SMLC-NYSBC / TARDIS

Transformer And Rapid Dimensionless Instance Segmentation
https://smlc-nysbc.github.io/TARDIS/
MIT License
10 stars 2 forks source link

@GFabig asked for additional extension to compared Amira and Tardis #50

Closed RRobert92 closed 1 year ago

RRobert92 commented 1 year ago

The extension should compare Amira and Tardis spatial graphs and filter out only overlapping filaments.

Example:

Missing tasks:

RRobert92 commented 1 year ago

Minutes of the Zoom meeting held with @GunarF to discuss methods and suggestions for integrating a new package into Tardis v0.1.0-RC1.

Topic: A single spatial graph with splines that are a positive match between the 2 graphs will be produced when the package compares 2 spatial graphs.

Decisions 1) It was decided to give this package a separate entry and combine it with existence MT prediction (automatic detection see point 2)

2) MT prediction during the initialization will detect if the image file has an additional amira spatial graph file and run a comparison if so. If the comparison is conducted independently, each of the spatial graphs will need to have two prefixes entered in order to be recognized as belonging to the same data. E.g. Spatial_graph_cell1_amira.am ; Spatial_graph_cell1_tardis.am

3) Run independently will be initiated by:

tardis_compare_amira -dir str -o str -pf1 str -pf2 str

where:
-dir: is folder location with spatial graph
-o: is output folder location
-pf1: is prefix for Amira data set
-pf2: is prefix for Tardis data set

4) It was decided to limit user control to the comparison threshold.

5) Output file: A Single Amira spatial graph file containing MT annotated with 5 labels will be included in a single output file:

Label 1: MT taken from the Tardis (matches Amira) 
Label 2: MT taken from the Amira (matches Tardis) 
Label 3: MT with no matches (aka noise) 
Label4: MT in Tardis without match 
Label5: MT in Amira without match