Cloud Hosting

By Shreekanth Joshi

Artificial intelligence (AI) and machine learning (ML) have seen exponential growth in their ability to consume large amounts of data and produce accurate insights that approach human level accuracy. This has largely been possible due to the availability of cloud-based resources that are scalable, more cost effective and readily available.

In the healthcare industry, the analytics of health-related data is improving care from super specialized tertiary care centers to secondary and primary care. Telemedicine is making these insights available at the point of care, leading to better and more specialized diagnosis. Both developments make more reliable care available in real time across the last mile and help bridge the gap between large numbers of patients and a limited number of healthcare providers.

Access to care is a major issue in the case of geographically distributed populations. But digital tools and resources can be provisioned on the cloud and made available over the last mile to these areas with data network coverage. This allows primary centers to diagnose and collect digital samples and send them for analysis to tertiary centers.

Machine learning models can be made more robust and accurate using cloud infrastructure. The flexible resourcing available in the cloud can track more last mile data from devices, wearables and health trackers, then stream and aggregate it cheaply in cloud-based storage. The heavy-duty analysis of this large amount of data can be done efficiently using cloud-based compute infrastructure. This in turn allows the ML models to be trained more effectively and their accuracy improves over time.

The large amount of data available for training makes ML models scale even better. For several tasks in image analysis, for example, the model accuracy is already reaching human level. ML models can be made more personalized to start generating recommendations that are very specific to individual patients.

Regulatory Considerations

All this computation comes at a regulatory cost. Data must be secured at rest and in motion, and be anonymize before feeding into the ML models and recommendations must be re-identified to make them specific to a particular patient. This involves resources from not one but multiple cloud providers working in a hybrid manner.

The National Digital Health Blueprint calls for strict adherence to regulations with respect to privacy and protection of patient data. As a result, advanced technical check points need to be implemented to prevent patient data from being accidentally disclosed to unintended recipients.

Its also necessary to enforce consent-related policies that allow patient data to be used only by healthcare professionals with consent and for a specified duration.

This requires a heavy emphasis on securing cloud environments and enforcing controls for data access, processing and the dissemination of insights.

ML models also ingest significant amounts of personal data from each patient device, such as the health tracker on a mobile phone or wearables like Fitbit or sleep monitors, insulin monitors and even blood pressure monitors. All these devices must be integrated with the cloud resources to enable end to end data processing.

Given stringent regulation and privacy requirements, data privacy and access must be monitored and governed. This calls for the use of technologies for data and cloud management. IT teams must have comprehensive management frameworks that can integrate personal and corporate devices and implement the necessary safeguards.

Modern unified management frameworks provide a way to make this problem more manageable and can implement the necessary governance best practices to manage the consumer devices, mobile apps and back-end cloud platforms.

The author is Vice President, Engineering, Persistent Systems

Visit link:
Integrating AI & ML in cloud services for healthcare: The benefits and risks - ETCIO.com