AI and digital tools in pathology – Impact on the industry

Artificial intelligence (AI) has begun to be a part of our daily lives in varying forms and degrees. Healthcare is no exception, and why should it be?

While the promise of AI in medicine is panoramic, some medical specialties lend themselves naturally to AI applications, due to their large emphasis on pattern recognition, such as radiology, pathology, dermatology, and ophthalmology. These areas have seen the development of several AI applications for medical imaging analysis, as a tool for case triage and to establish more precise diagnoses.

Clinical AI solutions through (deep learning) pattern recognitions are augmenting diagnostics – imaging, pathology, and sequencing. In addition to applications in diagnosis and personalized medicine, machine learning (ML) has the potential to be used for better drug discovery and manufacturing, clinical trials, development of smarter electronic health records (EHRs), and the prediction of epidemic outbreaks.

Administrative AI tools are reducing provider burden and increasing efficiency by use of NLP for recording digital notes and automating laborious tasks, such as retrieval of required information from unstructured medical records, lab results, or medical history.

Wearables produce new types of longitudinal data that cross over into diagnostics, by monitoring vital signs, collect large amounts of data via Internet of Medical Things (IoMT) to provide personalized guidance to individuals, and to develop longitudinal views of triggers for disease onset or deterioration. By providing patients with an array of medical devices, such as implantables, blood pressure cuffs, sensors, and wearables, doctors can access real-time patient data remotely. This supports the needs of patients with a chronic condition. The increased adoption of EHRs and personal healthcare records that can connect longitudinal and genomics data provides further possibilities for AI to grow as a diagnostic tool.

Digitization is an essential first step to enable the use of computational analytics and AI. This remains a significant challenge as healthcare is one of the least digitized economic sectors, and digitization is the first step to building AI. The costs of digitization, computation, and storage, though decreasing, are still a significant deterrent to adoption as healthcare organizations need to invest significantly before they can start to realize the potential benefits of AI.

At Dr Lal PathLabs, deployment of AI solutions into the digital histopathology workflow has enabled our histopathologists to quickly prioritize cases through AI-generated cancer heat maps on the biopsy slides. The measurement of tumor and grading by AI enables higher accuracy and efficiency in reporting prostate and breast cancer biopsies, eliminating subjectivity among pathologists.

We are also validating an AI tool for predicting IHC status on breast cancer histopathology slides without the use of IHC staining. This has the potential to bring down testing costs for breast cancer reporting.

AI tools are only as good as the data used to develop and maintain them, and access to high-quality data is needed to create effective AI tools. Limitations and bias in data can reduce the safety and effectiveness of AI tools and paucity of evaluations in clinical settings can make AI tools lack transparency. There is need to put in place robust data governance, ensure interoperability, standardize data formats, and bring clarity to consent over data sharing.

As more AI systems are developed, large quantities of data will be in the hands of more people and organizations, adding to privacy risks and concerns. Clear guidelines on the required levels of pseudo-anonymization or anonymization, and on the linking and de-linking of records needed for certain applications, would be of value.

The undeniable potential of AI will augment, support and in some cases, challenge, the expertise of diagnosticians. New professionals with hybrid clinical and data-science skills will need to be integrated into healthcare, working not in silos, but as a single AI-empowered workforce.