A HYBRID ANALOG AND DIGITAL VLSI NEURAL-NETWORK FOR INTRACARDIAC MORPHOLOGY CLASSIFICATION

Current Implantable Cardioverter Defibrillators (ICD's) use timing bas ed decision trees for cardiac arrhythmia classification, Timing alone does not distinguish all rhythms for all patients. Hence, more computa tionally intensive morphology analysis is required for complete diagno sis. An analog VLSI neural network has been designed and tested to per form cardiac morphology classification tasks. Analog techniques were c hosen to meet the strict power and area requirements of the implantabl e system while incurring the design difficulties of noise, drift and o ffsets inherent in analog approaches. The robustness of the neural net work architecture however, to a large extent, overcomes these inherent shortcomings of the analog approach. The network is a 10:6:3 multilay er perceptron with on chip digital weight storage. The chip also inclu des a bucket brigade input to feed the Intracardiac Electrogram (ICEG) to the network and a Winner Take All circuit for converting classific ations to a binary representation. The training system trained the net work in loop and included a commercial implantable defibrillator in th e signal processing path. The system has successfully distinguished tw o arrhythmia classes on a morphological basis for seven different pati ents with an average of 95% true positive and 97% true negative detect ions for the dangerous rhythm. The chip was implemented in 1.2 mu m CM OS and consumes less than 200 nW maximum average power from a 3 V supp ly in an area of 2.2 x 2.2 mm(2).