Artificial Neural Networks (ANNs) are increasingly deployed in safety-critical applications including autonomous vehicles and medical diagnostics. To avoid unexpected ANN behaviour and provide evidence for their trustworthy operation, ANNs should be thoroughly tested. In this study, representing ANNs with a programming language such as Java, enabled us to utilize software engineering tools for comprehensive testing of ANNs. Inspired by fault localization in software, the whitebox ANN testing approach that we presented in our paper, establishes the hit spectrum of neurons and identify suspicious neurons which are responsible for inadequate ANN performance. Moreover, we also modified original inputs to synthesize adversarial inputs that increase activation values of suspicious neurons.