Sites Inria

Version française

Computer sciences


Nicholas Ayache's opening lecture at the Collège de France

© Inria - Asclepios

Nicholas Ayache will give his inaugural lecture at the Collège de France on Thursday, 10 April 2014, at 6 pm. The lecture is entitled “From Medical Images to the Digital Patient”.

Nicholas Ayache is the newly appointed professor on the Chair in Informatics and Computational Sciences at the Collège de France. His course is entitled "The Personalised Digital Patient: Images, Medicine, and Informatics" . He will use the opportunity to present some of the most advanced research in computational medical imaging.

The goal of this fast-growing field of research, at the crossroads of informatics, computational sciences and medicine, is to design and develop software algorithms to process medical images to assist doctors in their clinical practice. Such algorithms aim, in particular, to enhance diagnosis by extracting objective and clinically useful information from medical images. It also seeks to assist therapeutic practice with planning and simulation applied to a digital model of the patient.

Rapid growth of medical imaging

Divergence of Registration-based Flux in Alzheimer's Disease: this color map demonstrates the regions undergoing volume changes with Alzheimer's disease evolution. - © Inria - Asclepios

Medical images are now ubiquitous in everyday and hospital-based medical practice. In addition to x-rays, four major imaging methods are commonly used: scanner, MRI, ultrasound, or scintigraphy1. The images produced by these four methods are volume-related: at each point of the human body they provide information in a small element of volume called voxel , the volume extension of the pixel .

There are other existing methods human body imaging and new techniques regularly emerge. Elastography2, to name one, which enables the elasticity of tissue to be measured from an MRI or ultrasound, or endomicroscopy, which enables the microscopic architecture of cells to be viewed at the end of optical fibres.

Most medical images contain a vast volume of data. The anatomical image of an organ, or even whole body may contain between a few million and several hundred million voxels, stored in huge 3-D matrices of numbers. The quantity of information grows rapidly when several images are acquired on the same patient to take advantage of the complementarity of different methods, or to monitor change over time; they then become 4-D images with three spatial dimensions and one temporal dimension.

As if this flood of images were not enough, large databases of images are gradually becoming available on the Web. These images are often accompanied by metadata on the patient's medical history and pathology.

The role of Informatics and computational sciences

Computational 4 chambers mesh of the heart for electromechanical simulations. - © Inria - Asclepios

In light of all these images and their complexity, doctors are often only able to visually retrieve incomplete and qualitative information. Large volume images are often only viewed in the form of 2-D cross-sections. It then becomes almost impossible to accurately quantify the volume of a tumour, to detect an isolated anomaly in an entire organ and follow its subtle development between two exams, or to quantify the movement of a dynamic body such as the heart in a sequence of images. It is even more difficult to plan a delicate operation without the computer's help.

Informatics and computational sciences play a crucial role in making thorough and optimal use of this overabundance of information. They are essential for the analysis of reconstructed images, with the purpose of objectively extracting clinically relevant information and presenting it in a unified and intuitive manner to the doctor. They also provide the chance to build a computational paradigm of the patient for simulation : simulating the development of a pathology or the effect of therapy for example, or medical or surgical procedures for practitioner training.

Cardiac fibre orientations measured in vivo with diffusion tensor MRI. - © Inria - Asclepios

Computational analysis and simulation of medical images rely on algorithms that take the specificity of anatomy and human physiology into account using mathematical, biological, physical or chemical models, adapted to the resolution of images. These models of the human body themselves depend on parameters that allow modification of form and function of the simulated organs. Used with a standard set of parameters, the models are generic : they describe and simulate the average shape and function of organs in a population. But with the medical images and all of the data available on a specific patient , the configuration of a generic model can be adjusted through algorithms to more accurately reproduce the shape and functioning of this individual's organs. This then gives a personalised model.

The personalised digital patient and computational medicine

Fibre tracking using diffusion tensor imaging of the brain, in order to reveal brain connectivity. - © Inria - Asclepios

The personalised digital patient 3  is none other than this set of digital data and algorithms that help reproduce, on various scales, the shape and dynamic functioning of the main tissue and organs of a given patient. It is also the unified framework allowing information from the patient's anatomical and functional images to be integrated, as well as information describing the singular medical history of the patient and his/her illness.

Computational and personalised paradigms are designed to assist doctors in their medical practice: they help diagnosis by quantifying this information in the images; as well as prognosis by simulating the change in a pathology; and therapy by planning, simulating and controlling an operation. This is what computational medicine foreshadows in the future, a computerised component of medicine not intended to replace doctors, but rather to provide them with digital tools to assist in their practice of medicine for the benefit of patients.

From medical images to the digital patient

Variability of the brain sulci measured from 98 healthy brains, red areas represent regions of large variability. - © Inria - Asclepios

In his opening lecture, entitled "From medical images to the digital patient", Nicholas Ayache has chosen four examples that demonstrate the progress made in algorithms and models used with medical images. The first two examples, morphometry and endomicroscopy , fall within the scope of computational anatomy . The algorithms used are based on geometric, statistical, and semantic models of the human body. The next two examples, oncology and cardiology , involve computational physiology . Their algorithms also rely on biological, physical or chemical models of the human body.


Mesh of the liver for laparoscopic surgery simulation. - © Inria - Asclepios

Computational medical imaging, at the crossroads of informatics and medical imaging, provides new digital tools at the service of doctors and patients, in the broader framework of computational medicine.

Ongoing progress in these areas allows a glimpse of how information technology and computational sciences can support the transition from a standardised and reactive medicine to a more personalised, preventive and predictive medicine4. They are based to a large extent on algorithmic progress in image processing and in the computational modelling of the anatomy and physiology of the human body.

Upcoming courses, as well as seminars and the closing symposium will delve more deeply into algorithmic, mathematical and biophysical principles in this fast-growing research field, while illustrating its multi-disciplinary character and the latest progress. There will be scientists and doctors of various specialities, at the digital patient's bedside.

1. Scanning or computed axial tomography by X-ray; MRI or magnetic resonance imaging; ultrasonography or ultrasound imaging; scintigraphy: PET (positron emission tomography) and SPECT (single photon emission computed tomography).

2. M. Fink, Renversement du temps, ondes et innovation ; opening lecture, Collège de France, 2009.

3. The term virtual patient is also used, particularly in the context of simulating medical operations and surgery. Of course, the terms virtual patient and hypochondriac should not be confused.

4. Elias Zerhouni, Grandes tendances de l'innovation biomédicale au XXIe siècle ; opening lecture, Collège de France, 2011.

Keywords: Nicholas Ayache Patient numérique Chaire informatique et sciences numériques Collège de France Imagerie médicale