Image Analytics and Artificial Intelligence Architectures for Medical and Forensic Applications

Artificial Intelligence (AI) has emerged as a critical technology that enables machines to learn from their experiences, adapt to new inputs, and perform tasks similar to those performed by humans. With improvements in methodology and the availability of large databases, AI systems are capable of achieving exceptional performance on a variety of complex tasks. Despite these groundbreaking advances, challenges continue to exist that inhibit or even prevent the widespread adoption of AI in certain applications. For example, (1) the high complexity and energy demands of current AI models make it difficult to deploy them in resource-constrained environments, (2) the poor quality of data impacts the performance of the AI systems and leads to inaccurate predictions, and (3) the lack of understanding of the underlying process reduces the trust in and the verifiability of the decisions made by the AI system.

This dissertation addresses the present challenges and opportunities by promoting the development of high-performance, low-cost AI-inspired architectures. In particular, this dissertation focuses on addressing the following questions:

1. Can the enhancement and augmentation of low-quality data improve the performance of AI systems?

2. Is it possible to reduce the number of parameters in a neural network while increasing its accuracy and robustness?

3. Can the integration of visual perception and cognition of the expert into current architectures demystify the “black-box” element of AI?

The contributions of this dissertation are expected to facilitate the development of efficient models that will accelerate the adoption of AI in medical and forensic applications.