Table 5
From: Automated cytometric gating with human-level performance using bivariate segmentation
Input: | |
\(\left\{{X}_{1},{X}_{2},{X}_{3},\ldots,{X}_{{n}_{{training}}}\right\}\in {{\mathbb{R}}}^{m*n}\) :protein expression matrices for all training subjects | |
\(\left\{{Y}_{1},{Y}_{2},{Y}_{3},\ldots,{Y}_{{n}_{{training}}}\right\}\in {{\mathbb{R}}}^{m*n}\) :cell type label for all training subjects | |
\(\left\{{X}_{1},{X}_{2},{X}_{3},\ldots,{X}_{{n}_{{testing}}}\right\}\in {{\mathbb{R}}}^{m*n}\) :protein expression matrices for all testing subjects | |
Output: | |
\(\left\{{Y}_{1},{Y}_{2},{Y}_{3},\ldots,{Y}_{{n}_{{testing}}}\right\}\in {{\mathbb{R}}}^{m*n}\) :predicted cell type label for all testing subjects | |
\(\left\{{H}_{1},{H}_{2},{H}_{3},\ldots,{H}_{{n}_{{testing}}}\right\}\) :target cell population convex hull with 3D density map | |
UNITO Training | |
for i=1:ntraining do | |
\(\bar{{X}_{i}},\,{Y}_{i}\leftarrow {Normalize}\left({X}_{i}\left[{channel}1,{channel}2,{gate}\right]\right)\triangleright\bar{{X}_{i}}\in {{\mathbb{R}}}^{m*2},\,{Y}_{i}\in {{\mathbb{R}}}^{m*1}\) | |
\({D}_{i},\,{M}_{i}\leftarrow {Image\; Construction}\left(\bar{{X}_{i}},\,{Y}_{i}\right)\triangleright\)Converting tabular data to density and mask image | |
End for | |
UNITO Classifier←UNITO Network(D, M) | |
UNITO Prediction | |
for j=1:ntesting do | |
\(\bar{{X}_{j}}\leftarrow {Normalize}\left({X}_{j}\left[{channel}1,{channel}2\right]\right)\triangleright\bar{{X}_{j}}\in {{\mathbb{R}}}^{m*2}\) | |
\({D}_{j}\leftarrow {Image\; Construction}\left(\bar{{X}_{j}}\right)\) | |
\({M}_{j}\leftarrow {UNITO\; Classifier}\left({D}_{j}\right)\) | |
\({Y}_{j}\leftarrow {Mask\; Interpolation}\left(\bar{{X}_{j}},\,{M}_{j}\right)\) | |
\({H}_{j}\leftarrow {UNITO\; Visualization}\left({X}_{j},\,{Y}_{j}\right)\) | |
End for | |
Return \({Y}_{{testing}},\,{H}_{{testing}}\) | |
