An emerging innovation is the application of AI to cancer imaging. In particular, in AI, Deep Learning (DL) Technology (in short, DELT) can support the qualitative interpretation of cancer imaging by expert clinicians, such as volumetric delineation of tumors, stage of cancer, mutations as well as evaluation of the impact of diseases on adjacent organs and strength of anti-cancer treatments. This study has two main goals. The first goal is to analyze technological trajectories of DL technologies in two cancer fields to show evolutionary growth of this emerging technology in medicine. The second goal is the identification of main barriers to the adoption of deep learning technology in cancer diagnostics and suggest best practices of innovation management in healthcare systems. Innovation management for a widespread diffusion of this new technology to support organizational efficiency of healthcare faces a number of challenges related to three macro level aspects. Firstly, economic factors to consider are given by the costs of technology ‘research and development’, ‘searching’, ‘acquisition’, maintenance’, ‘control and oversee’ and by ‘transition’ costs. Innovation management needs to justify these costs to government and third-party payers developing reimbursement strategies and showing a high efficacy and efficiency of deep learning technology in clinical practice. Secondly, organizational factors to consider in digital pathology are that ‘organizational workflow’ of pathologists and other physicians is likely to change and it may be costly at the initial phase of diffusion owing to additional workflow in hospitals (e.g., equipment, research personnel, data storage of digital pathology, etc.). Further, best practices for healthcare innovation management should invest in the promotion of data and knowledge exchanges between healthcare institutions, tech companies and data scientists for the promotion of positive ‘network externalities’. Technological transformation requires a paradigm shift toward a patient centered ecosystem where medical demand can be best met by interdisciplinary teams made of all the key healthcare stakeholders, including new technology developers, data scientists, and healthcare institutions. Third, human resources are indeed a critical factor for the implementation of deep learning technology in clinical practice. Innovation management need to promote change in the ‘organization culture’ and invest on employees ‘training’ and ‘education’. Next, psychological barriers associated with issues of ‘self-efficacy’, ‘trust’, ‘risk perception’, ‘uncertainty’ and ‘fear of replacement’ must be promptly addressed as human capital might be the greatest challenge and require the longest time to be dealt with. Finally, institutional factors to consider are concerned with ‘regulation’ for the definition of ‘evaluation criteria’ to guarantee ‘accountability’, patients ‘privacy’, ‘information’ and ‘consent’. Overall, then, this study shows that organizational change is imminent and best practices of
innovation management need to implemented as a matter of urgency for the benefit of all the stakeholders.