According to a 2017 ABI Research report, comprehensive healthcare damages from cybercrime will exceed $1 trillion in 2018. Since healthcare firms want to avoid the huge costs of a breach (and all the ramifications related to HIPAA compliance and reputation), there is a huge incentive to adopt more intuitive and adaptive security protections.
Enter artificial intelligence, and more specifically, machine learning.
These two concepts are very similar, so to clarify the definitions:
- Artificial intelligence – This branch of computer science is committed to creating machines that are “smart”; i.e., that have the same characteristics as human intelligence. Another way to look at it comes from Stanford, which defines it as “the science of getting computers to act without being explicitly programmed.”
- Machine learning – This application of artificial intelligence, which is experiencing faster growth than any other part of AI, allows computers to pick up new information (“learn”) and refine their capabilities in response, without any code directing them to do so (per Expert System).
The future of the AI & machine learning market
Saying that artificial intelligence (AI) and machine learning are growing rapidly in healthcare does not capture the magnitude of that growth. A 2017 report from analyst Market Study Report forecasts that machine learning will grow to $10 billion by 2024.
One of its biggest areas will be diagnosis and medical imaging, which is expected to hit $2.5 billion (USD) by that same point. In 2016, the market size was $320 million, so to get to that total figure of $10B by 2024, the market will be skyrocketing at greater than a 38% compound annual growth rate (CAGR); during that same period, the diagnosis and imaging segment is projected to increase more than 40%.
The promise of vast machine learning expansion is even more impressive. MarketsandMarkets predicts that machine learning spending will hit $8.81 billion in 2022, up from $1.41 billion in 2017, at a CAGR of 44.1%.
Beyond medical imaging and diagnostics, other primary areas of healthcare AI that will expand rapidly in the years to come include genomics, pharmaceutical discovery, and personal assistants.
The ABI Research report noted that machine learning will not just cause ripples but large wakes; by 2021, it will have propelled the analytics, intelligence, and big data market to $96 billion.
How AI will grow
As Intel senior principal engineer Ted Wilke noted, healthcare AI started as a way to make day-to-day tasks easier and manage big data. Now, the AI market will expand dramatically because big data is growing, so storing and managing data must be as effective and efficient as possible. Applications have become much more diverse and complex: improving treatment plans, performing genomic sequencing, and uncovering sophisticated patterns within medical images.
Furthermore, many healthcare entities are wading (or diving) into predictive clinical analytics. This branch of AI helps users estimate the likelihood that a patient might experience a return of their illness by analyzing patient records with increasingly powerful algorithms.
Broadly speaking, healthcare AI applications can be categorized as fulfilling needs for automated reasoning, but demand for the less tantalizing search and classification applications is more prominent.
What is extraordinary about artificial intelligence is its ability to use methods such as natural language processing and text analytics to make sense of big data quickly. Medical researchers and clinicians can use these systems to find similarities in data and patterns. Reasoning capacity allows these platforms to advise security personnel of steps they may want to take.
How machine learning will grow
Machine learning will grow for similar reasons tied to innovations. The massive growth in data generation and technological advancement will be key reasons businesses increasingly turn to machine learning, as MarketsandMarkets indicates. Additionally, the market has strong expansion potential because more modern systems are being implemented in healthcare and because efforts to make services and operations intelligent are becoming a central focus.
The ABI team noted that machine learning will replace traditional AV, heuristics, and signature-based systems. As a result, security information and event management (SIEM) systems and protocols will become more sophisticated, enabling automatic threat discovery by incorporating analytics techniques such as machine learning.
ABI researchers expect deep learning algorithms (a type of machine learning and an evolution of task-specific algorithms) and user and entity behavior analytics, or UEBA (automatic threat discovery that incorporates analytics techniques such as machine learning, an evolution of rule-based systems), to grow. In this climate, security solution providers are beginning to incorporate deep learning, UEBA, and other machine learning methods into their offerings.
Today, SIEM is considered a best-practice approach to the threat landscape, providing healthcare organizations with the protection they need. SIEM programs pull in data from an organization’s infrastructure across the full spectrum (including network security elements such as antivirus and firewalls, as well as general software and hosting platforms). After collecting this data, the software aggregates logs, giving security teams access to a complete monitoring log and critical insights. Security analytics enable the company to identify and mitigate threats immediately through real-time processing.
With the advent of the machine learning era, insider error will become a less common cause of compromise because unsupervised methods reduce reliance on manual review. The log-centered approach of SIEM will be joined with additional UEBA programs, both isolated from one another and integrated in their approaches and findings.
How healthcare can prepare for the AI & machine learning world
Organizations do not need to adopt Silicon Valley systems to embrace these technologies. Instead, machine learning techniques can be leveraged by organizations to build custom AI big data analytics environments.
Many healthcare companies are approaching AI by incorporating semantic data lakes, explained Albert Einstein College of Medicine professor Parsa Mirhaji, MD, PhD. The semantic approach is an upgrade from relational databases, which have very strict structural requirements. The relational method requires a pre-established, organized schema. Isolated data silos result.
Data lakes are collections of data in a wide variety of formats. Insights are derived by bringing together data with no known relationship by applying unique, standardized identifiers to each. This capability is permitted by the incorporation of natural language processing within data lakes. All of these applications will require a HIPAA-compliant database.
HIPAA-compliant AI infrastructure
For healthcare teams exploring AI-based security monitoring and analytics, the hosting environment needs to support both the software and the protection of sensitive data. Atlantic.Net’s managed HIPAA GPU hosting includes the NVIDIA A2 with 16 GB of GPU memory, offering a lower-cost option for lighter inference workloads. Managed security, backups, and a BAA support the underlying infrastructure. Discuss your chosen software, expected processing demand, and patient-data requirements with our team to identify a suitable configuration.
