Many healthcare practices today, including cancer diagnosis, still rely heavily on manual activities and processes – and even those that do utilize digital technologies use them on an independent basis, not under one integral environment. Pathology, the medical specialty of diagnosing disease in patients, most notably cancer, is a good example. Traditional pathology involves manual processes that have remained unchanged for years, where glass slides with tissue samples are analyzed by pathologists using microscopes.But that is changing today with a growing trend of moving toward digitized infrastructures and workflows (for example digital pathology). This trend is expected to accelerate as a result of the Covid-19 pandemic, increasingly saving time and labor costs while providing better and more cost-effective care. Moreover, adding an AI layer to an already digitized workflow can help make processes even more efficient on several fronts, including:
Triage: AI tools can help determine which cases should be reviewed more urgently and which physicians or resources to assign to them. This will ensure that each case is diagnosed by the physician whose capabilities can best be applied to the situation at hand, assigning cases based on a physician’s sub-specialty or level of experience, which becomes especially important in complex cases.
Diagnosis: AI tools can also help with cancer diagnosis and assessment, pointing instantly to particular areas of interest, for instance those that include cancerous cells, and signaling the ‘needle in the haystack’ factor that can shed light on a patient’s situation – and what may be required to treat them. As a result, turnaround times for case reporting to the referring physician can be reduced significantly. In addition, smart AI tools will allow for the automation of some diagnostic tasks that are currently performed manually, such as counting cells, measuring features and automatically filling up parts of the report.
Next-generation diagnostics: With advanced machine learning capabilities and as more digital datasets become available, AI tools will be able to analyze more data – and thus provide more insight – than is currently possible. These tools will go beyond mimicking a physician’s diagnosis; AI models will integrate extensive amounts of data from diverse sources (e.g., imaging, pathology and clinical data), acting as a sort of tumor board, where experts from multiple fields share their findings and knowledge to decide on the most accurate diagnosis and treatment. In the near future, we can expect to see AI serve as the “perfect companion” to the physician, with an unparalleled ability to combine huge amounts of fragmented information accurately and effectively.
Telehealth and Physician Access
Telehealth and Physician Access
With all the advancements in medical science, prognostic assessments and treatment decisions are many times not much better than a shot in the dark. For example, up to 75% of oncology patients (in a number of cancer types) do not respond to at least one of the available treatment drugs. Developing new biomarkers and genomic tests to more accurately predict prognosis and enable physicians to choose the right treatment is a costly, unpredictable, and painfully slow process.
Enter AI. Biomarkers based on artificial intelligence and machine learning can be an important, cost-effective and efficient way to develop tools for precision medicine. AI algorithms could provide new prognostic insights and help oncologists stratify patients into smaller, more homogeneous groups and assist in selecting personalized treatment for each patient based on a multitude of data types. These algorithms will combine pathology, radiology, genomic, clinical and demographic data, and analyze huge databases of medical records, treatments and outcomes. AI-driven solutions could also provide a more accurate and objective analysis of medical data, leading to new computational tests that replace or augment today’s molecular tests.
Obviously an essential part of any quality-driven system, quality control is a real challenge in labor-intensive processes. For example, for pathologists diagnosing cancer, an effective “near bulletproof” quality control requires having a second pathologist review the same biopsy again — a time consuming and costly task. As a result, we find many clinical processes where quality control is done only on a small fraction of cases, if at all, often failing to detect errors.AI and advanced algorithms can help create concurrent processes that rapidly review medical procedures and diagnoses, detecting mistakes early enough to avoid any damage. Fast and automated AI tools will enable pathology labs – for the first time – to apply rigid quality control standards to all their cases. Currently, the majority of cases are viewed by a single pathologist in almost all labs, which means that errors and misdiagnosed cancers may not be detected on time. AI can help transform this completely.
Algorithms can also alert on inconsistent findings, making sure the different medical disciplines provide a coherent picture of the patient’s condition and treatment plan, and that there are no “holes” in the process. An electronic medical record (EMR) that contains all the medical information of a patient could be constantly monitored by AI – whenever something suspicious is detected, such as a lab test result that is inconsistent with a previous diagnosis, or a drug prescription that does not match the patient’s condition, the AI would highlight the inconsistency, prompting a thorough review by the relevant physician.
Decisions on issues like which patient should be sent for a specific test – CT, MRI, colonoscopy and others – are currently conducted based on demographics or other general variables that often have nothing to do with a patient’s medical record. Going forward, we will see more medical screening that’s based on specific characteristics identified by AI algorithms trained on large providers’ datasets. Such predictive algorithms will identify specific patterns, risk factors or correlations between multiple variables using advanced machine learning techniques.Science-based medicine is one of the miracles of the modern age; lives are being saved today that just a few years ago would have been unsalvageable, and new, advanced treatments are providing a much better quality of life for many more than would have been possible in the past. Still, there are many domains in medicine and healthcare that could benefit from adopting the latest developments in artificial intelligence and machine learning and we see their role in improving healthcare becoming ever more important in the future.