Suggestions
Robert Bakos
HOPPR Co-Founder and CTO, Entrepreneur & Advisor in Digital Health, AI, IoT and Medical Imaging
Robert Bakos is the Co-Founder and Chief Technology Officer (CTO) of HOPPR, a company specializing in AI-powered medical imaging solutions. He has over 22 years of experience in software development, particularly in digital health and medical imaging. Before co-founding HOPPR in May 2019, he served as CTO at higi, where he led the development of a HIPAA-compliant cloud-based platform, managed FDA Class II medical kiosks, and integrated IoT solutions. His leadership contributed to higi's acquisition as part of a $4.2 billion SPAC transaction.12
At HOPPR, Bakos focuses on developing multimodal foundation models for medical imaging, leveraging over 15 million diverse imaging studies to address challenges like radiology backlogs and diagnostic accuracy. The company emphasizes ISO-compliant development to integrate seamlessly into medical workflows.13
Highlights
Today, Oliver Chen MD, Julie Bauml and I had the opportunity to attend some truly insightful talks at the Emory Health AI Bias Symposium. Engaging with thought leaders on the critical issues of bias in AI within healthcare was both inspiring and thought-provoking. As we continue to push the boundaries of AI in medical imaging, it’s essential to keep these discussions at the forefront.
I'm grateful for the chance to learn and collaborate with such a passionate and knowledgeable community.
Learn more about the Emory Health AI Bias Datathon at https://lnkd.in/gynea6T4
#AI #Healthcare #EthicalAI #HealthTech #BiasInAI HOPPR
Researchers from IIT Bombay recently published an approach to federated learning for medical images, which they refer to as "Cross-Client Variations Adaptive Federated Learning".
Federated learning on medical image data often faces challenges due to variations in data across different client sites. The proposed method aims to align the feature space across different clients by using diffusion models with transformers to generate synthetic images. These images are then shared across clients and used to translate local client images into a common feature space. The approach appears to improve the performance of the federated learning model in certain cases, offering another novel approach to working around data privacy concerns in the healthcare industry.
Here are the key takeaways:
- Transforms local images into a unified feature space, minimizing cross-client variations.
- Utilizes synthetic images to maintain privacy while standardizing data.
- Outperforms traditional federated learning and centralized learning in limited data scenarios.
Learn more about this innovative framework here: https://lnkd.in/garMcfGZ
#GenAI #FederatedLearning #MedicalImaging HOPPR
