Amir Dehsarvi

Amir Dehsarvi
Ludwig-Maximilians-University of Munich | LMU · Institute for Stroke and Dementia Research (ISD)

PhD in Electronic Engineering


How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
Citations since 2016
6 Research Items
2 Citations
I work on the application of machine learning/deep learning for the analysis of biomedical data for disease diagnosis, target identification, and validation, using different types of large, complex, biomedical datasets.
Additional affiliations
February 2019 - present
Trinity College Dublin
  • PostDoc Position
  • My current research aims to identify speech markers and their underlying neural correlates (brain structure and functional connectivity) for the detection of cognitive impairment in MCI and AD.
February 2014 - July 2018
The University of York
Field of study
  • Electronics
October 2012 - September 2013
The University of York
Field of study
  • Digital Signal Processing
October 2007 - September 2010
The University of Science and Art, Yazd, Iran
Field of study
  • Electrical Engineering - Electronics


Publications (6)
Background Overt sentence reading in mild cognitive impairment (MCI) and mild-to-moderate Alzheimer’s disease (AD) has been associated with slowness of speech, characterized by a higher number of pauses, shorter speech units and slower speech rate and attributed to reduced working memory/ attention and language capacity. Objective This preliminary...
This paper reports an automated approach to the clinical monitoring of Parkinson’s disease (PD) by applying Evolutionary Algorithms (EAs) to resting-state functional magnetic imaging (rs-fMRI) data. The novel application of EAs to both map and predict the functional connectivity is considered in patients receiving the drug Modafinil versus placebo....
Full-text available
Background: Increasing efforts have focused on the establishment of novel biomarkers for the early detection of Alzheimer’s disease (AD) and prediction of Mild Cognitive Impairment (MCI)-to-AD conversion. Behavioral changes over the course of healthy ageing, at disease onset and during disease progression, have been recently put forward as promisin...
Full-text available
Background and Objective: It is commonly accepted that accurate monitoring of neurodegenerative diseases is crucial for effective disease management and delivery of medication and treatment. This research develops automatic clinical monitoring techniques for PD, following treatment, using the novel application of EAs. Specifically, the research que...
Full-text available
Accurate early diagnosis and monitoring of neurodegenerative conditions is essential for effective disease management and delivery of medication and treatment. This research develops automatic methods for detecting brain imaging preclinical biomarkers for Parkinson's disease (PD) by considering the novel application of evolutionary algorithms. A fu...
Conference Paper
Full-text available
Accurate early diagnosis and monitoring of neurodegenerative conditions is essential for effective disease management and treatment. This research develops automatic methods for detecting brain imaging preclinical biomarkers for olfactory dysfunction in early stage Parkinson's disease (PD) by considering the novel application of evolutionary algori...


Question (1)
I am running a script based on different batches in SPM on a dataset with 26 participants (2 sessions and 3 runs). This dataset has been analysed before in FSL (2 publications with statistical analysis) and I am doing a reanalysis using SPM12. The code runs smoothly for some of the participants and does the spatial preprocessing and the processing by defining a GLM and then moving on to DCM to extract the DCM.Ep.A (I have modified the sum_run_fmri_spec in config so that it stops asking for a confirmation for over writing the SPM file).
Can this be because for regressing out the WM and CSF I have two different jobs? This is the same when I want to extract the four VOIs.
I have done this according to the sample script in the practical example for rs-DCM.
Unfortunately, it works well with the exception of a few participants/runs. 
The error message is:
Running job #1
Running 'Volume of Interest'
Warning: Empty region.
> In spm_regions (line 155)
In spm_run_voi (line 63)
In cfg_run_cm (line 29)
In cfg_util>local_runcj (line 1688)
In cfg_util (line 959)
In spm_jobman>fill_run_job (line 458)
In spm_jobman (line 247)
In spDCM_fun_test (line 207)
Failed 'Volume of Interest'
Reference to non-existent field 'v'.
In file ".../spm12/config/spm_run_voi.m" (v6301), function "spm_run_voi" at line 76.
The following modules did not run:
Failed: Volume of Interest
Error using MATLABbatch system
Job execution failed. The full log of this run can be found in MATLAB command
window, starting with the lines (look for the line showing the exact #job as
displayed in this error message)
Running job #1
Is there anything that I should do to fix this? 
On a different note, I had to modify the code in DCM specification:
Sess = SPM.Sess(xY(1).Sess);
as this was causing errors as xY(1).Sess is 2 at points, and not 1. Therefore, I changed it to:
Sess = SPM.Sess(1);
Is it going to be problematic? And, if yes, is there any other way of fixing the issue?
I have three "for" loops for the dataset, one for the subjects, one for the sessions, and the last one for the runs... I managed to get the DCM.mat for 16 subjects out of 26 (and then it stops for the error I get in the last section). Then I am just taking the average of the DCM.Ep.A for each subject in each session (which I am not sure is the right thing to do - the alternative method can be doing a BMS RFX to find out which model is best and just use the DCM.Ep.A for that model). The data includes samples taken from 26 subjects in two sessions and three runs.
Unfortunately, the RFX BMS does not seem to be working outside the GUI (when I change the values) either. The script generated by the batch seems to be like (I have changed it slightly):
clear matlabbatch;
mkdir([session_folder_name '/func/BMS']);
matlabbatch{1} inference.dir = ...
            cellstr([session_folder_name '/func/BMS']);
matlabbatch{1} inference.sess_dcm{1}.dcmmat{ 1,1} = ...
            cellstr([session_folder_name '/func/Run01/GLM/DCM_DMN.mat'] );
matlabbatch{1} inference.sess_dcm{1}.dcmmat{ 2,1} = ...
            cellstr([session_folder_name '/func/Run02/GLM/DCM_DMN.mat'] );
matlabbatch{1} inference.sess_dcm{1}.dcmmat{ 3,1} = ...
            cellstr([session_folder_name '/func/Run03/GLM/DCM_DMN.mat'] );
matlabbatch{1} inference.model_sp = {''};
matlabbatch{1} inference.load_f = {''};
matlabbatch{1} inference.method = 'RFX';
matlabbatch{1} inference.family_level.family_ file = {''};
matlabbatch{1} inference.bma.bma_no = 0;
matlabbatch{1} inference.verify_id = 0;
However, this gives me an error as well (The Model seems to be empty and I cannot fix it, even though all the variables and the structure in the script above seem to be fine).
On a different note, I was wondering whether there is a way of finding out which model is the best one, other than just looking at the graphs generated using the GUI - I realised that this is a question some of the other SPM users in the list have as well. If not, it seems like I have to run the script for each session and each subject and look at the graphs... which does not seem reasonable. If there is a quantitative way of finding the best model for each session in BMS, then I can easily select the DCM.Ep.A for that model and use it in my further analysis. Also, what if the best model is different for each subject, e.g., model 3 for subject 1, model 2 for subject 8? Can they then be used in the analysis?
Many thanks and apologies for the very long email with millions of different questions!
Thanking you in advance,


Cited By


Projects (2)
Developing a technique for accurate diagnosis of PD using EAs on rs-fMRI data.
Developing a technique for accurate monitoring of PD using EAs on rs-fMRI data for participants prescribed a drug.