Definiens Community - User Forum

Customized feature dll that works in 32 bit does not work in 64 bit?

fritzdev — Thu, 16 Feb 2012 23:04:35 +0100

"Stanislav" wrote:

It looks like different versions of CRT library are used for our software and your dll. In this case, the memory for DString is allocated by one version of CRT and deallocated by another.

Which version of compiler do you use?

Do you use "manifest" for your dll? Could you try to disable it? In this case OS will use manifest from exe file and find correct version of CRT.

Our original features DLL was compiled with VS2005 (release mode). I recompiled with VS2003 (release mode) and was able to get it working in both 32bit and 64bit windows. Using depends I can can see that our DLL and definiens DLLs are now both using the same version of the CRT. I can only assume we were getting lucky with the VS2005 compiled version under 32bit windows (ie sysWOW64 better detects memory corruption). Thanks

Customized feature dll that works in 32 bit does not work in 64 bit?

Stanislav — Tue, 14 Feb 2012 18:20:38 +0100

It looks like different versions of CRT library are used for our software and your dll. In this case, the memory for DString is allocated by one version of CRT and deallocated by another.

Which version of compiler do you use?

Do you use "manifest" for your dll? Could you try to disable it? In this case OS will use manifest from exe file and find correct version of CRT.

H&E .scn files+Developer

amalysa — Tue, 14 Feb 2012 16:52:36 +0100

Hi,

I need help figuring out how to process 40x H&E stained .scn files using Developer to batch process a large number of slides. What I would like to do (for now) is the following with these images:

i) detect the nuclei and determine their shape and sizes
ii) determine nucleus to cytoplasm ratios

However, I'm having a really hard time gettin past the segmenting stage as well as classifying the different parts of the tissue. For instance, when I open the high quality .scn files in Developer 2.0 and try to segnment it, the process takes hours. I was recently told that I should make a map of the low quality image and then the high quality image in after.

When I've segmented the tissue, classifying the different cells has become almost impossible because I don't know what features to use to be able to detect differences to classify different objects.

I know I've just asked for a lot of help but I really am lost as to how to get past these points in my analyses. Does anyone have any experience with working i) on high quality .scn files, and ii) analyzing H&E stains using Developer. My background is basic in working with Definiens and writing rulesets.

Thanks for taking the time read my message and for all your help

Customized feature dll that works in 32 bit does not work in 64 bit?

fritzdev — Tue, 14 Feb 2012 16:38:50 +0100

"Stanislav" wrote:

You need to recompile your dll with 64bit compiler. Unfortunately 64bit processes cannot load 32bit dlls.

Thanks, but I am running the 32 bit version of Definiens (over SysWOW64). I have been able to debug our DLL up to a point. There is a function call for "GetShortName" that has some local DString variables generated from a DValue::ToStr calls. When they go out of scope I get the following message:

Debug Assertion Failed! _BLOCK_TYPE_IS_VALID(pHead->nBlockUse)

This is generated on the delete[] data call in the DString destructor. I've found some documentation that suggests this is due to double deletion, yet a DString I create does not have this problem.

Still not sure where to go from here.

Customized feature dll that works in 32 bit does not work in 64 bit?

Stanislav — Tue, 14 Feb 2012 15:15:21 +0100

You need to recompile your dll with 64bit compiler. Unfortunately 64bit processes cannot load 32bit dlls.

Customized feature dll that works in 32 bit does not work in 64 bit?

fritzdev — Thu, 09 Feb 2012 17:01:18 +0100

We have a customized features DLL that has been working under 32 bit Windows XP with Definiens Developer 7. We have installed the same Definiens Developer 7 on Windows 7 64 bit and copied the same features DLL to the plugins folder, but when we try to load the rule-set we get a simple dialog box with the message "Could not load the process". What does this error message mean?

I have been able to debug things to the point where I see calls are made into our DLL to create the plugin and set parameters. There are also calls to get Short Name, etc. and then we get the error dialog. The last call I could trace was Info::GetName. Any help resolving this issue would be greatly appreciated.

DAB stain area greater than ROI area?

kaihartmann — Thu, 19 Jan 2012 13:32:52 +0100

Hi Olu,

I have just seen a ticket from Raj in our support portal. You don't need to send any files or open a new ticket. We will have a look at it.

Best wishes,
Kai

DAB stain area greater than ROI area?

kaihartmann — Thu, 19 Jan 2012 13:12:39 +0100

Hi Olujimi,

1) The export for ROIs is not correct. This is a bug and will be fixed with the maintenance release scheduled for end of this week. Please contact support(at)definiens[dot]com if you encounter this problem again.

2) The output should be in these folders:
results\Statistics\Custom
results\Statistics\Custom\ObjectStatistics.report

If you don't find it there, maybe you can send the solution (.dax) file to support(at)definiens[dot]com with a short description such that we can have a look at it.

Best wishes,
Kai

Supported image formats in eCognition 8.7

Johannes — Tue, 17 Jan 2012 10:35:27 +0100

Hi Mustak,

you have entered the online community for biomedical image analysis using Definiens Developer and Tissue Studio. Since your field of work seems to be geospatial imaging, I would recommend to post your questions in the online community for users of the eCognition software, community.ecognition.com.

Best
Johannes

DAB stain area greater than ROI area?

Olujimi — Sat, 14 Jan 2012 00:44:08 +0100

Hi,
I am new to definiens. I ran my first DAB analysis for an antibody that stains both inside cells and outside.
1. In result output that I got lists a "sum area" of staining that is greater than the region of interest (ROI) area in square microns. How is this possible? 2. Although I used a custom export for Marker Areas, none of the output data shows marker area. What am i missing?

Supported image formats in eCognition 8.7

mustak — Fri, 13 Jan 2012 04:27:39 +0100

Hi there

Does anyone know whether ERMapper ECW image format is supported in eCognition 8.7? I was using ecw images in the earlier version (8.6) but the new version is rejecting ecw images.

Any suggestions/advise appreciated.

Thank you
Mustak Shaikh

Job Offer at Definiens: Technical Support Engineer for our new office in the Boston area

DanielS — Mon, 02 Jan 2012 12:01:00 +0100

Definiens Inc. - for our new office in the Boston area

In this exciting role you will be responsible for covering the whole support process for the world’s most innovative image analysis technology and software platforms.
We expect you will have experience working in technical customer support with a “hands on” attitude. An initial, 3-4 weeks trip to the company’s Munich, Germany headquarters for training may be required.

Responsibilities

Provide client support and technical issue resolution via E-Mail, phone and other electronic medium.
Track, prioritize and document requests using a support request system.
Reproduce customer problems in-house, gathering of all necessary data for tracking issues at customers.
Planning and support of installation at customers including pre and post sales activities.
Supporting integration into customers IT environment.

Qualifications & Skills

Excellent knowledge of Networks and Administration for both Windows and Linux.
Ability to understand clients network setups including security systems.
Experience with software licensing, ideally Flexera.
Knowledge of modern architectures and technology including scripting technologies.
Professional experience in customer service business is an advantage.
Highly self-motivated and independent.
A strong interest in image based or image analysis applications (any previous experience would be a bonus).
Ability to work in a fast-paced, small company.
Strong verbal and written communication skills in English.
Readiness to travel and to work on customers’ site.
Flexibility with working hours with respect to our world-wide customer base.

We offer

A challenging and rewarding position working with customers with our pioneering software and technologies.
A highly motivated, international and well-qualified team.
A stimulating cosmopolitan environment.
Excellent working conditions with comprehensive benefits and ongoing career and professional development.

If you are interested in a career at Definiens, please send your CV indicating your desired position and salary expectations to:

Human Resources
Definiens AG
Bernhard-Wicki-Straße 5
D-80636 München
Germany

jobs(at)definiens.com

http://www.definiens.com/about-definiens/career/technical-support-engineer-mf.html

H&E alone Vs cytokeratin for tumour detection

kaihartmann — Fri, 09 Dec 2011 11:53:06 +0100

Hello Peter,

You can use a 2-step approach: a combination of Composer ROI detection with the Nucleus Morphology Filter, which is new in Tissue Studio 3.0. This could give you a good distinction already on H&E between healthy and dysplastic cells. I would expect that there will be border-line nuclei between the two classes (healthy and dysplastic) that can not be judged by the morphology filter - mainly because the morphology is almost identical.

A cytokeratin stain could be used to improve the result of the initial Composer ROI detection, but would be of no effect to the second step.

Best wishes,
Kai

3rd International Definiens Symposium in Munich - June 14-16, 2012

Florian — Fri, 09 Dec 2011 11:50:33 +0100

Dear everyone,

we're happy to announce the 3rd International Definiens Symposium in Munich. It will be held from June 14-16, 2012 in the Definiens headquarters and the close-by Ernst & Young auditiorium. You're heartily invited.
Read more at: http://www.definiens.com/community/symposium.html

best regards,
Florian

Individual Tree delineation

Florian — Fri, 09 Dec 2011 11:46:41 +0100

From your question I assume that you might be interested in Definiens former geospatial business that is now part of Trimble. You will find all relevant information at http://www.ecognition.com/. It's still the same technology applied to both geospatial and biomedical images but we're separate companies now.

best regards,
Florian

Tissue staining for accurate cell segmentation

kaihartmann — Fri, 09 Dec 2011 11:46:01 +0100

Hello Peter,

If you would like to have exact cell boundaries, we recommend using a membrane IHC stain for use with Tissue Studio. In other cases, you can use the Cell Simulation function, which can either be based on a fixed distance from the nucleus or based on a whole-cell stain.

I am not aware of any customer that has used a reliable stain cocktail for multiple cell types (stroma, healthy and tumor cells). Maybe other users can comment?

Best wishes,
Kai

classification map

Florian — Fri, 09 Dec 2011 11:45:48 +0100

H&E alone Vs cytokeratin for tumour detection

pdcaie — Wed, 07 Dec 2011 14:45:45 +0100

Hello,

I was wondering if anyone had differentiated colorectal stroma from healthy epithelial from dysplastic (cancer) epithelial using H&E staining alone in TS3? Can this be done by teaching the programme to recognise these ROIs using the composer software?

If H&E is not specific enough would you need to apply a cytokeratin stain?

Thanks,
Peter

Tissue staining for accurate cell segmentation

pdcaie — Wed, 07 Dec 2011 14:17:34 +0100

Hello,

I am wanting to accurately and robustly segment both nucleus and cells in colorectal tissue using Tissue studio 3. Is H&E good enough for this or will I need a specific membrane/whole cell stain? Is there stains I can multiplex with H&E for this or is IHC the way forward along with a nuclear dye e.g.DAPI? As I am interested in stroma, healthy and tumour cells cytokeratin alone would not be suitable.

Many thanks,
Peter

classification map

sparkling106 — Sun, 27 Nov 2011 22:52:58 +0100

i have some doubts about how to merge my segments after make a multiresolution segementation, i choosed my samples, i run the classification +descirption classes, and now i want to merge the classes and create a shp to export in another GIs programm, how to do that?
thanks a lot

Individual Tree delineation

FlushKlaus — Fri, 25 Nov 2011 16:16:46 +0100

I am trying to delineate individual trees on a 8-Band World View 2 image.
I am thinking of the best way to perform this task:
appropriate methods could be:
- Window-based smoothing or normalisation filters;
- local spectral maxima as indicators of the likely location of a tree crown (Maybe on the panchromatic Band)
-delineated tree crowns using seed identification and a region-growing algorithm
-texture parameters like co-occurrence contrast, dissimilarity, and homogeneity texture measures.

I would need some help for finding the best approach, or some new ideas!
Thank you

Job Offer at Definiens: Software Developer

andreasprinz — Thu, 24 Nov 2011 10:38:31 +0100

To join our expanding Software Development team, we are seeking for our headquarters in Munich a highly qualified.

Responsibilities:

Develop, unit-test and maintain software components for the Definiens Developer Platform and related Definiens Solutions in Life Science Applications

Write, edit and revise detailed technical design documents to accurately describe the technical details of the components to be developed

Explain and document major application design decisions and their rationales to technical and non-technical audiences

Qualifications & Skills

Degree in Computer Science, Biomedical Engineering, Mathematics or Physics

Work experience in Software Development dealing with sophisticated and technically focused client server products

Superior in-depth knowledge of
o C and C++
o Win32 API and MFC
o Microsoft Visual Studio
o Linux Operating Systems

Professional background in at least one of the following areas
o Linux Environments / Cross Platform Development
o Middleware and Component Technologies
o Databases
o XML/SOAP/HTML

Ability to adapt well to changing priorities in a fast paced development environment

Ability to work efficiently as part of a development team with minimal supervision

An interest in image based or image analysis applications (any previous experience would be a bonus)

Strong verbal and written communication skills in English and German
Must be a team player

We offer

A challenging and rewarding position at the cutting edge of image analysis

The opportunity to work within a highly motivated and qualified team of experts

A stimulating cosmopolitan working environment in Munich

Excellent working conditions with comprehensive benefits and ongoing career and professional development

If you are interested in a career at Definiens, please send your CV indicating your desired position and salary expectations to:

Human Resources
Definiens AG
Trappentreustrasse 1
D-80339 München
Germany

jobs(at)definiens.com

Please also visit:
www.definiens.com/about-definiens/career/software-developer-mf.html

stain isolation (prototype)

bzengerlandolt — Thu, 24 Nov 2011 08:26:27 +0100

Hello Krissy,

the stain isolation (prototype) algorithm allows you to isolate the contributions of two different stains (in brightfield images).

For the perhaps most common case (?) of isolating Hematoxylin and DAB you can try entering cx/cy values for Hematoxylin (0.4/0.4) and DAB (-0.3/-0.3), which is a good starting point.
The staining layer (entered in Stain 1 parameter field) will then return values between 0 and 1, showing the relative contribution of one Stain to the overal optical density (returned to the layer in Stain 2 parameter field).
To obtain more intuitive staining layers, multiply the optical density layer once with staining, and once with (1 - staining), using the layer arithmetics algorithm.

But now to your question of how to estimate the cx and cy values in general:
(I have to warn that this takes a minute to implement, but it is worth the effort.)

Using the developer, you can create a customized object feature that evaluates cx and cy values according to the hue-saturation density transformation for every object.

For the hue-saturation density transformation we need to know the characteristics of the (white) background. This can be done

* by using a default RGB value of (255, 255, 255)
* by hardcoding a reasonable estimate for the slide of interest
* by using the mean RGB of an already identified image background

The corresponding numbers are best stored in scence variables, e.g.

calib_Rmax = 222
calib_Gmax = 230
calib_Bmax = 218

(These numbers also need to be entered later in the stain isolation (prototype) algorithm.)

From this we can compute an object's densities using customized features:

Density_Red = - ln ([Mean Layer 1]/[calib_Rmax])
Density_Green = - ln ([Mean Layer 2]/[calib_Gmax])
Density_Blue = - ln ([Mean Layer 3]/[calib_Bmax])

Using the following customized features, we can implement the hue-saturation density transformation as follows:

cx (HSD Transform) = ( [Density_Red]/([Density_Red]+[Density_Green]+[Density_Blue]) ) * 3 - 1
cy (HSD Transform) = ( ([Density_Green]-[Density_Blue])/([Density_Red]+[Density_Green]+[Density_Blue]) ) * 3^0.5

Now just click on objects of color 1 to find out typical cx/cy values for the first stain, and click on an object of color 2 to find typical values for the second stain. Or better yet, export the cx and cy value for all objects, and look at a cx/cy scatterplot (e.g. in Excel) to estimate the endpoints of the line on which your objects fall.

Hope this helps.
Good luck,

Barbara

stain isolation (prototype)

krissyburke — Tue, 22 Nov 2011 15:47:33 +0100

Can someone give me a brief description of the algorithm, stain isolation? What is the staining characteristics, cx stain1, cy stain1, etc...

Thanks,
Krissy

Aperio to Studio Annotations

GillianBrown — Mon, 31 Oct 2011 23:59:42 +0100

Yes, you did understand. I have begun to use ImageScope much more for preliminary observations, as you can have a virtual slide tray of the your session batch of slides, but we dont have Spectrum. Its just curiosity at the moment though.

assigned image layer gets lost when updating another image layer in an action

teinert — Thu, 06 Oct 2011 13:14:30 +0200

Hi Mark,

your problem and its origin is not completely clear to me. Can you provide us a ruleset exhibiting this problem so that I can forward it to our support team.

Kind regards
Thomas

assigned image layer gets lost when updating another image layer in an action

maasland — Wed, 05 Oct 2011 16:07:46 +0200

Hi!

I would like to split-up some of my actions. By doing so I would like to overwrite image layers in an action by applying one or more preprocessing steps (e.g. image smoothing). I.e., I would like to update for instance the layer Layer_Nucleus_Stain. When I do so, Layer_Nucleus_Stain is being updated by the preprocessing action but as a result also a image layer is created containing the original (non-processed) Layer_Nucleus_Stain. By this new image layer my assigned image layers (General Settings) get mixed up and I lose the initially defined assigments of Layer_Marker1_Stain.

How can I avoid this undesired >noname> image layer?

Thanks in advance!

Mark

Aperio to Studio Annotations

kaihartmann — Wed, 28 Sep 2011 09:54:54 +0200

Hi Gill,

At the moment, we do not support the Hamamatsu annotation format.

If you do annotations on Hamamatsu files in ImageScope, these can not be read into the current version. Is this what you mean? I can check - this might be an easy feature for the next release.

Thanks for the request.

Best wishes,
Kai

Aperio to Studio Annotations

GillianBrown — Tue, 27 Sep 2011 23:42:38 +0200

Does the Link to 3rd Party Annotations only work for Aperio Scans held in Spectrum? We have a Hamamatsu Nanozoomer and do not load into Spectrum, but it might be useful to mark up areas in the same way for processing in Studio,

Who can help me?

teinert — Thu, 18 Aug 2011 15:56:14 +0200

Hi dxqianwei,

Image object fusion is a very powerful algorithm, which allows the user to have maximum control on which objects should be merged.

There are three terms associated with image objects:
- Seed: The object, which is looking for partners (which are called Candidate) to merge with.
- Candidate: Objects that are considered to be merged with the seed. Whether the Seed and the Candidate are merged depends on the Fitting Function in the Image object fusion algorithm.
- Target: This is the resulting object when the Seed and the Candidate are merged.

I hope this helps you.

For further explanations I would like to refer you the Reference Book, chapter 7.3 on page 85.

Kind regards
Thomas