Anatomical Predictions using Subject-Specific Medical Data Marianne - - PowerPoint PPT Presentation

Anatomical Predictions using Subject-Specific Medical Data

Marianne Rakic, John Guttag, Adrian Dalca

Evolution is captured by deformation field, 𝜚!:

Short Paper, MIDL 2020 Anatomical Predictions using Subject-Specific Medical Data 2

x0 φv + ✏ = xt

Baseline scan, 𝒚! Follow-up scan, 𝒚" Deformation field, 𝜚#

Model

Baseline scan, 𝒚! Follow-up scan, 𝒚" INPUTS Attributes, 𝒃!

November 26, 2019 Anatomical Predictions using Subject-Specific Medical Data 3

𝑤𝒚!→𝒚"

Baseline Scan (𝒚!) Velocity field Deformation field

Spatial Transform &

! $

Integration Layer

U-Net

𝜚𝒚!→𝒚#

𝒃𝟏 Concatenate

Attributes Follow-up Scan (𝒚") (𝒚! ∘ 𝜚𝒚!→𝒚")

𝑕&(𝒚!, 𝒃!, 𝑢)

Baseline scan Attributes

φv = gθ(x0, a0, t)

Model Architecture

• Use convolutional neural network to predict 𝜚$

Short Paper, MIDL 2020 Anatomical Predictions using Subject-Specific Medical Data 4

Our model variants with input: Our model variants with input:

Experiments: More external data helps

• Experiments using the ADNI dataset

Conclusion

• It is possible to predict anatomical changes
• Adding non image information helps
• In recent work, we have improved results
• Paper available at https://arxiv.org/abs/2006.00090

Short Paper, MIDL 2020 Anatomical Predictions using Subject-Specific Medical Data 5

Large variation among subjects

Baseline scan, 𝑦! Follow-up scan, 𝑦" Predicted scan, " 𝑦"