An adaptive reconfigurable hybrid bioinspired classifier and its hardware implementation for efficient medical image diagnosis

Abstract

Many real-time applications such as computer vision, optical newlineimaging, healthcare e-diagnosis, smart vehicles, iot and so on are utilizing newlinethe fpgas for hardware implementation. optical and medical imaging newlinetechniques require an efficient hardware architecture to enhance the newlineperformance in real-time and highly depend on image data for processing and newlinediagnosis. various health related parameters including ecg, eeg, heart-rate, newlineinfants growth monitoring and other diseases are monitored and stored the newlineelectronic data for future use. in recent days, disease prediction and newlineclassification issues are processed using bio-inspired techniques. most newlinecommon techniques are regression methods ann, svm, dt, rf, nb and newlinemlp neural network and deep learning algorithms such as dnn, cnn, rnn newlineetc. these ai models outperforms in terms of accuracy, specificity and false newlinerate prediction and other cataloging problems. despite of its benefits, hardware evaluation for these ml and dl techniques are still deficient. many researchers are utilizing the cpu-based newlinecomputational devices which lead to low-performance, high time-complexity, newlinehigh power utilization and resource wastage. few researchers focused on newlineoptimizing the power and area-inefficiency together. newlineto overcome these challenges, bio-inspired hybrid classifier, newlineelm is boosted with bat algorithm, is designed and implemented in da newlinebased fpga architecture. in the hardware implementation, a novel fpga newlineaccelerator for implementing bio-inspired algorithm is designed with fsm newlinecalculation in order to perform the bio-inspired ai algorithms with minimum newlinepower and to speed up the alu operations newline newline