Converted JPEG Image

[Anthony Leo]

Hello Everyone,

I am doing small project on Medical Diagnois using AI.
Since this dataset is relatively new , i want to try and learn from it.

However , i am finding it very hard to convert the .mxrs files to JPEG.

I tried converting the some of the files using Slide Converter to TIFF Format but the files are very big.
It does not even open in my desktop computer.

Any help will be great .

Thanks,

Anthony

[econdesousa]

Hi Anthony

Indeed they are huge files. You can choose one of the following possible options to overcome this problem:

• use 3dHISTECH software to generate huge tiffs and then tile them at reading time 
• use OPENSLIDE (https://openslide.org/api/python/) to load images (also tiled) directly to python (or c++ or any other lang you use)
• use QuPATH to load, annotate and export just the relevant part of each slide to smaller tiff images

[Vincenzo Della Mea]

However , i am finding it very hard to convert the .mxrs files to JPEG.

I tried converting the some of the files using Slide Converter to TIFF Format but the files are very big.
It does not even open in my desktop computer.

there is no way of making them smaller, and you have to manage them as such; read about digital slides to understand their features. If your project is an educational one, this dataset is not particularly recommended because it is large and also in a format that requires some extra effort. A much more manageable alternative that is good for learning is the ICIAR Bach Challenge. 

[Andreas Turzynski]

Hello Anthony,

with the 3D Histotech Converter you first have to annotate a small region of interest in the WSI using the CaseViewer. Depending on the desired final magnification the annotation area should be limited. The annotated area is then to be converted from mrxs to mrxs using the Converter. The resulting mrxs dataset has then to be converted from mrxs to tiled tiff. You can choose different down scaling factors in the software. To achieve a readable tiff file depends on the annotation area and the downscaling factor.

Alternatively, you can check out the recent PANDA challenge on kaggle. There you can find some very helpful public notebooks for the usage of openslide.  

Kind regards

Andreas

[Anthony Leo]

Thanks Andreas . I will have a look at the PANDA challenge on Kaggle.

Regards,

Anthony

[Anthony Leo]

Thanks Vincenzo. 
I was bit curious on the data format and want to start from the original source of the dataset .
But looking at the file size , i think i am better off working with cleaned dataset like  https://www.kaggle.com/c/siim-isic-melanoma-classification/data?select=jpeg .

Regards,
Anthony