Google Chrome for Mac and Linux

Posted by Martin P in Development, Open Source      No comments

Sep
26

Having written a little about Google Chrome, it came as a disappointment to many that Google hasn’t produced (yet) a version for Mac and Linux.  Being Open Source, of course, it is entirely possible that others could step in and fill that hole.  Enter Cross Over Chromium.

Now it should be said that even the native Windows beta-version turned out to be too much of a pain in the ass for me to continue and after a week or so I went back to Firefox.  A port of Chromium via the emulation layer provided by WINE is likely to be even less appealing, and CodeWeavers themselves describe it as a proof-of-concept only – not to be used as a main browser.

Errors in speech recognition

Posted by Martin P in PACS General      No comments

Sep
25

A news item in AuntMinnie alerts us to the dangers in changing business processes without careful thought.  Errors introduced by the use of speech recognition made their way into the released report, increasing the error rate compared to traditional transcription.  While the study has limited relevance -

…its value is limited by the fact that the reports represent the week of May 20-27, 2002. Major improvements have been made to speech recognition software in the past six years. Additionally, speech recognition had only been in use at Aberdeen Royal Infirmary since 2001, and while properly trained, the radiologists were relatively inexperienced using the technology.

- it does reinforce a previous AM item on the same subject.  But the question is not with the technology.  Yes, the technology has improved beyond recognition (pun not intended) over the last 5 or 6 years, but nobody seriously expects a zero error rate. The issue is in the Quality Control process, which is well understood in the context of the wider record-keeping community.

A periodic audit of a relatively small number of radiology reports, such as five to 10 reports per radiologist, can identify significant voice recognition error patterns among the group and by individuals, and can assist in efforts to mitigate these problems

… is good practice, as is regular training refreshers (we all get into bad habits), but an issue that isn’t mentioned in the AM articles is in the process (often, not all the time,) to self-review prior to sign-off. The transcriptionist performs one thing well that automated systems don’t – s/he knows when a string of text is complete garbage – as with many of the errors that subsequently slip through the review.

Does the answer lie in a pre-review, prior to the sign-off review by the reporting Rad, to identify glaring errors? A double-blind approach is taken in many organisations when QC-ing document scanning operations. But this may have issues:

• Is the knowledge that a report has already been checked by someone give the Radiologist less incentive to pay full attention to the sign-off review?
• Was the system not paid for through the reduction in transcriptionists? Admittedly amongst other factors, but one of those factors includes reduction in error rate, surely!

The solution isn’t with technology – although the improvement of recognition engines helps, and other elements can facilitate a solution – at some point, its about people.