Digital Pathology Blog

Quality is Job 1

This of course was Ford Motor Company’s marketing slogan in the early 1980s. The message was clear – in the wake of some issues, Ford was going to build in quality to its automobiles rather than make it a checkbox at the end of the process. Over the years, Ford prospered while its major domestic competitors, General Motors and Chrysler, struggled to compete with Ford as well as foreign automakers.

There is a lot of talk currently about artificial intelligence. AI- in short, is the ability for computing systems to be programmed to perform tasks AND to learn from those tasks to refine their abilities the next time a problem or question or decision is faced by the system tasked to perform the operation.

Until recently, the practice of pathology has been entirely “human-driven”.  Well-trained pathologists examine all tissues and arrive at diagnoses based on their application of learned criteria and experience.  However, it is well-known that the accuracy of human interpretation can be hampered by subjectivity (inter-observer variability), inconsistency (intra-observer variability), and fatigue. Recently, the rise of digital methods in pathology has led to a growing interest in applying artificial intelligence (AI) to aid or even improve on the analysis of medical specimens.  For the pathologist, AI has the potential to improve accuracy, productivity, and workflow by allowing the computer do what it does well: consume lots of data, recognize patterns, and perform automated analyses.  Objective and reproducible specimen examination, along with certainty  that all material has been “seen,” leads to greater accuracy.  AI can be trained to identify specific features and present them selectively to human observers, leading to prescreening (“guided screening”) of specimens and promising improved workflow and productivity.  And, finally, the “holy grail”: under the right circumstances, AI-driven systems can make novel observations of morphologic patterns in pathology specimens, potentially leading to new knowledge and ultimately computer-generated diagnosis.  In this blog, we describe our general approach to applying AI to digital pathology specimens, detail some of the important steps involved in establishing an AI workflow, and discuss the potential benefits of AI.

I’ve spent a lot of time in Starbucks, and over the years, it seems little has really changed with the basic operations. The line to the register is flanked by food options and souvenirs leading to someone who takes your order and your money (The Register). Your order moves to someone standing at the espresso machine (On Bar). Other Starbucks associates are simultaneously working the Drive-Thru customers (On Drive-Thru).

Another associate is grinding beans, ensuring the coffee urns are full, teas are brewed, cups, lids and cup protectors are in place, and coolers are well stocked with sandwiches, croissants and breakfast sandwiches (Customer Support).

So, who has the most important job at Starbucks? The Register? On Bar? On Drive-Thru? Customer Support?

Liquid biopsy is defined as, “A test done on a sample of blood to look for cancer cells from a tumor that are circulating in the blood or for pieces of DNA from tumor cells that are in the blood. A liquid biopsy may be used to help find cancer at an early stage. It may also be used to help plan treatment or to find out how well treatment is working or if cancer has come back. Being able to take multiple samples of blood over time may also help doctors understand what kind of molecular changes are taking place in a tumor.”*

Often times the test is regarded as “non-invasive” as it can be performed with simple venipuncture specimens or voided urine samples.

These tests are gaining traction within the industry as a viable alternative to traditional screening methods for cancer. Study results are turning up positive data in favor of the tests, and a recent report by financial services firm Cowen & Co. said annual sales for the tools could surpass $10 billion.

During training, I had an attending who actually called the autopsy the "awfultopsy," as in "awfultopsies are not that bad if you only have to do one every other year" – which I think was his personal goal. Sometimes the autopsy pathologist would be out and others would cover if there happened to be a case; so many of my attendings did not have conducting posts as part of their routine practice.

It doesn't even seem like a real number but the year 2017 is close.  High school freshman this year were not born yet on September 11, 2001, which does not seem like that long ago.  15 years have passed since the day when my generation will always recall where they were when they heard/saw what was unfolding – much like my parent’s generation can tell you about Kennedy’s assassination like it was yesterday.

On a recent trip, to present at grand rounds at a well-known midwestern medical school, I noticed a dapper gentleman on my flight.  He was wearing a three-piece suit and carrying an impressive attaché case. He stood out among the many travelers dressed in shorts and flip-flops, and who probably wished they hadn’t when we reached our destination where the weather was not as friendly.

There is a decades-old theorem first derived at the Mayo Clinic that although based on little actual data has been popularized over the years stating:

• 70% of patients in a hospital have laboratory data.
• 70% of clinical data points (objective information) come from laboratory data.
• 70% of clinical healthcare decisions result from laboratory tests.
• These data account for less than 5-10% of healthcare costs.

This is a lot of bang for the buck. Imagine 70% of objective information assisting in 70% of healthcare decisions for as little as 5% of total healthcare expenditures.

Almost 3 years ago I wrote a piece for Advance for Administrators of the Laboratory in their "Perspectives in Pathology" entitled Death of the General (Surgical) Pathologist.

This piece discussed the issues surrounding what is perceived by many as the increasing (and necessary) sub-specialization in pathology.  The model of "everybody doing everything” has evolved into “everybody doing one thing".

Historically, practicing pathologists were solid general surgical pathologists, often times with their own expertise/interest in a particular area, but they prided themselves on being "surgical pathologists" rather than "liver", "hemato-" or "dermato-" pathologists as in today’s world. A number of factors have driven and resulted in this trend.  

First and foremost, the needs of the market.  Dedicated specialists in breast or gastrointestinal diseases are commonplace in medicine including the radiologist, surgeon, oncologist and radiation therapist.  Pathology appropriately responded accordingly. Secondly, pathologists have driven some our own marketing and needs.  The resulting potential problems are paramount and based on many fallacies.