Using bite-sized examples, the workshop presents the fundamental building blocks of the R and Python function: function arguments, global and local environments, and other essential function components. Next, we investigate key R and Python debugging tools like traceback, debug and browser and venture into the detective work to identify the bugs. 

I presented a CPD workshop Scale it with SQL at the Royal Statistical Science Conference 2024 in Brighton. The workshop demonstrates the implementation of SQL in open source platform (R and Python) and popular applications (like Microsoft Excel, Microsoft Access, and SAS). The workshop will be useful for statisticians, data scientists, data administrators and researchers to efficiently handling reasonably large-sized data (simple or relational) in multiple programming landscapes.