The study has been well designed and articulated. A new Deep learning algorithm developed was tested with standard performance analysis in pattern recognition such as receiver operating curve (ROC). Validation in primary data sets with very good performance of 1% false positive rate indicates that this algorithm is much suited in practical implementation. As authors already suggest such system can aid to expert radiologists in better interpretation of medical tests and save human lives during critical situations.
I would appreciate the authors for taking pain in disseminating this research as open-access and indicating competing interests honestly such that research community worldwide gets better impacted.