AIPD-DC11-UniPi

• Host Institution: University of Pisa
• PhD Enrolment: University of Pisa
• Start Date: October 2025
  
• Duration: 36 months
• Official PhD Supervisor: Mirco Cosottini

Research Objectives

1. To expedite the early diagnosis of PD;
2. To improve the differential diagnosis among degenerative parkinsonism with overlapping clinical presentation;
3. To predict the forthcoming pheno-conversion in patients at risk of developing synucleinopathies.

In more detail:

A) 7T MRI: identify radiological markers which significantly differ between PD patients, iRBD patients and control subjects, and to explore possible markers able to predict the time to phenoconversion in iRBD patients. This approach might enable the identification of iRBD patients who are going to develop motor symptoms. The secondary aim is to explore the differences between PD and MSA patients, other than cerebellar and putaminal atrophy, to increase the number of features able to differentiate the two diseases.

B) 3T MRI: we aim to identify radiological markers of clinical outcome in parkinsonian patients. We will focus on patients with parkinsonism but without a definite diagnosis at the time of the MRI examination, searching for markers that can predict the type of parkinsonism that was diagnosed at follow-up appointments. We will initially focus on the NTUA and PPMI datasets. PPMI provides access to more than 1700 brain MR images (>900 PD, ~ 600 prodromal, >200 controls). We will implement recent AI techniques and compare their discriminative capability with results already reported in literature using the same datasets.

Particular attention will be given to XAI techniques (e.g. saliency maps) to detect the most promising regions of interest to be addressed in the 7T data analyses. 7T MRI data: We will use an MRI dataset already acquired in patients with PD, iRBD and MSA and in control subjects, including iron-sensitive and whole brain T1-weighted images. 3T MRI data: We will use an MRI dataset already acquired in patients with PD, MSA and PSP, and in parkinsonian patients without a definite diagnosis at the time of the MRI examination. Data available include T1-weighted, T2-weighted and iron-sensitive images. The research will be mainly focused on the midbrain, as the substantia nigra is the first brain site to be affected by pathology, but all basal nuclei will be explored to improve the diagnostic accuracy in differentiating between parkinsonisms. Additional regions of interest may emerge during the preliminary analyses performed on the public datasets.

Expected Results

• 7T MRI: Radiological markers differentiating between PD, iRBD and control subjects, and between PD and MSA.
  Furthermore, we expect to find a marker able to predict the forthcoming motor conversion in iRBD patients.
• 3T MRI: Radiological markers differentiating between PD, MSA and PSP. Additionally, we expect to find radiological markers able to predict the clinical evolution of patients without a definite diagnosis at the time of the MRI examination (i.e., the kind of parkinsonism that has been diagnosed in the follow-up).
• Novel AI/ML algorithms.
   

Planned Secondment(s)

• Host: GE Healthcare
  • Duration: 24 months
    
  • Purpose: Learning about AI/ML applied to MRI

This project is part of the "Digital Health" work package.