Removing NaN values from measurement table before computing feature_correlation_matrix

[marabuuu]

Currently, before computing the feature_correlation_matrix all rows with NaN values are removed (see documentation and code)

napari-accelerated-pixel-and-object-classification/napari_accelerated_pixel_and_object_classification/_custom_table_row_classifier.py

Line 264 in 68da4c9

correlation_matrix = pd.DataFrame(selected_properties).dropna().corr()

However, in my case I had the measurement moments_normalized in my table. This measurement had NaN values for each label (see screenshot). This means, all rows were removed and the feature_correlation_matrix only contained NaN values.

Maybe in some cases, it would make more sense to remove the column and not the row. But I am not sure if this would still be good scientific practice.

Best,
Mara