The Executive Suite: How artificial intelligence can assist primary care physicians in cancer diagnosis

In recent years, improvements in prevention, screening, diagnosis, and treatment have greatly lowered cancer incidence and mortality. Early diagnostics played an instrumental part in reducing mortality from cancer in the US by 29 percent from 1991 to 2017, annual statistics reported by the American Cancer Society show. However, a model created by the National Cancer Institute (NCI) projects nearly ten thousand more people will die of cancer over the next decade as a result of missed screenings, postponed diagnoses, and cutbacks in oncology care caused by the COVID-19 pandemic.

Identifying Cancer Cells in the Blood

Catching an early-stage cancer before it spreads based on a few drops of blood is considered the “holy grail” for cancer diagnostics but has been beset by many challenges. As cancer is more curable at early stages, its early detection is crucial. One of the most studied bio sources so far for early diagnosis is circulating tumor cells (CTCs), which are shed into the blood from metastatic sites and can therefore give a snapshot of the cancer progression at a given time. However, their extremely low numbers in the blood and molecular heterogeneity make it nearly impossible to use them as a diagnostic tool. Abnormal blood cells per contra, that are also indicative of different blood cancers, are more abundant in the blood stream and are routinely detected and enumerated by experienced technicians that inspect stained blood films using microscopes in hematology labs.

Current Challenges in Blood-Based Early Cancer Detection

While some cancers, such as prostate cancer, require specific antigen testing (PSA), or identification of immunoglobins for myeloma, blood cancers may be the first type of cancer that can be detected early based on basic blood testing that generally takes place in primary care. The most commonly performed blood test in the world, the complete blood count (CBC) that is mostly used to understand a patient’s overall health status, could become a much stronger indication of blood cancers if they incorporate the ability to flag and count abnormal cells, including nucleated red blood cells (nRBCs), Immature Granulocytes (IGs), and blast cells.

Today, this type of blood analysis is only partially available by using lab-based hematology analyzers, which are mostly large, expensive and located far from the frontlines of primary care. Furthermore, CBC testing is performed using analyzer technology that has not had a major upgrade since the 1960s and does not allow for precise identification and counting of abnormal cells. Standard hematology analyzers only measure 3 to 4 optical or electrical properties of each cell, which makes it very difficult to differentiate normal and abnormal cells that vary only in nuance. As such, abnormal cells cannot be differentiated from their normal counterparts and are often confounded with other types of cells, making the results unreliable. When the presence of abnormal cells is suspected a “flag” is raised by the analyzer and the sample is sent for manual inspection by microscope.

AI’s Role in Abnormal Cell Flagging and Enumeration

New point-of-care technologies that leverage machine vision and AI can now provide rapid analysis of blood cells, including comprehensive flagging and even counting of abnormal blood cells, to provide a better option for early diagnosis.

Making these diagnostics available through a point-of-care system means that physicians in primary care can have a significantly earlier indication of an issue through rapid blood testing in their office. This also means that patients do not have to go into specific labs to have routine blood testing that can be handled in the doctor’s office. Furthermore, a trend towards treating oncology patients at home, makes such POC diagnostic technologies highly valuable. Patients receiving chemotherapy must have their blood counts done regularly, being able to administer treatment at home demands portable and robust POC analyzers that can deal with abnormal cells that are frequent in such patients. Moreover, these patients are immunosuppressed and allowing them to be tested and treated at home eliminates the risk of infection in hospitals or other clinical environments.

Rapid and accessible abnormal cell enumeration can play a greater role in early cancer detection than the present cumbersome and time-consuming process. Making it available through a point-of-care diagnostic system that is easily managed and offers a fast turnaround time for lab-accurate results would both lower pressure on the healthcare system, as well as improve care options for patients. The ongoing COVID-19 pandemic has further underlined the need for such a solution to be implemented.

Developments that have gone through the clinical rigor of the peer-review and evaluation process will start making an impact in the “one-drop diagnostics” market. It will be to everybody’s benefit once the companies that have successfully showcased their technology in clinical trials are included in the mainstream healthcare system.