Virtual Cochlea Exploration

Virtual exploration of the cochlea: anatomical variations across individuals

Luca Ferrarini
Julien Milles


In November 2003, we started developing an autonomous virtual mobile robot for the exploration of 3-dimensional medical images of tubular structures. The robot is equipped with a virtual camera and virtual sensors: an artificial intelligence module processes feedbacks from the sensors in order to keep a central position through the tubular structure during the navigation. Moreover, the information from the sensors is used to estimate anatomical properties such as the diameter of the structure. Finally, feedbacks from the frontal sensors are used to automatically detect bifurcations, at which points the robot could clone itself to continue to the exploration in parallel along both ways.

During the last years, we have applied the virtual robot to the analysis of several structures, such as colon, bronchi, carotid arteries, and cochlea. This latter application provided the most promising and interesting results.

The cochlea is part of the inner ear. Filled with watery liquid, it is responsible to translate the vibration coming from the middle ear into electrical impulses to the brain. Knowing the anatomy of the cochlea for a given subject is extremely important for cochlear implants. The tubular structure of the cochlea is ideal for virtual robot exploration. Along its journey through the cochlea, the robot can assess the local diameter of the structure, as well as determining its central path. Common imaging techniques, such as MR and CT, are capable of visualizing the entire canal, without though further differentiating between the different scale (sub-structures) composing the cochlea. Post-mortem micro-CT, on the other hand, can provide more detailed information: in such images, the virtual robot can separately explore the different scale, reporting information on their diameter and central path. The combination of such information allows us to establish an anatomical relation between the diameters visible in clinical set up with the diameters of the scala in which cochlear implants must be placed.

Major results

Applying the virtual robot to micro-CT and CT cochlea images, we achieved the following results:

1.      Assessment of local cochlear diameter both for the entire canal (in micro-CT and CT), and scala timpani and scala vestibule (in micro-CT only).

2.      Assessment of the relation between the cochlea canal visible in clinical set up (pre- and post-operative CT) and the scala in which the electrodes are placed.

3.      Finding of local anatomical change along the modiolus axis. This finding promises to have important consequences in cochlear implants development and surgery. It was known from literature that cochlear implants could go wrong in a specific location along the cochlea, where the electrodes could break the membrane separating the two scale, ending up in the wrong scala. However, a convincing explanation for this pattern was missing. The anatomical change in the cochlea detected by the virtual robot explains now why some surgical operation go wrong: electrode implants are usually developed to be flexible on one plane (in order to follow the spiral form of the cochlea), but quite rigid on the other plane (along the modiolus axis): thus, having a dip along the axis could cause the breaking of the membrane.




Figure 1a 

Figure 1b 



Fig. 1 (a) Local coordinate system of the mobile robot and steering vector: the desired direction is fully described by two angels in the local system. (b) Sensory system of the AVMR: frontal sensors are described by two angles, g and d; lateral sensors are described by an angle a and a relative distance dr (b and dt for top and down sensors).

Figure 2
Fig. 2 (top) General scheme for the NFC: the input variables are fuzzyfied before being fed to the RNN neural network. The fuzzy output membership functions are defuzzyfied by the ORNN neural network. (bottom) Membership functions for position (a), orientation (b), and desired output angle (c).

Figure 3
Fig. 3 Central lines detected in medical datasets. From left to right: cochlea, carotid artery, colon, and two internal views of the colon obtained with the virtual camera.

Figure 4
Fig. 4 (Left) Volume rendering of a carotid artery (CT scan); (Right) Local radii estimated by the AVMR.

Figure 5 
Fig 5. (A) Given the subsampled micro-CT dataset, a 3-D region growing is applied to segment the cochlear canal. (B) A local coordinate system is given. (C) The skeleton presents several side branches because of the surface irregularities. (D) In distance map, each voxel is associated with its distance to the closest surface points. (E) A wave-prop applied to the skeleton and modulated by the distance map provides the final central path. (Otology & Neurotology, 2009)


For further information, please contact:
Dr. JR Milles, PhD
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 5342
Fax. +31 (0)71 526 6801


Journal/Conference Papers

  • BM Verbist, L Ferrarini, JJ Briaire, F van Poucke, F Admiraal-Behloul, H Olofsen, JHC Reiber, JHM Frijns, Anatomical considerations of cochlear morphology and its implications for insertion trauma in cochlear implant surgery, Otology and Neurotology, 30(4):471-477, 2009
  • Ferrarini L, Verbist B.M., Olofsen H, Vanpouke F., Frijns J.H.M., Reiber J.H.C., Admiraal-Behloul F., "Autonomous Virtual Robot for 3-Dimensional Medical Image Exploration: Application to Micro-CT Cochlear Images", Artificial Intelligence In Medicine. 43(1):1-15, 2008
  • Luca Ferrarini, Hans Olofsen, Johan H.C. Reiber, and Faiza Admiraal-Behloul, "A NeuroFuzzy Controller for 3D Virtual Centered Navigation in Medical Images of Tubular Structures", International Conference on Artificial Neural Networks - ICANN 2005, LNCS 3697, pp 371-376, Warsaw - Poland
  • Ferrarini Luca, Hans Olofsen, Johan H.C. Reiber, Faiza Admiraal-Behloul, "Exploration of 3D Medical Images of Tubular Structures: A NeuroFuzzy Controller for Virtual Centered Navigation", ASCI Conference 2005, Heijen The Netherlands
  • Faiza Admiraal-Behloul, Boudewijn .P.F. Lelieveldt, Luca Ferrarini, Hans Olofsen, Rob van der Geest and J.H.C Reiber, "A Virtual Exploring Mobile Robot for Left Ventricle Contour Tracking", International Joint Conference on Neural Networks 2004 (IJCNN 2004), Budapest – Hungary

Book chapters

  • L. Ferrarini, H. Olofsen, J.H.C. Reiber, F. Admiraal-Behloul, "Autonomous Virtual Mobile Robot for the Exploration of 3D Medical Images", In Medical Robots - Advanced Robotic Systems (ISBN 978-3-902613-18-9)

Conference abstracts

  • B.M. Verbist, L. Ferrarini, J.J. Briaire, A. Zarowski, J.H.M. Frijns, "New insights in insertion trauma during cochlear implant surgery based on 3-dimensional image exploration of the cochlea", ESHNR, October 2009, Verona, Italy.
  • B.M. Verbist, J.J. Briaire, L. Ferrarini, R. Joemai, J. Snel-Bongers, J.H.M. Frijns, "Pre- and postoperative assessment of cochlear implantees by means of multidetector row computed tomography (MD-CT)", Conference on implantable auditory prostheses, July 12-17 2009, Lake Tahoe, California (U.S.A.)