Subscribe via RSS Feed

Research and Expertise

Accomplished scientist and project leader with more than ten years of experience in the fields of high energy physics, numerical simulation and medical physics. Strong problem solving abilities and data analysis skills including statistical analysis and error propagation. Sophisticated computer expertise and proven ability in C++ programming. Excellent communication skills including management of a group, participation within large scientific collaborations and student supervision. Extensive experience with scientific writing and presentations to wide international audiences.

PANACEE project (Institut Curie, LITO) – Orsay, France (since December 2020)

Develop methods & tools to identify a small group of patients with non small cell lung cancer and similar clinical and radiomic characteristics

  • PANACEE: PANomic Atlas for non-small CEll lung cancer managEment.
  • Segmentation of PET/CT images and extraction of radiomic features.
  • Prediction models (type of cancer and/or response to treatment).
  • Development of Artificial Intelligence tools for oncology (lung and breast cancer).

Mathematical Oncology global project (CRCM, SMARTc) – Marseille, France (October 2018 – October 2020)

Mathematical modeling for oncology treatments optimization

  • Web application (soon to be launched) that checks drug-drug (pharmaceuticals and/or chemotherapies) compatibility when using Y-shaped perfusion.
  • Haemato toxicity modelisation using machine learning techniques (patients treated for rhabdomyosarcoma).
  • Mathematical modelisation of metronomic chemotherapy coupled with checkpoint inhibitors (immunotherapy).
  • Personalised treatments for patients with pancreatic adenocarcinoma using omics, allowing to predict the clinical course and chemo-sensitivity of this disease.

Development of predictive algorithms in the field of vehicle valuation – autobiz – Suresnes, France (August 2017 – August 2018)

Machine learning techniques for predictive data analytics

  • Vehicle valuation prediction (forecasting) using the Auto Regressive Integrated Moving Average model (statsmodels, ARIMA).
  • Impact of geographical regions on the vehicle valuation (matplotlib basemap, shapefiles, python-visualisation/folium).
  • Vehicle cost of repair modelisation using regression analysis for predictive modelling (scikit-learn) – in production mode.
  • Computation done through Amazon Web Services (storage, EC2 instances).
  • Big data processed using Apache-spark framework.

Simulation of nanoparticle-mediated hyperthermal therapy – CEA/I2BM/SHFJ – Orsay, France (June 2015 – Dec. 2016)

Monte-Carlo simulations of hyperthermia

A new approach in cancer treatment is the use of nanoparticles for the induction of intracellular hyperthermia. Near infrared absorbing nanoparticles enhance photothermal therapy of tumors. Computational modeling is an important tool for investigating and optimizing parameters such as nanoparticle size and shape, excitation wavelength and tissue properties.

    • We extended GATE (Geant4 based simulation code) to model nanoparticle mediated hyperthermal therapy.
    • We validated the diffusion of heat with experimental data (collaboration with the IMNC in Orsay).

    PET/CT quantification in lung imaging – INM-UCL – London, United-Kingdom (2013 – 2015)

    Quantification of pulmonary PET/CT data

    Idiopathic Pulmonary Fibrosis (IPF) is an interstitial lung disease characterized by an increase in the quantity of extra cellular matrix and the destruction of parenchyma structure. There is no effective treatment. There has been an increased interest in imaging pulmonary disorders using Positron Emission Tomography (PET) techniques.

      • We studied the PET radiotracer uptake in lung regions exhibiting obvious fibrosis. This may provide a disease progression biomarker which is sensitive over the duration of the clinical trial.
      • Air content in the lung influences reconstructed PET images. It is important to correct for this effect, called ‘tissue fraction effect’ (TFE). We extended the TFE to include the blood component of the lung tissue.
      • Respiratory gating (gate the data into different motion states) of lung PET/CT data is important for motion correction for example. We performed a data driven gating of cine-CT data using the Principal Component Analysis.
      • Lung density changes during respiration. We investigated the effect of gated PET/CT image registration associated with the breathing cycle. We looked at the relationship between density changes measured in CT, tracer concentration changes measured in PET and local volume changes obtained from the Jacobian of the registration deformation field.

      hGATE project – CEA/I2BM/SHFJ – Orsay, France (2011 – 2013)

      Monte-Carlo simulations of optical imaging

      GATE is an advanced open-source software dedicated to Monte-Carlo simulations of pre-clinical and clinical scans acquired in emission tomography, transmission tomography and radiotherapy therapy. Optical imaging is currently drawing growing interest for molecular imaging, as a non-invasive, efficient and low-cost imaging technique allowing for real time study of biological processes.

        • We extended GATE to support simulations of optical imaging experiments such as bioluminescence or fluorescence and validated it against the Monte-Carlo for Multi-Layered media (MCML) simulation tool. We have shown an excellent agreement between GATE and MCML.
        • Monte-Carlo simulations of optical imaging are computationally demanding. Recently, Graphics Processing Units (GPU) became a cost-effective solution to access to high power computation. We extended GATE to support optical imaging simulations using GPU architectures.

        Compact Muon Solenoid experiment at CERN (2006 – 2010)


        Top quark physics

        • Combination of Top Pair Production Cross Sections in proton-proton Collisions at 7TeV and Comparisons with Theory.
        • Measurement of the tt production cross section at 7TeV data by looking for semileptonic decays of the Top-quark pairs in the electron channel with reliance on b-tagging (b-quark jet identification technique).
        • Estimate of the background from light quarks or gluon jets that could be tagged as heavy flavor jets, using negative tags (when the flight distance of the reconstructed secondary vertex is negative)

        Pixel Simulation


        D0 experiment at Fermilab (2006 – 2008)

        Electroweak sector and searches for new phenomena

        • New physics predicting particles that decay into (could be observed as a bump in the Zγ mass spectrum)
        • Search for a scalar or vector particle decaying into Zγ in proton anti-proton collisions at center of mass energy of 1.96TeV
        • 95% Confidence Level limits on cross section times branching ratio into Zγ for a scalar/vector particle that decays into Zγ as function of the scalar/vector resonance mass


        Phenomenology (2000 – 2004)