modifying the ester or ether chain length leads to optimized membrane characteristics. Adverse events are mainly seen, when complement activation is twinned with the presence of endotoxins from bacterial cell walls. Here, cytokines are released in a synergistic manner from white blood cells (cell- activation), which may lead to inflammatory reactions. Hypersensitivity reactions. Allergic reactions are frequently observed in those patients who have been in contact with resi- duals of the sterilizing agent ethylene-oxide (ETO). ETO is bound spontaneously to albumin. ETO/Albumin provoke the formation of IgE-antibodies which are responsible for the majority of seen hypersenitivity reactions. Biometarials, such as polyurethane (PUR) or dialysis membrane polymers from PMMA store ETO in their bulkstructure and show a slow release pattern. Further, polymer extractables eluted by blood may also induce hypersensivity reactions. Clinical conse- quences in chronically treated patients, such as in hemodialysis have to be seriously considered. Hemodynamic effects. Surfaces bearing negative charges of a defined density stimulate the “contact phase” of coagulation. As a consequence, the vasodilator bradykinin is formed and— if not degraded by angiotensin-converting enzyme (ACE)— severe blood pressure drops are observed,. This reaction is amplified in those patients, who are treated by ACE- inhibitors. Some negatively-charged polymers exhibit these effects and have exerted fatal incidences [4,5]. Interaction with administered medicinal drugs. Several bio- compatibility pattern of dialysis may also be influenced by the simultaneous administration of drugs, e.g. the use of ACE- inhibitors amplifies the effect of bradykinin-formation. Aspirin may reduce the formation of thromboxane initiated by dialysis membranes and vitamins may interfere with the release of tumor necrosis factor (TNF). As a consequence, membrane polymers should be checked for the capacity to interact with old and newly developed medicinal drugs. Conclusion. – It is possible to adapt the polymer composi- tion of modern dialysis membranes such that adverse patient reactions can be excluded or at least be minimized. The basis is a better understanding of the underlying mechanisms of blood material interaction. Further smart geometries and design of dialysis membranes are the strategies for further develop- ment of biomaterials for dialysis membranes [6,7]. References [1] ANSI/AAMI/ISO-Standard 10993-4. Biological Evaluation of Medical Devices. 1993. [2] Consensus Conference on Biocompatibility. Nephrol Dial Transplant 1994;9(Suppl 2). [3] Diamantoglou M, Vienken J. Strategies for the development of biocompa- tible dialysis membranes. Macromol Symp 1996;103:31–42. [4] Tielemanns C, et al. Anaphylactoid reactions during hemodialysis on AN69 membranes in patients receiving ACE inhibitors. Kidney Int 1990; 38:982–8. [5] Kammerl, et al. Extracorporeal therapy with AN69 membranes in combi- nation with ACE-inhibition causing severe anaphylactoid reactions: still a current problem? Clin Nephrol 2000;53:486. [6] Vienken J. Polymers in dialysis: characteristics and needs. Medical Device Technology 2001;12:18–27. [7] Vienken J, Bowry S. Quo vadis dialysis membrane? Artif Organs 2002; 26:152–9. Medical devices 1 Wearable system for post stroke rehabilitation of the upper-limb F. Lo Russi a , A. Tognetti a , T. Giorgino b , S. Quaglini b a Information Engineering Department, University of Pisa, via Caruso 2, Pisa, Italy b Laboratory for Medical Informatics, Dipartimento di Informatica e Sistemistica, DIS, University of Pavia, Pisa, Italy Aims. – Conductive elastomers (CE) are a novel strain- sensing technology which can be embedded unobtrusively into an elastic garment’s fabric. Elasticity of the garment leads to strong cross talk phenomena between sensors. In order to address this problem, in this paper an analysis of the attributes (attribute selection), by computing the information gain (IG) is performed. A prototype was realized to simulta- neously measure the strains at multiple points of a shirt cover- ing the thorax and upper limb. This work describes preliminary experiments with machine learning techniques, employed to analyze the strain measures in order to reliably reconstruct upper-limb posture. The aim of the application is to detect execution, correctness and progress of physical exercises per- formed as part of neurological rehabilitation therapy. Methods. – When CE sensors are deposited on a garment as stripes, the impedance of each deposited segment varies as a function of the strain to which it is subject. Advantages of CE sensors over solid-state and other types is their negligible weight and thickness, and the fact that any number of “measur- ing points” can be let on a garment in a single setup. The gar- ment is able to detect body postures taken by patients while performing prescribed exercises for physical rehabilitation after discharge by the rehab clinic. After the garment is worn a subject is asked to perform one of several rehabilitation exer- cises foreseen in the rehabilitation protocol. This is done in order to “snapshot” the current sensor readings. The values are stored and later used as the training set to build a posture- recognition model. The aim of the final system is to discrimi- nate not only the progress along the correct execution, but also warn in case one of the improper positions is attained. In this simplified setup, therefore, we are dealing a supervised classi- fication task, with 19 numeric attributes (sensors). Class labels, or target postures, will be assigned among a set which includes n intermediate steps taken in the “correct” exercise execution path, and m “incorrect” positions. For tests to be described in the following we have preliminarily fixed N = 4 and m = 3. Results. – Before testing actual classification schemes, a preliminary exploration on attributes (attribute selection), by computing the IG of each single attribute, has been performed. Abstracts / Biomedicine & Pharmacotherapy 60 (2006) 468–479 473