Dr. R2D2--Computerized pathologist?

Meet C-Path, the iJobPlodd version of the 21st century pathologist.

The December 3rd-9th, 2011issue of The Economist has an article in this "Technology Quarterly" section entitled, "Indolent or Aggressive: A computerised pathologist that can outperform its human counterparts could transform the field of cancer diagnosis."  Actually, this a is review/interview on an article (abstract) published on 9 November 2011 in Science Translational Medicine by Daphne Koller and colleagues from Stanford.

Unfortunately, it appears that the author of The Economist article failed to consult a practising pathologist to get some context for this study.  Here's an example:

LOOKING for needles in haystacks is boring. But computers do not get bored. Contracting out to machines the tedious business of assessing the dangerousness of cancer cells in histological microscope slides ought thus to be an obvious thing to do. Cervical-cancer smear tests aside, however, such electronic intrusions into the pathology laboratory are limited. Grading cancer cells into “indolent” and “aggressive”, and hazarding an opinion about whether they spell a treatable condition or an untreatable one, has remained the realm of the human expert.

Perhaps I'm being a bit sensitive, but I don't consider looking at a patient's slides "tedious business" and resent this characterization of my practice of medical care!  According to this logic, perhaps patients would be better served having computers undertake the "tedious business' of taking a history or pelvic exams or listening to the heart.  In brief, this is a cutting-edge research article--not a roadmap for radically changing pathology practice in general (or even for the narrow application for breast cancer) .  As an example, it looks only at breast cancer tissue microarrays--not real clinical specimens.

I absolutely do not wish to take away from this article, however.  It is brilliant!  As I re-read it, apart from being green with envy (imagining how great it would be to do this with my own pet, non-small cell lung cancer!), I am picking up more and more fine points of trying to do this kind of work.  Here are my two main points which make this worth recommending:

1. The image features to be measured were not predetermined by a pathologist: the system made a very large set of measurements of the epithelial and stromal components--and their relationships--of the tumor and then the program determined which of these were related to survival outcomes.
2. The confirmation of the crucial role of the stroma in tumor outcome.

The bottom line, after reading this article (which I highly encourage anyone interested in digital pathology or anyone planning on practicing anatomic pathology in the next 20+ years to do), is that pathologists can be reassured of having a "job."  But--the article points to how our job might be significantly impacted by digital imaging and robust image analysis (ahem!--without regarding to FDA, CAP, payment, etc., i.e. the real drivers for adoption and implementation of these technologies).