Assisted System for Decision-Making in Parkinson Disease Surgery
Parkinson’s disease is a degenerative disease of the Central Nervous System (CNS) characterized by a progressive deterioration of motor functions, with rigidity, hypokinesia, bradykinesia, tremor and alteration of postural reflexes. This disease progresses inexorably towards disability and death, the rhythm of that progression being unpredictable.
Although the initial treatment of Parkinson’s disease is pharmacological, in some cases adequate control of symptoms is not achieved despite the various combinations of currently available medications. In other cases, the medication causes disabling side effects, such as the appearance of abnormal movements (dyskinesias) or intolerance. These two situations appear more frequently as the treatment time increases. In these patients, a surgical
intervention aimed at controlling Parkinson’s symptoms is indicated.
The most used surgical techniques are aimed at suppressing the activity of nuclei that remain hyperactive (Vim, GPi or STN). For a certain time the suppression was achieved by coagulating said nuclei, but recently it has been demonstrated that its chronic stimulation with high frequency alternating current is equivalent, with the advantages that this procedure is associated with fewer complications and side effects, it is reversible, and its
effect is more easily modulated.
The method currently used in almost all centers with experience is basically:
1. Locate the initial target point from standard brain atlases
2. Record a path of about 15 to 20 mm with microelectrodes
3. Confirm the structures located by this procedure with sensory stimulation (driving) or macro stimulation and 4, choose a new target point from the data
obtained.
This project aims to design a system to simplify surgical intervention and allow its dissemination to a greater number of patients.
The philosophy of this system is based on three fundamental points:
- The use of deformable brain atlases. They consist in the adaptation of a deformable atlas to the concrete anatomy of the patient, choosing homologous points in the atlas and in the MR image.
- The use of automated recognition systems for microregistration patterns. There are algorithms that allow teaching a computer to recognize patterns and classify them.
- The integration of the neurophysiological information of the microregisters as anatomical data in the deformable atlases.
• Preparation of a micro-records database, duly labeled.
• Development of a deformable atlas, which makes an adjustment to the desired point or structure starting from two point clouds: one in the atlas and another in the patient’s brain.
To obtain this objective, the composition of two applications (f ● g) is used. The first (f) is an affine transformation, while the second (g) uses five different approaches. In the first of these approximations, the application g is defined by a displacement weighted by the inverse of the relative distance to the reference points. In the second, to obtain the application g a similar idea is used, but in this case a system of equations must be solved. In the latter approaches, the g application uses offsets that conserve approximately the volume using three different vector fields.
• Characterization of the activity of the different regions by time-frequency techniques of deep intracranial records. Parameterization from the spectrogram itself and from the Choi-Williams distribution. Evaluation of the parameters selected in the classification using ROC curves. The classification by means of neural network (MLP) on the parameters of the transformation T-F obtains, in the majority of patients, percentages greater than 94% in the
identification of the studied regions.
• The integration of the neurophysiological information of the microregisters is not achieved as anatomical data in the deformable atlases.