v1.3.7: [Update] model output calibration

[Update] Model output calibration: We have implemented a new feature based on isotonic model calibration, which was fitted to different age bins. The "calibrated probability" of Alzheimer's disease is now estimated based on the raw CNN model output and the ratio of true positive to true negative cases in the normative calibration subcohorts.
For more information, see: Martensen, O.H., Strauß, T., Ramedani, R., Dyrba, M., (accepted). Comparison of post-hoc calibration methods for neural network likelihood scores, in: Handels, H., et al. (Eds.), Bildverarbeitung für die Medizin 2026. Springer Fachmedien Wiesbaden, pp. 443–449. https://doi.org/10.1007/978-3-658-51100-5_87

v1.3.6: [New Feature] model output calibration

[New feature] Model output calibration: We have implemented a new testing feature based on isotonic model calibration, which was fitted to four different cohorts. The "calibrated probability" of Alzheimer's disease is now estimated for new data based on the raw CNN model output and the ratio of true positive to true negative cases in the calibration cohorts.
For more information, see: Martensen, O.H., Strauß, T., Ramedani, R., Dyrba, M., (accepted). Comparison of post-hoc calibration methods for neural network likelihood scores, in: Handels, H., et al. (Eds.), Bildverarbeitung für die Medizin 2026. Springer Fachmedien Wiesbaden, pp. 443–449. https://doi.org/10.1007/978-3-658-51100-5_87

v1.3.5: Merge pull request #23 from J-Chandra-py/demo

• Improvement: Fixed a race condition in the GUI code (which caused the waiting label "processing..." to vanish too early when processing user-defined upload data)

Added German translation

• Added dictionary (JSON files) of Strings for GUI translation
• Added German translation

Final version used in Alz Res Ther paper

This is the version of the InteractiveVis used in our paper "Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: Evaluation in Alzheimer’s disease", Alzheimer’s Research & Therapy, DOI: 10.1186/s13195-021-00924-2

Changelog v1.3.3

• changed colormap to red-yellow-green

process user data using multiple processes

Changelog v1.3.2

• process uploaded user data using multiple parallel processes for the residualization procedure to reduce waiting time (Linux/MacOs only)
• disabled multithreading/ multicore processing under Windows due to problems

process user data using multiple threads

Changelog v1.3.1

• process uploaded user data using multiple threads for the residualization procedure to reduce waiting time

visualization procedure updates

Changelog v1.3.0

• added scroll wheel navigation to slice views
• changed scaling to fixed factor of raw values for better comparability across people
• changed visualization algorithm to LRP-CMP (alpha=1,beta=1)

Added user data upload functionality

Changelog v1.2.0

• Added user data upload functionality and modified GUI elements to provide covariate information for the uploaded scan

Substantially improved visualization performance

Changelog v1.1.0

• Improved (re)plotting performance by updating Bokeh's figures via ColumnDataSource instead of raw figure content
• Drawing of crosshairs via CSS in web browser