The objective of our project is to develop a walking simulation software which helps the surgeon and the patient to make a decision.
It has been developped by Brice Prosperi, Mélissa Daudé, Emeric Jeandupeux, Guilhem Bavre, and Jerôme Bouchard.
Cerebral palsy is a very complicated disease and what can cause that sickness is very difficult to identify. Moreover a very heavy operation is required to treat it.
That's why we develop our software in order to ba able to suggest the good treatement and to display a customized 3D model corresponding to the gait of the patient. That last part of our work is the reason why our project is named AVATAR.
This project is leaded with the collaboration of Eric Dasailly from the Poidatz institute and Vinvent Vigneron who is a lecturer and researcher in the Evry Val d'Essonne University. We were given a database containing manny datas about previous medicine cases.
There were different steps during the development of our software. First, we analyzed the database in order to understand the elements which were important of us and on which we could work. The results of our analyze were that the database includes 41 patients and each of them has 459 kinematic datas to describe his gait cycle. Moreover, the database also contains the different surgeries received by these patients. There are 6 different surgeries and as many combinaison as it is possible to make with these 6 surgeries.
Secondly, we worked on extracting the datas. This step is very important because it simplifies a lot the work on the data. We used the Clustering algorithm to determinate the 10 most important values in the gait cycle for every patient. We also noticed that only 11 surgical combinaison were important.
Then, our work was to be able to make a surgical forecast with two different method : the kmean algorithm which is able to predict the closest patient from any patient in the database. The second method is the neuron network which is able to create a relationship between the patients and the surgeries. There is a learning process and then the neural network can suggest a surgery.
Lastly, we worked on the 3D model which uses the kinematic vector containing 459 values and describing the gait cycle of a patient in order to model the gait after having the operation.
If we summarize, this is our works our software: When we analyze a new patient, we search for the closest patient in the database and we give the surgeries that he received to the surgeon. Then, we obtain his kinematic vector after having the operation and we display the 3D model of that patient after the operation.
So the software reminds to the surgeon what a similar patient from the one he is examining. Then he can show to the patient how he could walk after having the operation.