Opening Doors to AI-Enabled Pathology
Finding a low-resource route to AI-enabled digital pathology
Talat Zehra | | Opinion
The COVID-19 pandemic changed what we define as “normal” in our lives – and many of those changes may be permanent. Take, for instance, digital pathology, which allowed us to continue working while avoiding unnecessary exposure during lockdown. Its adoption may have been rapid, but our use of digital tools will continue long after the crisis is over. Pathology is not new to the digital world – it has been over two decades since the introduction of whole-slide imaging (WSI) scanners. But as the world grappled with a deadly pandemic, digital pathology rose to the occasion and traveled like wildfire around the globe.
Digital pathology lets pathologists in their homes connect to colleagues in any part of the world in an instant. Once the glass slides are digitized using WSI scanners, the images can be sent anywhere for second opinions, remote diagnoses, or educational purposes – removing physical barriers that cost time and money. Digitizing also ensures long-term slide preservation, whereas glass slide quality can fade over time and necessitate re-staining. And, of course, digitization opens the door to automated, artificial intelligence (AI)-enabled disease diagnosis and prognosis models in laboratories with sufficient resources and infrastructure. These automated methods can pick up small pathologies that are more easily missed on a busy slide in a busy lab. In the near future, AI-enabled digital pathology tools will achieve diagnostic and prognostic capabilities beyond the scope of traditional microscopy.
Developing countries contain more than two-thirds of the world’s population – and more than half of its cancer and endemic disease burden. Pathologists are scarce everywhere, but the situation is especially grave in the developing world. Digital and computational pathology can help – but AI-enabled techniques are beyond the reach of most low-resource organizations. Moreover, regulatory barriers and staff training present further challenges to adoption.
But all hope is not lost. Low-resource organizations can start their journeys toward digital pathology by leveraging resources from the open-source community. Many organizations offer free access to their WSI archives (for instance, The Cancer Genome Atlas, The Cancer Imaging Archive, the Digital Pathology Association’s Whole-Slide Imaging Repository, and more), although rapid downloading and local storage of these large datasets still presents a challenge. Another option is to use a microscope connected to a camera. A pathologist can photograph a region of interest for a particular pathology and then annotate it to train an automated AI model using open-source software (for instance, QuPath, ImageJ, Cytomine, Orbit, ASAP, or others). In this way, pathologists can make disease models even without high-tech scanners, large hard drives, or high-speed Internet.
Worldwide, one thing is clear: developing countries are the biggest source of data that must be preserved to make reliable disease models, analyze trends, and predict outcomes. Information is the fuel on which modern algorithms run – and the foundation on which the future of precision medicine depends. If we take responsibility for saving this data now, pathologists, patients, and technology innovators all win. Let’s work together to open the door to a new era of diagnostic medicine for all of humanity.