![]() |
KnowBrainer Speech Recognition | ![]() |
Topic Title: Random Review of DME (Dragon Medical Embedded) in Meditech Expanse Topic Summary: With personal workarounds, hotkeys, thoughts, etc. Created On: 12/28/2020 12:10 AM Status: Post and Reply |
|
![]() |
![]() |
- ax | - 12/28/2020 12:10 AM |
![]() |
![]() |
- ax | - 12/28/2020 01:02 AM |
![]() |
![]() |
- ax | - 12/28/2020 01:16 AM |
![]() |
![]() |
- ax | - 12/28/2020 01:37 AM |
![]() |
![]() |
- ax | - 12/28/2020 11:11 PM |
![]() |
![]() |
- kernman | - 12/30/2020 08:59 AM |
![]() |
![]() |
- Lunis Orcutt | - 12/30/2020 01:36 PM |
![]() |
![]() |
- ax | - 12/30/2020 03:52 PM |
![]() |
![]() |
- ax | - 12/30/2020 04:26 PM |
![]() |
![]() |
- dilligence | - 01/03/2021 10:05 PM |
![]() |
![]() |
- kernman | - 12/31/2020 08:57 AM |
![]() |
![]() |
- Noimus | - 01/03/2021 02:46 PM |
![]() |
![]() |
- MDH | - 01/03/2021 04:20 PM |
![]() |
![]() |
- ax | - 01/03/2021 09:10 PM |
![]() |
![]() |
- kernman | - 01/05/2021 08:41 AM |
![]() |
![]() |
- R. Wilke | - 01/05/2021 02:46 PM |
![]() |
![]() |
- dilligence | - 01/05/2021 08:08 PM |
![]() |
![]() |
- mppcarey | - 01/08/2021 09:03 AM |
![]() |
![]() |
- dilligence | - 01/08/2021 09:52 AM |
![]() |
![]() |
- mppcarey | - 01/08/2021 06:08 PM |
![]() |
![]() |
- dilligence | - 01/08/2021 06:57 PM |
![]() |
![]() |
- ax | - 02/07/2021 06:10 PM |
![]() |
![]() |
- Mav | - 02/08/2021 04:36 AM |
![]() |
![]() |
- ax | - 02/08/2021 09:44 PM |
![]() |
![]() |
- Mav | - 02/09/2021 04:08 AM |
![]() |
|||||
What is it? It is the "dystopic" future that is the present.
Does it work? Well enough under the Chrome browser which is the medium/wrapper/skin for Meditech 6.x. The lack of "snappiness" in browser response in general is the biggest "meh". Aside from the usual network/server factors, responsiveness is perhaps partially dependent on the client side Javascript performance. Therefore, despite that actual speech recognition is rendered almost entirely local device indedpendent, a poorly spec'd client may still fall short of delivering a good experience.
Full Select-and-Say? That is the point.
Is it "better" than DMO? Yes and No.
Yes if you use a Chromebook or Macbook. On these non-Redmond platforms DMO/DMPE are simply unavailable. DME works on those as it is as "cloud-based" as it gets. However, Mic control would have to be through the on-screen virtual button. It wouldn't suprise me if workarounds to physically control SpeechMike buttons exist for Macbooks.
No as it is super-restricted since it is browser only (actually the Expanse browser only ... doesn't work in any other Chrome window). It hasn't even got the step-by-step commands which DMO and MModal have. It, of course, has the basic "Select-and-Say" commands and rudimentary vocabulary commands. We acknowledge the "Select-and-Say" commands are well implemented and just work. It doesn't provide any dictation box per se.
[Edited] Does it require any installation? Two main ones come to mind (both are "portable" as in NOT requiring admin privilege to install):
1) Obligatory Chrome browser extension called "PowerMic Web Extension", which basically allows for any Mic to "speak" into Meditech Expanse in a browser. Despite the PowerMic moniker, a PowerMic or any handheld is not required for this browser extension to work. It is also "cross-platform" with the same extension being used for Chromebook or MacBook, which allows for even the built-in microphone to work through the on-screen virtual button (or through keyboard shortcut - see later).
2) Windows-only, on-device Mic plugin called "NuancePowerMicWebAdapter", which allows for the use of selected buttons (more on that later) on PowerMic, Philips SpeechMike, and Grundig (SonicMic ?). The download/installation of this "adapter"/driver is just prompted once, upon detection of a PowerMic/SpeechMike as input device. For those who ready up Chrome from outside the hospital to use DME for the first time, if this plugin is missed, Meditech Expanse often does not prompt you again. This is a common pitfall which leads to the complaint that PowerMic record button doesn't work.
NB: "PDCC" is not required to drive the 3 buttons on Philips SpeechMike which the above "PowerMic Web Adapter" recognizes, and its simulataneous use might even lead to conflicts (I never downloaded the PDCC so I didn't test).
Curiously, when a SpeechMike (SMP 3700/4000) is plugged into a PC for the very first time, its mic gain defaults to 0 for some reason. Nothing remotely deal-breaking. Just something to be aware of.
Do I like Dragon Medical "Embedded"? It's "free" ...
Have there been "Phase IV" hardware/software integration problems encountered post-deployment? There have. I won't go into details. We will say that from strictly a user perspective the whole system works as well as could be expected on Windows 10 "fat clients".
I will focus on WHAT I DO to get around some of the limitations.
|
|||||
|
|||||
![]() |
|||||
First off, even with the second Windows-only utility mentioned above, PowerMic buttons still DON'T all work under DME. Only the most important record button and 2 field navigation buttons work, as I recall. For those who are on this platform and mildly interested, refer to this Nuance document from 2019:
I have seen one colleague demostrating that, if he is concurrently logged onto MModal, pressing the PowerMic recording button in the Expanse browser results in dictation transcribed twice - first from MModal (which takes up less bandwidth due to compression, similar to DMO, while sans benefit of Select-and-Say), second from DME (more bandwidth-intensive as the speech data is sent off apparently "uncompressed").
P.S., I want to give a "shout-out" to the maker of OUC - Octopus USB Controller, who also makes the vaunted TableMike (although as far as sheer prowess is concerned, the SpeechMike/PowerMic rigged to a Lamicall LS05 Gooseneck, controlled by keyboard, blows all others out of the water ... notwithstanding the TableMike/SoundTech winning the svelteness contest).
The Octopus can program PowerMic buttons just as competently as Nuance's own PMIIControl.exe, and is more versatile. OUC delivers button programming with more granularity with "single press event", "double press event", "long press event", and "release event". Apparently not all controls are availale to each button, however.
Moreover, it claims an ability to "learn buttons" on a variety of USB HID, in ways the AutoHotKey-based HID control utility (AHKHID) could not. In fact AHKHID could not recognize any of the PowerMic buttons other than the ones defaulting as mouse buttons. I have little doubt that OUC could program SpeechMike buttons just as adeptly. I still have not installed Philips' own "PDCC" software as I read (from this forum) that it is buggy and also as I don't need it.
As the OUC is trialware, I did not test it extensively. |
|||||
|
|||||
![]() |
|||||
Saving grace for me personally has been the Meditech coded "short-cuts", which they call "hotkeys":
I do feel it is "insane" to call these keyboard shortcuts "hotkeys" as no one in their right mind would use them as "hotkeys".
Another saving grace was that these commands can be sent "out of focus" to a given Chrome window, thus enabling shoehorning the setup into a simulated "hidden speech box", especially with the testing session one is given. Otherwise one can only run a single DME session at any given time. |
|||||
|
|||||
![]() |
|||||
AutoHotKey to the rescue then:
There are some side benefits to the remapping. Heck, a lot of benefits. With a tiny bit more "transfer codes" bound to hotkeys I got my "hidden speech box", too, mainly by simple clipboard operations (very very AHK 101 stuff).
To make the text transfer close to 100% reliably across Citrix (as opposed to 95% of the times), I couldn't see any way around having to send a "dummy character" before the clipboard content. Hence IMO no citrix is still the best Citrix.
|
|||||
|
|||||
![]() |
|||||
A few random operational observations: |
|||||
|
|||||
![]() |
|||||
I am interested in this. I believe our hospital will be going to this in the future. What is your specialty. I know in pathology we use a a lot of synoptic reports and DMO just does not cut it. I am still using Dragon medical practice edition. I am afraid that is going away soon though.
|
|||||
|
|||||
![]() |
|||||
Nuance will officially discontinue DMPE 4 on March 31 but support it for an additional year. We will support DMPE 4 much longer ------------------------- Forum Mission Statement |
|||||
|
|||||
![]() |
|||||
I am in one of the Internal Medicine subspecialties with a fair share of procedures. As such we are a major "customer" of anatomical pathology. I am at the receiving end of synoptic reports from our pathology colleagues. So I know a teeny bit of what you speak of.
However, you should elaborate if you could.
My quick feel is that if DMO "doesn't cut it", I am hard-pressed to see how DME is going to meet the challenge. DME is even more of a one-trick pony than DMO. The latter at least has some "step-by-step" commands. DME basically just replaces typing, while it is nicely implemented with minimal footprint. Nothing more and nothing less. And it runs on colleagues' Macbooks and Chromebooks without going through Citrix, which they tend to be pleased with.
The boilerplates provided by DME are better than nothing. They can't do hotkey-programmable subject/pronoun substitution and the likes. At least they can bring up fields which allow for some interaction.
The real workhorse for any "synoptic reporting" will IMO be provided by the EHR itself. I thought something like PowerPath serves up their own. Epic, fwik, gives you several different autotext/templating methods. Currently my own "templates" are based off of Meditech "Typicals". These "Typicals" cannot be launched by voice. But once launched, DME works nicely on them, with full Select-and-Say control.
Pg. 47 of late Larry Allen's Scripting for Dragon 11 book talks about "Dialogue Commands". That kind of workflow is what I think "synoptic reporting" ought to be. I never implemented any Active Scripting myself. But skimming through that book, and especially the section on "Dialogue Commands", drove me to learn about AutoHotKey's GUI commands, and made productive use out of them.
Thus to me any good synoptic reporting should be "form-to-letter". The regrettable kind, unfortunately also the most commonplace, which many pathologists and radiologists, and certain allied health providers use, are full of:
xxxx: no yyyy: yes
It detracts from readability and "comprehensibility". That's why we all flock to the sections of "conclusion" or "summary" or "comments". Anyway, I digress. Our pathology and radiology colleagues will say that it is the "practice standards" or medicolegal considerations that are driving them to do this.
One of the best, most accessible, and above all, as-PORTABLE-as-it-gets platforms for generating "form-to-letter" is actually the web browser. I found that out myself, by dipping my toes in Javascript. Again there are other options. MS Word with a hefty dose of VBA will do it. But the browser gives you the lowest barrier to entry. Mark that "no barrier to entry".
Therefore I could see why Meditech chose the browser as the medium. Can't beat Epic in either depth of pocket or market share, piggy-back on Google! The actual click-click-combined-with-select-and-say Meditech templates aren't bad to navigate at all, owing to the power and versatility of Javascript I am sure.
Without knowing your specifics, I would suggest the following:
1. Don't give up on your DMPE, especially if you already got all your templates for "synoptic reporting" and what not, in PowerPath and what not.
2. Make that particular EHR you (or your institution) are using take the primary responsibility in delivering "form to letter" templates. That's of course a bit easier said.
|
|||||
|
|||||
![]() |
|||||
Lastly, if you are forced to migrate to DMO, you might actually want to take a look at something such as Speech Productivity. Pathologists usually sit in front of 1 or 2 computers and can often convince IT to install software that they insist on, as a group. The rest of us who roam from workstation to workstation usually have to be content with a "portable" life.
P.S., Seems as though formatting is especially a strong suit with the SP take on dictationbox. However, the real crux is not whether any box can do formatting, but whether any formatting is transferrable with the text into the EHR.
For myself, the primary reason I stick with the Meditech "Typicals" is precisely because it reasonably preserves formatting with respect to fonts, bold/underline, etc. |
|||||
|
|||||
![]() |
|||||
|
|||||
![]() |
|||||
Our governing body "CAP" Requires synoptic reporting and actually supplies examples. That being said I understand your frustration. The biggest problem with DMO is that you cannot use formatted text. I realize that DME will fall short of our expectations. I hope to continue using DMPE until I retire in a couple of years.
|
|||||
|
|||||
![]() |
|||||
How was a your experience with M-Modal Fluency Direct? Our hospital offers that for us through EPIC integration, however it is limited to just that. At one point we had the actual software installed on the PCs and I felt like it was much faster and responsive that way. I was consider M-Modal, or going with DMPE 4 for my own personal use without having to rely on EPIC. Just curious to hear your experience with them. |
|||||
|
|||||
![]() |
|||||
For straight dictaion, MMODAL is very accurate. For creating and using custom commands, it is primitive compared to DMPE.
MDH ------------------------- |
|||||
|
|||||
![]() |
|||||
I echo MDH that MModal was to me as accurate in a raw sense as the current flavour of DME, with the proviso that I last used it in 2019. This is also the impression of colleagues who are using both (MModal and DME) on the same PC currently. There are some button sorting issues etc they have to contend with. Hence some of them just use their own solutions exclusively outside the hospital, exactly as you are thinking.
|
|||||
|
|||||
![]() |
|||||
Very interested if this can do formatted text dilligence....
|
|||||
|
|||||
![]() |
|||||
Bottom line: If it can be done with the tool from dilligence, it can be done with any other text editor. At the same amount of functionality, and very often at lower pricing, or also at not cost at all. -------------------------
|
|||||
|
|||||
![]() |
|||||
|
|||||
![]() |
|||||
I did install the newest version of speech productivity. Unfortunately when I issue the command "DB"Nothing happens. Any suggestions? ------------------------- Dragon Medical Practice Edition 4.3; Dragon Capture, Speechproductivity & Knowbrainer 2017 |
|||||
|
|||||
![]() |
|||||
|
|||||
![]() |
|||||
Yes I see them. Unfortunately has not been executing The "DB" Commands. I do see a open box which is labeled "dictation box V.6.252". I am just not able to execute the text translation nor can I get to the DB configuration box. Any suggestions?
------------------------- Dragon Medical Practice Edition 4.3; Dragon Capture, Speechproductivity & Knowbrainer 2017 |
|||||
|
|||||
![]() |
|||||
Hmm... could be a couple of things but most probably you have the following Dragon setting enabled?
This will require you to say "click" to approach the SP buttons and menus. I.e: "click TRANSFER", "click Configure" etc.
I recommend disabling this option. But since that's not for everyone you can now quickly turn it on or off with the following two voice commands (kindly provided by © 2020 PG Service):
Require Click Enabled Require Click Disabled
In addition, as a rule of thumb it's always good practice before applying SP 6 PRO updates to kill the Dictation Box process by saying "Kill Dictation Box" and then run SP Cache Manager (say "Cache Manager") to clean out obsolete SP Temp files:
If you then reboot your system and install the new version everything should work like a charm again. But usually it's enough to just run Cache Manager and adjust the above mentioned Dragon setting.
Let me know if that works. ------------------------- |
|||||
|
|||||
![]() |
|||||
A bit of follow-on:
Two eminently tangible benefits to interfacing through Chrome: 1. Enabling roaming or "follow-me" capability WITHOUT relying on VDI - whether Citrix or Horizon View (by VMWare). I was a bit derelict in neglecting to mention this biggie in my review above. Contrasting with my own EMR that I have to access through Citrix (it's another story why some of us have no choice but to contend with both an "EMR" and an "EHR"), the advantage of browser-based roaming is basically overwhelming. If roaming outside the hospital network, a trendy two-factor-authentication is required. 2. Chrome or any modern browser opens up possibility of some degree of automation through CSS Selectors or JavaScript. I am just starting to play with it, and getting to see what is conducive. |
|||||
|
|||||
![]() |
|||||
Hi!
Some additional, more technical information about DME from what I see (behold: I only can throw in my technical expertise from what I see in this thread without having 1st hand access to DME).
DME seems to be the Browser edition of Nuance SpeechKit SDK used by Meditech.
You'd have to include some javascript files and pass some cookies, but then you can dictate into most of the editors usually being used on web pages (e.g. <input> or <textarea>). You can also move focus to a textbox using a voice command (e.g. "go to findings") quite easily by assigning a conceptual name ("findings" in the example) to a given text field without any need for further programming.
Meditech could add additional voice commands if they want, but the user himself cannot add custom commands outside the integration. These commands can contain lists similar to DMPE, but you cannot specify full-blown speech grammars and also have to insert a short pause before the command.
SpeechKit can use the exact same servers DMO uses, so recognition accuracy would be identical if they got identical audio quality. DMO performs compression on its audio data before sending it to the server, that's why you need so little bandwidth. SpeechKit Browser SDK seems to stream uncompressed raw PCM to the server (thus the insane bandwidth requirement of 1Mbps for each parallel speaker!). Some browsers also had problems recording audio a while ago and with the high bandwidth requirements, loss of audio is more likely than in DMO. So in theory, DMO should perform better and more stable than DME.
One adidtional "Meh" is the ugly "SpeechBar" which cannot even be customized in a meaningful way anymore. With the Browser SDK you can at least embed the SpeechBar on your page, but with the Windows Desktop SDK you always get an additional ugly window flying around your desktop and outside of your integration
Regarding support for client-side hardware, DME suffers from the same hardships every browser-based solution suffers from: Access to custom hardware depends on specialized drivers available on the client in combination with special browser extensions. In combination a very costly and error-prone mechanism.
Perhaps knowing a bit more about the technical background of DME lets you know more easily what to expect from the solution.
hth, mav |
|||||
|
|||||
![]() |
|||||
Thanks Mav for chiming in some technical background. What you put forth jibe with the Meditech FAQ I linked above and largely in keeping with first-hand user experience. The 1 Mbs/user bandwidth requirement in today's broadband reality is no problem at all. Overall reliability is not an issue - outside of any VDI environment. Because some DMO setups save money by offloading microphone input to "PowerMic Mobile", which are iPhone/Android apps, colleagues who use both report that the DME setup I described is in fact more accurate. Then there is the operability on Mac, which is scoring unexpected points. Perhaps we should expect it. Docs wanna be jocks. Jocks like their Macs. Regarding "additional voice commands", Meditech definitely chintzed out. Indeed I believe that Meditech could have easily encoded many more keyboard shortcuts. Next to Epic's liberal use of keyboard shortcuts, Meditech Expanse is utter fail in this essentially a "1.0" offering. The number of keyboard shortcuts on Expanse was less than on the previous versions of Meditech. It totals exactly 4, which pertain to speech and boilerplate field navigation, as I described above.
From mwarddoc's description, in the area of short-cut enablement, Epic puts this version of Expanse to shame by a million miles. Fortunately, as I have been discovering, CSS Selectors (with a teeny bit of javascript) can salvage the situation to some extent. Productive DIY through CSS Selectors with combinators is fairly within reach for average doctors (willingness and motivation aside). CSS selectors are no more than glorified "wildcard" searches. Now "injecting" the Javascript/CSS Selectors through keyboard shortcuts is a different matter altogether. One can "get by" with bookmarklets. And there are popular plugin "monkeys" to inject in style. But in order to ensure maximal portability, I dipped into snippets of AHK codes that are quite a few rungs above my pay grade and mental capacity (based off of MSAA?). But it works like a charm. Once this desktop "pseudo-code-injection" has been worked out, there is very little barrier to further driving the automation by voice. In fact I am toying with the idea of getting a copy of DPI or whatever individual Dragon just to do commands. But then it defeats my own desire to maximize "portability".
I am also looking into WSR for this purpose. So far this is not yet urgent as long as I have my own javascript-based DIY shortcuts on hand. Indirectly, for me DMO would have been better. With DMO I wouldn't have to even think about using WSR to "walk that last mile". But as MDH had said repeatedly, for the "95%", despite being a one-trick pony, DME under Expanse satisfies in 2 important ways: Ma, look: no Citrix! Pa, look: runs on my shiny Macbook! What the "hail" are yo waing for? Git it, son! |
|||||
|
|||||
![]() |
|||||
Using WSR just for voice commands is a viable solution for me (although, of course, platform dependent).
Activating WSR and creating a few command grammars is no rocket science and you don't need any third party tools or cloud services. I even created a small application to turn on Dragon's microphone when it's off - it's been a matter of an hour, approximately.
On the other hand, HTML5's Speech Recognition API works at least in Chrome, so this could be a platform-neutral way (even though I must admit I don't know if Chrome on Mac uses the same Chromium engine. Under iOS they have to use the Webkit innards which means that many modern APIs don't work).
hth, mav |
|||||
|
FuseTalk Standard Edition v4.0 - © 1999-2021 FuseTalk™ Inc. All rights reserved.