SFD-Mamba2Net: Dual-Use Architectures

• SFD-Mamba2Net is a label for two distinct Mamba-family architectures: one for multimodal urban damage assessment and another for coronary artery analysis.
• In the blast-damage variant, the network fuses pre‐ and post-event imagery with simulated blast-loading data to achieve pixel-wise damage mapping for rapid disaster response.
• In the ICA variant, an asymmetric encoder-decoder utilizing CASE, AA-DS Mamba2, and PHFP modules delivers enhanced coronary segmentation and stenosis detection.

Searching arXiv for the cited SFD-Mamba2Net papers to ground the article in current records. SFD-Mamba2Net is a name used on arXiv for two distinct Mamba-family architectures introduced in different application domains. In one usage, it denotes a Mamba-based multimodal network for rapid blast-induced structural damage assessment (SDA) from optical remote sensing imagery, with the 2020 Beirut explosion as the motivating case (Ma et al., 13 Apr 2026). In the other, it denotes an end-to-end framework for coronary artery segmentation and stenosis detection in invasive coronary angiography (ICA) that combines structural priors, dual-stream state-space modeling, and frequency-domain reconstruction (Mu et al., 10 Sep 2025). The shared name can suggest a single model lineage, but the two systems differ in task definition, modalities, module design, loss functions, datasets, and evaluation protocols.

1. Name, scope, and disambiguation

The label SFD-Mamba2Net is not tied to one canonical architecture. The 2026 blast-damage paper presents a multi-task multimodal network built around Mamba / visual state space components, inspired by VMamba and ChangeMamba, for pixel-wise structural damage assessment over an urban area after an explosion (Ma et al., 13 Apr 2026). The 2025 medical-imaging paper presents an asymmetric encoder-decoder for ICA that targets coronary artery segmentation and stenosis detection, with modules named CASE, AA-DS Mamba2, and PHFP (Mu et al., 10 Sep 2025).

This naming overlap matters because the acronym “SFD” is domain-specific in each paper. In the blast-damage setting, the architecture is defined primarily by multimodal fusion of pre-event imagery, post-event imagery, and blast-loading maps. In the ICA setting, the architecture is defined by structure-guided enhancement, frequency-aware decoding, and a dual-stream Mamba2 bottleneck. A plausible implication is that citations or repository references to “SFD-Mamba2Net” require domain context to avoid conflating urban damage mapping with coronary image analysis.

2. Structural damage assessment formulation

In the blast-damage paper, the target task is pixel-wise structural damage assessment after a major explosion. The model must infer building localization / segmentation from pre-event imagery, damage classification from pre- and post-event imagery, and may exploit blast-loading information that varies spatially with distance and exposure (Ma et al., 13 Apr 2026).

The problem formulation is motivated by the operational need for fast, large-area, building-level damage maps after an explosion. Classical field inspection is described as too slow and dangerous, while remote sensing provides a scalable alternative. The paper identifies three limitations in existing SDA systems: dependence on labeled data, poor transfer to small local events, and lack of physical blast context. To address these, it adopts a two-stage pipeline consisting of pre-training on xBD and fine-tuning on a small Beirut target dataset, while injecting simulated blast-load maps as an additional modality.

The core architecture contains four components: an image encoder, a building-segmentation decoder, a blast-loading encoder, and a damage-assessment decoder using a Residual Attention–based Spatiotemporal State Space (RA-STSS) fusion block. The optical inputs are the pre-event image $\mathbf{I}_{pre}$ and post-event image $\mathbf{I}_{post}$, and the third modality is the blast-loading map $\mathbf{B}$.

At the encoder front end, each optical image is patchified and embedded as

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$

followed by hierarchical feature extraction with VSS blocks and downsampling:

$$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$

The encoder is applied independently to pre- and post-event imagery, producing multi-scale feature sets $\{\mathbf{F}_l^{pre}\}_{l=1}^4$ and $\{\mathbf{F}_l^{post}\}_{l=1}^4$. The paper attributes to Mamba/VSS efficient long-range spatial modeling with linear-time sequence processing while remaining lightweight enough for rapid deployment.

The first decoder reconstructs a binary building mask from pre-event imagery alone:

$$\mathbf{M} = \mathrm{Decoder}(\{\mathbf{F}_l^{pre}\}_{l=1}^4) \in \mathbb{R}^{H \times W \times 3}.$$

This auxiliary task supplies mask supervision so that the network learns building-aware features before damage classification. The blast modality is encoded separately as a spatial map $\mathbf{B}\in\mathbb{R}^{H\times W\times 3}$, interpolated to each pyramid scale and projected with a $1\times1$ convolution, yielding $\mathbf{I}_{post}$0.

The damage decoder first fuses pre- and post-event features:

$\mathbf{I}_{post}$1

then performs a residual update modulated by the blast feature at the same level:

$\mathbf{I}_{post}$2

with $\mathbf{I}_{post}$3. The paper identifies this as the key multimodal fusion mechanism: the blast map acts as a residual attention gate, amplifying or suppressing features according to expected blast exposure. The final damage map is obtained by

$\mathbf{I}_{post}$4

3. Blast-loading representation, training, and empirical results

A defining contribution of the blast-damage SFD-Mamba2Net is the explicit use of multiscale blast-loading information generated by physics-based simulation rather than inferred from imagery alone (Ma et al., 13 Apr 2026). The paper states that the blast load is simulated using Viper::Blast, a GPU-based CFD solver validated for urban blast events. The Beirut explosion is approximated as a 0.50 kt TNT charge detonated at a height of 10 m above ground, with a cylindrical charge shape and length-to-diameter aspect ratio 1.

Training is formulated as multi-task optimization with two pixel-wise cross-entropy terms, $\mathbf{I}_{post}$5 for building segmentation and $\mathbf{I}_{post}$6 for damage classification:

$\mathbf{I}_{post}$7

The target dataset, Blast-7, is constructed by combining BRIGHT with blast-loading data from the 2020 Beirut explosion and replacing BRIGHT’s post-event images with high-resolution optical imagery from the Maxar Open Data Program. Blast-7 contains 50 images of 512 × 512 pixels, covers about 7 km², and is split into training / validation / test in a 3:1:1 ratio. The pretraining dataset, xBD, contains 11,034 image pairs at 1024 × 1024 resolution, 850,736 buildings, 19 disaster events, and 45,361.79 km² of coverage. The implementation uses PyTorch, AdamW, a learning rate of $\mathbf{I}_{post}$8 for pretraining, and $\mathbf{I}_{post}$9 for fine-tuning.

Evaluation follows the ChangeMamba/xBD protocol with $\mathbf{B}$0, $\mathbf{B}$1, $\mathbf{B}$2, and class-wise $\mathbf{B}$3 for Intact, Damaged, and Destroyed. The reported best results are:

Metric	Value
$\mathbf{B}$4	88.98%
$\mathbf{B}$5	88.30%
$\mathbf{B}$6	88.50%
Intact	93.54%
Damaged	77.96%
Destroyed	95.64%

The comparison set includes CNN-based baselines (DeepLabV3+, U-Net, SiamAttnUNet, SiamCRNN), a Transformer-based baseline (DamFormer), and a Mamba-based baseline (Mamba-BDA-Small). The paper reports that, compared with UNet, $\mathbf{B}$7 improves by 27.55%; compared with SiamCRNN, by 12.59%; over DamFormer, overall classification improves by 7.3% and damaged-class detection by 14.8%; and compared with Mamba-BDA-Small, the method improves overall performance and especially the difficult “damaged” class by about 19.2%. Fine-tuning requires only about 13 minutes.

The ablation study isolates pretraining, fine-tuning, distance information, and blast-loading information. Pretrain only yields $\mathbf{B}$8 and $\mathbf{B}$9, indicating that global pretraining alone is insufficient for the Beirut scenario. Fine-tuning (FT) gives $$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$0, FT with distance information (w/D) gives 86.20%, and FT with blast loading (w/B) reaches 88.50%. This directly supports the paper’s claim that physics-informed blast features materially improve damage classification, particularly for the intact and damaged classes.

4. Coronary artery segmentation architecture

In the ICA paper, SFD-Mamba2Net is formulated as an end-to-end framework for coronary artery segmentation and stenosis detection in invasive coronary angiography (Mu et al., 10 Sep 2025). The motivation is clinical: coronary artery disease is a leading cause of mortality, ICA is the gold standard for diagnosing coronary anatomy and stenosis severity, and manual interpretation is slow, subjective, and variable. Automated analysis is difficult because ICA images exhibit low contrast, high noise, projection overlap, motion artifacts, and extremely fine, tortuous vessel structures, especially micro-vessels smaller than two pixels, bifurcations, and distal branches.

The architecture uses an asymmetric encoder-decoder. The encoder has five stages, the first four built from conventional convolutional blocks, and the shallowest stage augmented by the CASE module. Each encoder block consists of two CBL units, where a CBL is 3×3 convolution + batch normalization + LeakyReLU, with dropout after BN. The stage feature is defined as

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$1

and downsampling uses

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$2

At the bottleneck, the model inserts AA-DS Mamba2, and the decoder mirrors the encoder with four upsampling stages. Each decoder stage concatenates the skip feature from the encoder with the upsampled output from the previous decoder stage,

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$3

where upsampling is

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$4

At the start of each decoder stage, the PHFP module refines the concatenated feature by progressive wavelet decomposition and reconstruction.

The CASE module, or Curvature-Aware Structural Enhancement, provides an explicit geometric prior for vessels. It uses a multi-scale Hessian-based vesselness filter inspired by Frangi-style vessel enhancement, with Gaussian smoothing at $$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$5, followed by Hessian computation

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$6

and eigenvalues

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$7

The vesselness response is

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$8

with

$$\mathbf{F}_0 = \mathrm{PatchPartition}(\mathbf{I}) \in \mathbb{R}^{\tfrac{H}{4}\times\tfrac{W}{4}\times C_1},$$9

The paper emphasizes that CASE has no learnable parameters, takes the maximum vesselness response across scales, concatenates structural maps with early learned features, and applies a lightweight channel attention mechanism.

The PHFP module, or Progressive High-Frequency Perception, is designed to recover fine boundaries and small branches lost during downsampling. Using the Haar wavelet, it decomposes each decoder feature into $$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$0, $$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$1, $$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$2, and $$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$3 sub-bands:

$$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$4

High-frequency components are refined by a $$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$5 depthwise convolution $$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$6, and reconstruction proceeds by inverse wavelet transform:

$$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$7

The final PHFP output is

$$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$8

The paper’s interpretation is that low-frequency components preserve global vessel topology, while high-frequency components carry edges and fine branches.

5. AA-DS Mamba2, optimization, and segmentation performance

The ICA model’s global-context mechanism is the Axial-Alternating Dual-Stream Mamba2 (AA-DS Mamba2) bottleneck, which adapts Mamba2 to 2-D vascular anatomy (Mu et al., 10 Sep 2025). For input

$$\mathbf{F}_1 = \mathrm{VSS}(\mathbf{F}_0), \qquad \mathbf{F}_l = \mathrm{VSS}(\mathrm{Down}(\mathbf{F}_{l-1})) \quad l=2,3,4.$$9

one path reshapes along width,

$\{\mathbf{F}_l^{pre}\}_{l=1}^4$0

applies DS Mamba2, reshapes back, then unfolds along height and applies DS Mamba2 again. The second path reverses that order. Their outputs are concatenated,

$\{\mathbf{F}_l^{pre}\}_{l=1}^4$1

and fused by

$\{\mathbf{F}_l^{pre}\}_{l=1}^4$2

The core DS Mamba2 block is bidirectional and processes the original sequence and its flipped version in parallel. The Mamba2 pipeline is described as input projection, activation and splitting into $\{\mathbf{F}_l^{pre}\}_{l=1}^4$3, $\{\mathbf{F}_l^{pre}\}_{l=1}^4$4, $\{\mathbf{F}_l^{pre}\}_{l=1}^4$5, and $\{\mathbf{F}_l^{pre}\}_{l=1}^4$6, SSD state propagation, fusion and normalization, and output projection. The structured state-space component is based on

$\{\mathbf{F}_l^{pre}\}_{l=1}^4$7

with

$\{\mathbf{F}_l^{pre}\}_{l=1}^4$8

followed by chunking into intra-chunk and inter-chunk terms, yielding $\{\mathbf{F}_l^{pre}\}_{l=1}^4$9 and $\{\mathbf{F}_l^{post}\}_{l=1}^4$0 before reshape and output.

Training uses mean squared error (MSE),

$\{\mathbf{F}_l^{post}\}_{l=1}^4$1

which the paper justifies as providing smooth gradients and helping suppress noise while preserving topology. Optimization uses Adam with initial learning rate $\{\mathbf{F}_l^{post}\}_{l=1}^4$2, $\{\mathbf{F}_l^{post}\}_{l=1}^4$3, $\{\mathbf{F}_l^{post}\}_{l=1}^4$4, and weight decay $\{\mathbf{F}_l^{post}\}_{l=1}^4$5. A StepLR schedule keeps the learning rate at $\{\mathbf{F}_l^{post}\}_{l=1}^4$6 for the first 10 epochs, halves it every 10 epochs until epoch 30, and then fixes it at $\{\mathbf{F}_l^{post}\}_{l=1}^4$7. Training is performed in PyTorch on two RTX 3090 GPUs.

The dataset comprises images from 99 patients at Jiangsu Province People’s Hospital between Feb. 26 and July 18, 2019, acquired with a Siemens AXIOM-Artis system at 15 fps, with resolution $\{\mathbf{F}_l^{post}\}_{l=1}^4$8 and spacing from 0.258 to 0.390 mm. It contains 187 left coronary artery (LCA) images and 127 right coronary artery (RCA) images, annotated frame-by-frame by experienced interventional cardiologists. Segmentation is evaluated by Dice, Acc, F1, IoU, Sens, Spec, HD95, and ASSD.

The comparison set includes U-Net, U-Net++, Swin-Unet, DSCNet, VM-UNet, Perspective-UNet, and ConDSeg. The reported results are:

Metric	Value
Dice	88.10%
F1	89.22%
IoU	80.81%
Accuracy	98.80%
Specificity	99.50%
HD95	7.57 pixels
ASSD	1.25 pixels

Relative to VM-UNet, Dice improves from 87.30% to 88.10% and IoU from 80.06% to 80.81%, while VM-UNet remained better on HD95 (5.64 vs 7.57 pixels). Relative to ConDSeg, Dice improves from 87.02% to 88.10%, F1 from 88.61% to 89.22%, IoU from 79.80% to 80.81%, and Specificity from 99.48% to 99.50%. The ablation study reports that adding the structural module alone gives modest improvement, AA-DS Mamba2 improves Dice and reduces ASSD more substantially, PHFP improves boundary localization and especially lowers HD95, and the full model yields the best combined results.

6. Stenosis detection, qualitative interpretation, and limitations

In the ICA paper, stenosis detection is not framed as a separate end-to-end classification network; instead, it is a downstream pipeline operating on the segmentation output (Mu et al., 10 Sep 2025). The vessel mask is first skeletonized to obtain the centerline, then the Euclidean distance transform is used to estimate local diameters:

$\{\mathbf{F}_l^{post}\}_{l=1}^4$9

The skeleton is decomposed into segments based on endpoint/bifurcation graph structure, and for each segment a stenosis ratio is computed as

$$\mathbf{M} = \mathrm{Decoder}(\{\mathbf{F}_l^{pre}\}_{l=1}^4) \in \mathbb{R}^{H \times W \times 3}.$$0

A segment is marked as stenotic if $$\mathbf{M} = \mathrm{Decoder}(\{\mathbf{F}_l^{pre}\}_{l=1}^4) \in \mathbb{R}^{H \times W \times 3}.$$1 (10%). Additional constraints exclude segments with maximum diameter below 1.8 mm and segments shorter than 20 pixels; if no suitable inflection point is found, a global maximum diameter is used as reference. Severity is graded as minimal: 1%–24%, mild: 25%–49%, moderate: 50%–69%, and severe: 70%–100%.

The evaluation metrics are TPR, PPV, ARMSE, and RRMSE:

$$\mathbf{M} = \mathrm{Decoder}(\{\mathbf{F}_l^{pre}\}_{l=1}^4) \in \mathbb{R}^{H \times W \times 3}.$$2

$$\mathbf{M} = \mathrm{Decoder}(\{\mathbf{F}_l^{pre}\}_{l=1}^4) \in \mathbb{R}^{H \times W \times 3}.$$3

Across 132 test images, the paper reports TPR = 0.60, PPV = 0.64, overall detection ratio = 0.4474, ARMSE = 0.14, and RRMSE = 0.30. For severe stenosis, it reports TPR = 0.34 and PPV = 0.55.

Both papers include qualitative evidence that is tightly coupled to their module design. In the blast-damage study, the salient qualitative claim is that adding blast-loading information provides physically grounded context beyond optical change alone, and the ablation results support that interpretation. In the ICA study, Grad-CAM visualizations are used to show stronger and more continuous activation on vessel structures for CASE, broader and more continuous attention over long vessel paths and bifurcations for AA-DS Mamba2, and sharper activations along branches and trunks for PHFP. This suggests that the two SFD-Mamba2Net variants, though unrelated in application, each use auxiliary structure to compensate for weaknesses of image-only feature extraction.

The limitations are likewise domain-specific. The blast-damage paper notes room for improvement through fine-tuning large vision/LLMs and richer multimodal SDA pipelines, while implicitly depending on a sufficiently good pretrained foundation model, simulated blast-load estimation, and some local fine-tuning samples. The ICA paper states explicitly that the dataset is relatively limited in scale and diversity, that the stenosis pipeline is based on static single-view ICA rather than multi-view or dynamic sequences, and that the 2-D projection of a 3-D coronary tree can reduce quantitative accuracy. Future work there is directed toward dataset expansion, multi-view and time-resolved ICA, and 3-D reconstruction approaches.

The combined record therefore presents SFD-Mamba2Net not as a single architecture but as a recurring name for two technically distinct systems: one for physics-informed rapid structural damage assessment after urban explosions (Ma et al., 13 Apr 2026), and one for structure-guided, frequency-enhanced coronary analysis in ICA (Mu et al., 10 Sep 2025).