New tissue analyzer will probably need to do more


According to a report on the ScienceDaily web site this morning, a German company called Fraunhofer-Gesellschaft has developed a prototype of an analytic tissue processor that can reliably distinguish between benign and cancerous prostate tissue within 90 seconds. However, …

What the report does not tell us is whether this tissue processor is actually of any real, clinical value, because:

  • It still requires biopsy-based tissue samples on which to carry out the “black box” processing, and biopsies remain a “hit or miss” form of tissue sampling — even when carried out under MRI/TRUS guidance.
  • There is no mention anywhere of whether the processing system is able to distinguish with any degree of accuracy between indolent and aggressive tissue types for cancerous tissue (e.g., the Gleason grade of a specific cancerous area of tisssue).
  • There is also no mention of whether the processor can provide information on the quantity of cancerous tissue in a particular sample, thus allowing an estimate of the clinical significance of the tissue volume.
  • It is unclear whether tissue analyzed using the technology is recoverable for re-analysis if required (as is currently possible with a pathological sample on a slide).

We do not wish to suggest that this new tissue processor has no value at all. We are simply noting that, based on the information provided by the company to date, its value appears to be limited (at least in the pathological examination of prostate tissue for cancer). The report on ScienceDaily is simply a “pick up” of a media release issued by the developer of the processor. It is appears to be clear in the media materials (because there is a photograph of the processor’s “read out” available in those media materials) that all the machine is currently able to do is state whether a specific tissue is benign (non-cancerous) or malignant (cancerous).

The clinical utility of any technology that is intended to substitue for the skill of a pathologist in determining factors that influence the treatment of a man with prostate cancer had better include something that is at least as accurate as the Gleason score, which is known to be one of the most critical factors in guiding the urgency and aggressiveness of a patient’s treatment.

2 Responses

  1. This seems to me a significant step forward, if the claims are valid. Criticizing the device for lacking qualities it is not meant to have doesn’t seem quite cricket.

    I can see scenarios where this would quite helpful:

    1. As described in the press release, consider an office that lacks a pathologist (and possibly even lacks any physician at all) but has technicians who are competent to conduct a good biopsy sample and mount the specimens for the machine’s optical inspection.

    2. Consider a machine working 24 hours a day triaging specimens from a busy outpatient lab, with an API that allows apps to talk to it:
    — “Patient Q301. Twelve cores, all negative. Email physician with nonresults. Update patient’s chart. Prepare invoice.” Elapsed time: 20 minutes.
    — “Patient Q302. Twelve cores, two positive. Email pathologist, flagging the two specimens in question; cc physician. Update patient’s chart. Prepare invoice.” Elapsed time: 20 minutes.
    — …

    Generally speaking, machines are better than humans at very repetitive tasks whose results are usually negative. (Consider SETI.) This seems like such a task, at least in bulk.

  2. Dear Paul:

    But we have no evidence (at the moment) that once the samples go into this machine, they aren’t altered in ways that make them unavailable for later, accurate analysis by a qualified pathologist. Without such a guarantee, the mass processing you suggest is impossible.

    And with regard to the idea that any urologist’s office currently carries out any type of pathological examination is inaccurate. Even at somewhere as sophisticated as Johns Hopkins, the biopsy samples are sent to the pathology lab for assessment. No urologist is qualified to carry out such an assessment.

    I can’t see any point to this machine unless there are some very high levels of certainty that the processing and examination of biopsy samples using the (undescribed) techniques don’t alter the specimens in any way.

    I do agree with you that if there is no effect on the biopsy samples whatsoever (which, frankly, is hard to imagine), then the automated processing to identify the negative samples would be useful and valuable.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: