Semi-Automatic Manufacturing of Customized Hearing Aids Using a Feature Driven Rule-based Framework K. Sickel 1 , S. Baloch 2 , V. Bubnik 2 , R. Melkisetoglu 2 , S. Azernikov 2 , T. Fang 2 , J. Hornegger 1 1 Pattern Recognition Lab, Department of Computer Science, Friedrich-Alexander-University Erlangen-Nuremberg, Martensstr. 3, 91058 Erlangen, Germany 2 Siemens Corporate Research, 755 College Road East, Princeton, NJ 08540, USA Email: konrad.sickel@informatik.uni-erlangen.de Abstract We present a semi-automatic solution to the prob- lem of customized 3-D shape modeling using a rule based system. The solution is illustrated with the use case of hearing aid manufacturing. The cur- rent process of manufacturing customized hearing aids consists of many manual steps, all of which require expert knowledge. The idea is to translate this expert knowledge into machine readable rules using a procedural knowledge representation. To- gether with a set of anatomical features detected on the ear impression, the necessary surface shaping operations are defined and applied. The advantage of this approach is its high flexibility and customis- ability. It yields better consistency, efficiency and reproducibility of the resulting device. We validate the approach in a real production environment. 1 Introduction This paper addresses the problem of modifying shapes in a rapid prototyping environment. The goal is to transform a raw 3-D impression U of the ear canal (Figure 1) into the shape D = T (U ) of a hearing aid (HA), see Figure 2 for examples. The applied transform T is constrained in numer- ous ways. The first constraint is the preservation of anatomical features to ensure a good fit in the ear of a patient. Second, the impression must main- tain enough space inside for the placement of elec- tronic components, such as receiver and faceplate. For wearing comfort, a ventilation tube (vent) has to be integrated. Furthermore the form of the target device type has to be specified. Hence, the current so-called detailing and modeling is a challenging, Figure 1: Image of a 3-D raw ear impression. tedious and time consuming manual process. The complexity of the problem is further increased by the high variability of ear shapes, which has pre- vented automation of the process so far. Special modeling software [9] provides a workspace to an experienced operator to perform a sequence of oper- ations, such as cutting and rounding. In addition, it offers tools to place the electronic components and the vent. This virtual placement of components en- sures that the device is built with electronics suit- able for the patient’s needs of amplification. The size of the electronic components, in turn, is related to their amplification power. Consequently, patients with major hearing loss may only have larger de- vices than patients with minor hearing loss. Recent works on shape modification for hearing aid manufacturing concentrate on the active shape model [1] approach. The active shape model allows VMV 2009 M. Magnor, B. Rosenhahn, H. Theisel (Editors)