---
title: 'MoS Module: Adaptive Spline Upsampling'
url: https://www.emergentmind.com/topics/mixture-of-splines-mos-module
type: topic
---

# MoS Module: Adaptive Spline Upsampling

The Mixture-of-Splines (MoS) module is a content-adaptive, geometric-continuity upsampling operator designed for null-space estimation in the Dual-Prior Null-Space Learning (DP-NSL) framework for arbitrary medical slice super-resolution. MoS replaces black-box parameteric upsamplers (such as MLPs) with a mixture of analytic B-spline experts, each of configurable order, and leverages a dynamic routing mechanism to enable spatially-varying interpolation continuity. This ensures that homogeneous tissue is rendered smoothly while anatomical boundaries and fine structures retain sharpness, all under explicit measurement consistency constraints [2606.26716].

## 1. Mathematical Foundation of Mixture-of-Splines

At the core of MoS are tensor-product B-spline basis functions of various analytic orders. The zeroth-order (box) B-spline basis is given by
\[
\beta^0(x) =
  \begin{cases}
    1, & |x|<\tfrac{1}{2} \\
    0, & \text{otherwise}
  \end{cases}
\]
Higher-order B-splines are obtained by convolution:
\[
\beta^p(x) = (\beta^0 * \beta^0 * \cdots * \beta^0)_{(p+1)\text{ times}}(x)
\]
with the shifted, p-th order, i-th basis given by $B_{p,i}(x) = \beta^p(x-i)$. In 1D, this yields $C^{p-1}$ continuity.

In the DP-NSL framework, MoS appears within the null-space estimator as:
\[
V_{SR} = V_{\mathcal R} + V_{\mathcal N} = \mathcal U(V_{LR}) + \Pi_\mathcal N \big( f_\theta(V_{LR}) \big)
\]
Here, $\mathcal U(V_{LR})$ is a range-space upsampling anchor reproducing all observed slices and $\Pi_{\mathcal N}$ projects onto the null space of the downsampling operator, ensuring that only unobservable details are filled in.

## 2. MoS Module Architecture and Regression Mechanism

For each query coordinate $\mathbf q$ in the high-resolution target grid, MoS computes an upsampled feature vector as a weighted mixture:
\[
\mathrm{MoS}(\mathbf F_{LR},\mathbf q) = \sum_{k=1}^K \pi_k(\mathbf q) \; \mathcal U_{p_k}(\mathbf F_{LR};\mathbf q)
\]
where:
- $\{p_k\}_{k=1}^K$ are the selected spline orders (in practice, $K=3$, $p=\{2,3,4\}$),
- $\mathcal U_{p_k}(\mathbf F_{LR};\mathbf q)$ denotes expert $k$'s separable 3D B-spline interpolation of order $p_k$,
- $\pi(\mathbf q) = (\pi_1, ..., \pi_K)$ are adaptive weights parameterized by a routing network:
\[
\boldsymbol\pi(\mathbf q) = \mathrm{Softmax}\left(R_\pi\big(\tilde{\mathbf F}_{LR}^{\mathbf q}\big)\right)
\]
Here, $\tilde{\mathbf F}_{LR}^{\mathbf q}$ is a locally interpolated feature vector.

Each expert computes
\[
\mathcal U_{p}(\mathbf F_{LR};\mathbf q) = \mathbf c \odot (\mathbf b_x \otimes \mathbf b_y \otimes \mathbf b_z)
\]
with axis-aligned basis evaluations
\[
\mathbf b_d = \beta^p\left( (\Delta q_d - \mathbf k_d)\odot \mathbf s_d \right), \quad d \in \{x, y, z\}
\]
where the geometric parameters $\mathbf c, \mathbf k_d, \mathbf s_d, \mathbf o$ are functions of the local feature neighborhood, typically predicted by shallow convolutional sub-networks.

## 3. Content-Aware Order Selection and Continuity Control

Rather than fixing a global spline order, MoS dynamically adapts the mixture ratio per coordinate via the routing network $R_\pi$. Low-order experts (e.g., $p=2$) dominate in homogeneous (smooth) tissue, while high-order experts (e.g., $p=4$) are prioritized near anatomical boundaries and fine details. Empirically, transition zones leverage a blend of orders, producing spatially-adaptive continuity [2606.26716, Fig. 5].

A B-spline of order $p$ is $C^{p-1}$ continuous, providing the ability to specify locally the needed degree of smoothness. The composite MoS field remains globally continuous and regionally attuned, mitigating both the over-smoothing effect of high-order interpolation and ringing artifacts of low-order kernels, while preserving anatomical fidelity.

## 4. Algorithmic Workflow

The MoS upsampling process for a single coordinate is as follows:

```python
# Inputs: F_LR ∈ ℝ^{C×S×H×W}, query coord q∈ℝ³, spline orders {p₁,…,p_K}, routing net R_π, expert conv-nets E_p
# Output: high-res feature f_SR^q ∈ ℝ^C

1. tilde_f = trilinear_interpolate(F_LR, q)
2. [c, {k_d},{s_d}, o] = E_shared( neighborhood(F_LR,q) )
3. for k in 1…K:
4.     Δq = q - q_LR + o
5.     for d in {x,y,z}:
6.         b_d = β^{p_k}((Δq_d - k_d) ⊙ s_d)
7.     B = outer(b_x, b_y, b_z)
8.     U_k = c ⊙ B
9. π = Softmax( R_π(tilde_f) )
10. f_SR^q = ∑_{k=1}^K π_k ⋅ U_k
```
Parameters $\{c, k_d, s_d, o\}$ allow for local modulation of the spline basis. Parallelization is applied over coordinate batches in practical implementations.

## 5. Implementation Considerations

- **Expert Bank**: $K=3$ B-splines ($p=2,3,4$) are used, with analytic forms in the appendix of [2606.26716].
- **Geometric Parameter Net**: A shallow 3D CNN (e.g., $1\times1\times1$ and $3\times3\times3$ convolutions) maps the local $r^3$ neighborhood to interpolation parameters.
- **Routing Net**: Either one or two fully-connected layers, or $1\times1\times1$ convolutions, produce the mixture logits from a pointwise interpolated feature vector.
- **Loss Function**: The DP-NSL framework is supervised end-to-end using an $\ell_1$ loss on the reconstructed volume, $L=\|V_{SR}-V_{HR}\|_1$, without separate regularization for MoS.
- **Computational Characteristics**: MoS increases FLOPs by approximately $15$–$20\%$ compared to trilinear upsampling. It remains computationally cheaper than heavy spatial-attention modules and provides significant PSNR improvements [2606.26716, Table 9].

## 6. Empirical Effect and Ablation Analysis

Ablation studies in [2606.26716, Sec. 4.4, Tables 3–4, Fig. 5] demonstrate the unique contribution of MoS:
- Replacing MoS with trilinear upsampling in the null-space estimator reduces $\times 2$ PSNR by $0.63$ dB.
- The combination of MoS and measurement-consistent projection (MCP) obtains the highest fidelity, with an additional $0.23$ dB over MoS without MCP.
- Single-order splines (any fixed $p$) have comparable in-scale performance, but the multi-order MoS generalizes better to scales outside the training distribution.
- Expert-weight visualizations show that MoS order selection is semantically meaningful, aligning with anatomical structure boundaries and tissue homogeneity.

## 7. Comparative Positioning and Functionality

MoS provides a principled, analytic alternative to black-box MLP or fixed-order spline upsamplers within constrained super-resolution frameworks. Its per-location adaption via a lightweight routing mechanism delivers explicit $C^{p-1}$ continuity control consistent with anatomical topology in medical imaging. No auxiliary losses or supervision are introduced for MoS itself; end-to-end learning in the context of DP-NSL is sufficient for optimal spatial upsampling.

A plausible implication is that MoS can be generalized as a modular upsampling unit where explicit, content-adaptive continuity is desired, and black-box interpolation mechanisms are ill-suited for measurement-consistent inverse problems. It also suggests that further exploration of spline order ensembles and dynamic weighting strategies could extend to other geometric or scientific domains [2606.26716].

Source: https://www.emergentmind.com/topics/mixture-of-splines-mos-module