Digital pathology set to speed diagnosis, cures
Digitalizing pathology is the next step in digitizing patient records in the shift to personalized medicine. Until recently, lab results generally were recorded as text in the patient’s electronic medical record (EMR). When a diagnostic image was added to the EMR, it was usually as a static picture rather than as a rich set of digital data that can be mined and analyzed by analytics software.
But that situation is changing rapidly.
Digitalizing microscopic images requires the use of digitized glass slides and specialized scanners. The result is digital pathology: a dynamic, image-based environment of the captured information that can be interpreted and managed as digital data.
In other words, a picture of whatever is on the slide is only one possible visualization of the data that is captured in this way. The data can be mined and analyzed, and the array of results presented in numerous visualizations for myriad uses.
Healthcare is changing. Our latest report shows you how.
Machine assistants for pathologists
Working with images that are so complicated and varied is highly complexity. Diagnostic clues are often nuanced and veiled, and sometimes missed by the human eye. Think of the millions of possible images of slides that are viewed under a microscope every day. Now think of the trillions of total slides pathology professionals worldwide view through the course of their combined careers. It is obvious how easily something could be overlooked or misidentified.
The medical information load is staggering, yet each image and pathological report generated is essentially used only once and for one purpose: a single diagnostic instance in a single patient. That amounts to an incalculable number of missed opportunities, both in terms of helping individual patients and in adding to the general medical knowledge base. While human pathologists have saved many lives using whatever tools they had to work with, historically, that traditional method represents massive inefficiencies.
More patients can be saved using new tools such as computerized vision and artificial intelligence in conjunction with digitalized data. Digital pathology can ensure new, life-saving tools that rely on a growing pathology knowledge base, further refining medical discovery in diagnostics.
Another issue driving the digitalization of pathology is the continuing shortage of human pathologists. Machine assistance is a practical way to ensure that all patients get the correct diagnosis in the fastest amount of time.
The end benefits to patients are faster and more accurate diagnoses, faster and more accurate treatment, and less stress while waiting on test results, to name but a few.
On the professional side, some of the ways digital pathology can help pathologists in their real-world, day-to-day work are:
- Lightening the workload. By automating much of the basic diagnoses via digital pathology and analytics, routine diagnostic work no longer clogs the pipeline. This frees pathologists to work on the difficult cases, where their talent is most needed.
- Speeding turnarounds. Faster delivery of results to healthcare providers means patients are treated quicker and more accurately. This can also help stem the rise of antibiotic-resistant microbes if diagnosis is confirmed before antibiotics are prescribed. Of course, some tests take longer than others, and healthcare providers will still need to act promptly and preemptively. But many times, diagnosis is delayed by pathologists’ backlogs. Eliminating those are helpful in numerous ways.
- Making it easy to share. Getting pathology information to specialists quickly is crucial, especially when patients have life-threatening conditions. Digitalized pathology is a fast and easy way to free the data from one physical location, bringing the data to the experts, not the experts to the data.
- Creating efficient digital reviews. It’s easier to run comparisons, check for errors, and use numerous digital tools to gain further insights if the pathology is digitalized. Working with limited information that is manually entered relies on older image processing technology and is subject to human error. This slows processes and increases errors.
- Improving processes and risk management. Internal workflows from data availability and governance to risk management and glass management can all be drastically improved.
- Varying diagnosis models. Pathologists can choose to use standard modalities, image analysis, or a variety of algorithms to arrive at or confirm a diagnosis. This is likely to greatly improve accuracy.
- Enhancing ease of integration. Digitalized pathology can be integrated with a number of other systems ranging from EMRs and patient portals to ICT systems and insurance reporting tools.
These benefits are achievable through using digital pathology with existing analytics, systems, and tools. Even more can be achieved with the use of newer tools.
Computer vision, machine learning, and artificial intelligence
Computer vision overlaps with image processing, particularly on the front end, where image processing techniques such as image enhancement and noise reduction are necessary for a machine to see and recognize an object.
Seeing is a very complex process, yet one that humans typically take for granted. Traditional image processing isn’t very good at recognizing more than a few visual objects, if that.
Computer vision seeks to remedy that by attempting to mimic biological systems in how objects are recognized. For example, a cat, a dog, and a child can all recognize a ball, no matter the ball’s color, size, or condition (such as brand new, punctured, or covered in mud). Computer vision works to recognize objects in much the same way as the cat, the dog, and the child—that is, through neural networks to create machine learning.
Machine learning is a subset of artificial intelligence; deep learning is a subset of machine learning. Essentially, a machine "learns" by comparing an enormous amount of data and working to eliminate errors until its results are entirely accurate. In the case of identifying a ball, as mentioned above, the software compares every possible depiction and representation of a ball until its ability to recognize a ball is near perfect.
It handles pathology in the same way. It will analyze and compare every possible presentation of Staphylococcus aureus, for example, until it reaches the point where it will always recognize the microbe. In this way, machine learning can also find microbes we humans were not even looking for in the sample, since it can analyze, compare, and learn an almost incalculable number of things in the same datasets simultaneously.
But before any of that can happen, pathology must be in a digital form that machines can use both to learn from and analyze.
Digital pathology’s role in personalized medicine
Personalized medicine, also known as precision medicine, is medicine tailored to a patient’s predicted response or susceptibility to a disease. A patient’s specific DNA and epigenetics play a large role in both developing a customized treatment and predicting an individual’s response to it.
Take cancer, for example. DNA analysis can reveal a patient’s likelihood of developing any given form of the disease. Epigenetics, the mechanism that triggers genes in DNA to turn on or off, have an effect on the occurrence of the disease, its resistance to treatment, and the likelihood of recurrence too. Environmental factors like smoking have an epigenetic effect on lung cancer, for instance. But there are also epigenetic drugs that combat cancer.
Indeed, there are two new trials underway—one for lung cancer and the other for bladder cancer—to study the effectiveness of combining epigenetic with more traditional immunotherapy treatments. Combinations with chemotherapy are also being studied by researchers at Johns Hopkins. Researchers the world over see epigenetic drugs as the correct path toward an eventual cure for cancers of all types.
Further, DNA vaccines are being developed for several uses ranging from countering antibiotic-resistant diseases to preventing reoccurrence of the specific cancer a patient suffered from earlier. Imagine a vaccine that ensures your cancer will not come back after it’s cured the first time!
Digital pathology plays an important role in these medical advances in that it is crucial to identifying the disease and the specific nuances of that individual manifestation of the disease. This data is the essential input in diagnostic algorithms and customized medicine calculations.
Digitalized pathology data can also be anonymized and combined with data from other patients to help speed diagnosis for all patients and to advance medical discoveries at a much faster rate.
In other words, by gathering, mining, and analyzing information from many patients and then digitally combining it with the patient’s digitalized DNA and epigenetics, customized treatments, cures, and vaccines can then be made to work for any specific patient. That’s what personalized medicine is: customized to the person and the disease to maximize disease eradication and lower or eradicate side effects from the treatment.
Digital pathology applications
We’ve covered only a few of the reasons why digital pathology is the essential keystone in modern and future medicine. Even so, the list is breathtaking and still growing.
To recap, some of the many applications for digital pathology as it is used today include:
- Primary diagnosis
- Diagnostic consultation
- Intraoperative diagnosis
- Medical student and resident training
- Machine learning
- Manual and semi-quantitative review of immunohistochemistry (IHC)
- Clinical research
- Diagnostic decision support
- Peer review
- Tumor boards
In summary, digital pathology supports education, tissue-based research, drug development, and the practice of pathology. Bottom line: It is an innovation that reduces laboratory expenses, increases operational efficiencies, improves productivity, and enhances treatment decisions and patient outcomes.
Digital pathology: Lessons for leaders
- Digital pathology will improve the entire workflow of medical care.
- Combining the latest in computer technologies, such as machine learning, AI, and high-performance computing, will enable general-purpose use of digital medicine.
- Personalized medicine requires these technologies as an enabler.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.