Artificial Intelligence enables development of new diagnostic and prognostic models, supporting the move toward personalized medicine. But there are challenges to overcome first.

By Gordon Feller      

Summary: Artificial intelligence (AI) is transforming clinical laboratories by enhancing efficiency, accuracy, and clinical decision-making, while necessitating changes in infrastructure, operations, and workforce training.

Takeaways:

  1. Efficiency and Automation: AI-powered automation is improving efficiency in clinical laboratories by handling routine tasks such as specimen analysis, data entry, and reporting, allowing human experts to focus on more complex issues.
  2. Diagnostic and Prognostic Models: AI, through machine learning and deep learning, is enabling the development of new diagnostic and prognostic models, facilitating personalized medicine by analyzing large clinical datasets.
  3. Challenges and Adaptation: Successful implementation of AI requires adapting laboratory infrastructure, updating workforce training, and addressing ethical concerns and regulatory frameworks to ensure responsible use and data privacy.

Artificial intelligence (AI) is transforming clinical laboratories and expanding the scope of laboratory medicine. The combination of automation and AI has the potential to significantly enhance efficiency, accuracy and clinical decision making in laboratory medicine, but will also necessitate adapting laboratory operations, infrastructure, and workforce training.

Early adoption of AI-powered automation is already showing indications that it can lead to increased efficiency in clinical labs. This is being accomplished by automating routine tasks like specimen handling, analysis, and reporting. However, it will require changes to laboratory infrastructure and workforce training to adapt to automated systems.

Artificial Intelligence and Clinical Data

AI, particularly machine learning and deep learning, can be applied to the large clinical datasets generated through increased automation. This enables development of new diagnostic and prognostic models, supporting the move toward personalized medicine. Some analysts who are following the trendlines in labs note that the promise of automated image analysis is being coupled with AI for tissue-based disease diagnosis in pathology. AI is starting to aid in interpreting complex laboratory data and assist in disease diagnosis—across various disciplines.

If AI is to realize its full potential inside labs, changes in training programs for pathologists, clinical doctoral scientists, and laboratory professionals will be necessary. This will enable labs to fully leverage automation and AI capabilities in areas like digital pathology and interpretation of complex data.

AI is starting to be utilized in clinical laboratories for various applications, including:

  • Automated Spectral Analysis: AI models are being developed to automate the analysis of spectroscopic data for disease detection and multivariate analysis of disease conditions. For example, Mayo Clinic has implemented an AI model to automate the spectral analysis of kidney stones, classifying 708 unique stone types.
  • Digital Image Analysis: AI is being employed for digital image analysis in microbiology, hematopathology, immunology, and forensics. Mayo Clinic has also implemented an AI tool for automated fecal analysis, flagging positive samples for technologist review.
  • Data Entry Automation: AI can automate data entry tasks and processes, creating standardized reports for lab test results and automating entry into Laboratory Information Systems (LIS).
  • Test Utilization: AI models can minimize unnecessary laboratory testing by predicting test results from available patient data, reducing redundancy and duplication of tests.
  • Quality Control: AI can assist in auto-verification of test results for quality control purposes.
  • Laboratory Operations: AI is being used for data analytics and operational decision-making, such as predicting workflow volumes, staffing requirements, and instrument automation.
  • Interpretation and Diagnosis: AI algorithms are being developed to aid in the interpretation of complex laboratory data and assist in disease diagnosis, such as detecting tuberculosis from chest X-rays and cervical cancer from Pap smear images.

The Potential of AI

Speaking about AI’s uses inside both molecular pathology and digital pathology, one leading figure has kept an eye on the growing usage. 

“It’s been there, it just hasn’t been widely advertised,” says Carlos J. Suarez, MD, associate director of the molecular pathology laboratory at Stanford University Medical Center in California and co-director of the Genetic and Genomic Testing Optimization Service.

While AI adoption in clinical laboratories is still in its early stages, these examples help to demonstrate the potential of AI in several areas:

  • Enhancing efficiency
  • Improving accuracy
  • Sharpening decision making in laboratory medicine    

According to Samuel Hamway, Senior Analyst at Nucleus Research, AI is transforming the landscape of laboratory medicine, with implications for diagnostic accuracy, efficiency, and personalized patient care, specifically in low- and middle-income countries.

“AI technologies are increasingly utilized to analyze the vast data produced in clinical laboratories. These technologies are adept at identifying patterns in complex datasets, such as genomic information, microbial metagenomics, and imaging data,” he says. “It is particularly impactful, offering high precision which is crucial for both diagnosis and ongoing patient management. The predictive capabilities of AI extend to forecasting disease progression and patient outcomes by integrating various data types, including patient records and clinical outcomes. This comprehensive data analysis facilitates personalized treatment plans, improving therapeutic outcomes. Moreover, AI supports the optimization of laboratory operations, from enhancing data management to automating labor-intensive tasks such as data normalization, essential for maintaining efficiency amid resource constraints.”

Artificial Intelligence in the Lab: The Benefits

AI in clinical labs primarily excels at automating routine tasks and augmenting human decision making.

For instance, AI algorithms can analyze large datasets rapidly, identify patterns in medical images, and predict test results based on existing patient data. This automation of mundane work allows laboratorians to focus on more complex tasks that require human expertise and judgment.

Learn More: Molecular Pathology Workflows Set to Advance with AI-Powered Tool

One of the key benefits of AI in labs is its ability to improve and validate the accuracy of results. Machine learning models can be trained on vast amounts of historical data to detect anomalies, flag potential errors, and provide quality control checks. This can lead to more consistent and reliable test results, ultimately benefiting patient care.

However, it’s important to recognize that AI is not a panacea for all laboratory challenges. While it excels at pattern recognition and data analysis, it lacks the contextual understanding and critical thinking skills that human experts possess. Rather than replacing entirely the human-in-the-loop, AI should be viewed as a tool to augment human expertise.

The Challenge of Artificial Intelligence

There have been instances of pushback from laboratorians and pathologists regarding AI implementation in labs. One notable example is in the field of digital pathology. Some pathologists have expressed concerns about the reliability of AI algorithms in diagnosing complex cases, fearing that overreliance on AI could lead to misdiagnosis or missed subtle features which human experts might catch. Additionally, there are concerns about job security and the potential deskilling of the workforce as AI takes over more routine tasks.

Hamway outlines some of the challenges facing labs, “including the scarcity of digital infrastructure and the limited availability of structured and comprehensive data sets. To overcome these obstacles, substantial investments are necessary to improve digital capabilities and develop data-sharing networks that support AI’s needs. Additionally, there are ethical considerations regarding data privacy and the transparency of AI-driven decisions, necessitating robust regulatory frameworks to ensure responsible use. The successful implementation of AI in clinical labs requires a coordinated approach involving various stakeholders—government bodies, international organizations, and healthcare providers.”

Hamway believes that this collaboration aims to foster innovation and facilitate the integration of AI technologies that are sensitive to the specific needs and constraints of clinical environments.

Implementing AI in clinical labs with a more balanced approach would entail a focus on the uses of AI in these elements:

  • Collaboration: Involve laboratorians and pathologists in the AI development and implementation process to ensure the technology meets their needs and addresses their concerns.
  • Validation: Rigorously test and validate AI algorithms before deployment, comparing their performance against human experts and established gold standards.
  • Transparency: Ensure that AI decision-making processes are explainable and interpretable, allowing human experts to understand and verify the results.
  • Continuous monitoring: Regularly assess the performance of AI systems and update them as needed to maintain accuracy and relevance.
  • Education and training: Provide comprehensive training to laboratory staff on how to work alongside AI systems effectively.
  • Ethical considerations: Address privacy concerns and ensure compliance with regulatory requirements for handling sensitive patient data.

Implementing AI in unbalanced ways happens when there’s a rush for the adoption of unproven technologies, failing to involve key stakeholders, or viewing AI as a replacement for human expertise rather than a complementary tool.

The Mayo Clinic: Artificial Intelligence Success Story

One real-world instance of a clinical laboratory using specific AI tools comes from Mayo Clinic. Mayo Clinic has implemented AI technology to enhance information retrieval and analysis across their vast clinical data resources.

Specifically, Mayo Clinic has given thousands of its scientific researchers access to 50 petabytes of clinical data through Vertex AI search, a tool developed by Google Cloud. This AI-powered search capability allows researchers to rapidly retrieve and analyze relevant information from Mayo Clinic’s extensive clinical datasets.

“Our prioritization of patient safety, privacy, and ethical considerations, means that generative AI can have a significant and positive impact on how we work and deliver healthcare,” says Cris Ross, Mayo Clinic’s chief information officer in a statement to the media on the launch of Mayo’s AI implementation. “Google Cloud’s tools have the potential to unlock sources of information that typically aren’t searchable in a conventional manner, or are difficult to access or interpret, from a patient’s complex medical history to their imaging, genomics, and labs. Accessing insights more quickly and easily could drive more cures, create more connections with patients, and transform healthcare.”

Mayo’s current implementation of Vertex AI search aims to address several key challenges:

  • Scale: With 50 petabytes of clinical data, traditional search methods would be slow and inefficient. AI-powered search can quickly process and retrieve relevant information from this massive dataset.
  • Language barriers: The system can accelerate information retrieval across multiple languages, expanding the pool of accessible data for researchers.
  • Research efficiency: By providing faster access to relevant clinical data, researchers can spend less time searching for information and more time analyzing it, potentially accelerating the pace of medical discoveries.
  • Cross-disciplinary insights: The AI search capability likely allows researchers to uncover connections and patterns across different medical specialties and data types that may not have been apparent through manual searches.

Facilitating Change

AI’s emergence as a tool in medicine is already creating some pressures for real change. Consider, for example, the debates regarding AI and medical devices. Zach Rothstein, JD, executive director of AdvaMedDx, was asked his view about how well the current device regulatory framework support the review of diagnostics that are developed using AI or that incorporate AI.

He recommends that Congress and the FDA enact an innovative predetermined change control plans (“PCCPs”) program:

“If implemented in accordance with its statutory authority, we believe PCCPs will encourage the development of AI/ML enabled devices while ensuring they are safe and effective for patients,” Rothstein says. “However, the PCCP mechanism enacted by Congress is limited to changes made to a device without changing the device’s intended use. In many cases, a diagnostics platform can have multiple applications, such as the potential to identify multiple diseases, or provide decision support for multiple conditions. These use cases would potentially implicate multiple intended uses. A concept like Tech Cert in VALID would provide appropriate tools for FDA to be assured of the safety and effectiveness of the new use case without requiring new review cycles and submissions for each such use case. As far as AI regulation with respect to devices is concerned, we believe FDA should be the sole regulatory body with jurisdiction to avoid redundant oversight or regulatory confusion for medical devices.”

Artificial Intelligence and the Future of the Clinical Lab

Clearly, AI has the potential to significantly enhance clinical laboratory operations by automating routine tasks, improving accuracy, and supporting decision-making. However, its implementation should be approached carefully, with a focus on collaboration, validation, and continuous improvement. By striking the right balance between AI capabilities and human expertise, clinical labs can leverage this technology to improve patient care while maintaining the critical role of skilled laboratorians and pathologists.

AI tools can augment human expertise in the medical field, rather than replace it, by providing powerful data analysis and retrieval capabilities. 

ABOUT THE AUTHOR

Gordon Feller’s first published article, analyzing science/technology trends was published in 1980, before he’d finished his undergrad degree at Columbia University in NYC. He continued with this focus through his graduate degree, and into his years working at think tanks and corporations. During these decades he’s published hundreds of articles, advised dozens of governments and their agencies, assisted investors and their portfolio companies. He’s been working out of Silicon Valley for 40 years.