Slide
Centre Interdisciplinaire
de Recherche et d’Innovation
en Cybersécurité et Société
de Recherche et d’Innovation
en Cybersécurité et Société
1.
Allaoui, M. L.; Allili, M. S.
LG-TargetSeg: Deep semantic fusion via language-conditioned selective scanning for targeted medical image segmentation Journal Article
In: Information Fusion, vol. 139, 2027, ISSN: 15662535 (ISSN).
Abstract | Links | BibTeX | Tags: Computerized tomography, decoding, Deep vision-language fusion, Encoder-decoder, Gaussians, Medical image processing, Medical image segmentation, Modeling languages, Natural language processing systems, Reinforced gaussian dice loss, Selective scanning, Semantic fusion, Semantic segmentation, Semantics, Shallow levels, Signal encoding, State space methods, State-space modeling, State-space models, Training time
@article{allaouiLGTargetSegDeepSemantic2027,
title = {LG-TargetSeg: Deep semantic fusion via language-conditioned selective scanning for targeted medical image segmentation},
author = {M. L. Allaoui and M. S. Allili},
url = {https://www.scopus.com/pages/publications/105051242411?origin=resultslist},
doi = {10.1016/j.inffus.2026.104797},
issn = {15662535 (ISSN)},
year = {2027},
date = {2027-01-01},
journal = {Information Fusion},
volume = {139},
publisher = {Elsevier B.V.},
abstract = {Clinical language defines the semantic intent of targeted medical image segmentation, yet most vision–language methods incorporate it only at a shallow level or through training-time conditioning, leaving spatial reasoning largely language-blind. We propose LG-TargetSeg (Language-Guided Targeted Segmentation), a cross-modal framework that embeds clinical language directly into the selective scan of a Mamba-based encoder-decoder. Natural-language prompts, encoded by a frozen MedSigLIP foundation model, shape selective scanning throughout the encoder–decoder and additionally adapt the bottleneck state transition, alongside gated modulation and cross-attention, enabling semantic guidance to continuously shape feature propagation toward the target anatomy. A Text-Driven Selective Scan 2D (TD-SS2D) module achieves prompt-aware multi-scale fusion with linear complexity in the number of spatial tokens. To improve robustness to boundary ambiguity and class imbalance, we introduce the Reinforced Gaussian Dice Loss (RGDL) with learnable weighting parameters. Evaluated under both task-specific and universal generalization settings, LG-TargetSeg achieves state-of-the-art performance. Moreover, its prompt-driven selectivity remains reliable under genuine multi-target competition: on abdominal CT images containing up to 16 annotated organs, changing only the text prompt on a frozen model produces organ masks with less than 1% pairwise overlap. Additional materials and code are available at: https://github.com/MOHAMEDLamine77/LG-TargetSeg . © 2026 Elsevier B.V.},
keywords = {Computerized tomography, decoding, Deep vision-language fusion, Encoder-decoder, Gaussians, Medical image processing, Medical image segmentation, Modeling languages, Natural language processing systems, Reinforced gaussian dice loss, Selective scanning, Semantic fusion, Semantic segmentation, Semantics, Shallow levels, Signal encoding, State space methods, State-space modeling, State-space models, Training time},
pubstate = {published},
tppubtype = {article}
}
Clinical language defines the semantic intent of targeted medical image segmentation, yet most vision–language methods incorporate it only at a shallow level or through training-time conditioning, leaving spatial reasoning largely language-blind. We propose LG-TargetSeg (Language-Guided Targeted Segmentation), a cross-modal framework that embeds clinical language directly into the selective scan of a Mamba-based encoder-decoder. Natural-language prompts, encoded by a frozen MedSigLIP foundation model, shape selective scanning throughout the encoder–decoder and additionally adapt the bottleneck state transition, alongside gated modulation and cross-attention, enabling semantic guidance to continuously shape feature propagation toward the target anatomy. A Text-Driven Selective Scan 2D (TD-SS2D) module achieves prompt-aware multi-scale fusion with linear complexity in the number of spatial tokens. To improve robustness to boundary ambiguity and class imbalance, we introduce the Reinforced Gaussian Dice Loss (RGDL) with learnable weighting parameters. Evaluated under both task-specific and universal generalization settings, LG-TargetSeg achieves state-of-the-art performance. Moreover, its prompt-driven selectivity remains reliable under genuine multi-target competition: on abdominal CT images containing up to 16 annotated organs, changing only the text prompt on a frozen model produces organ masks with less than 1% pairwise overlap. Additional materials and code are available at: https://github.com/MOHAMEDLamine77/LG-TargetSeg . © 2026 Elsevier B.V.
2.
Allaoui, M. L.; Allili, M. S.; Belaid, A.
HA-U3Net: A modality-agnostic framework for 3D medical image segmentation using nested V-Net structure and hybrid attention Journal Article
In: Knowledge-Based Systems, vol. 327, 2025, ISSN: 09507051 (ISSN).
Abstract | Links | BibTeX | Tags: 3D medical image, 3D medical image segmentation, Diagnosis, Diagnosis planning, Disease diagnosis, Disease treatment, Generalization capability, Image segmentation, Magnetic resonance imaging, Medical image processing, Medical image segmentation, Nested volume-structure, Net structures, Self hybrid attention, Structures (built objects)
@article{allaoui_ha-u3net_2025,
title = {HA-U3Net: A modality-agnostic framework for 3D medical image segmentation using nested V-Net structure and hybrid attention},
author = {M. L. Allaoui and M. S. Allili and A. Belaid},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-105011370963&doi=10.1016%2Fj.knosys.2025.114127&partnerID=40&md5=d98a109f015445adb3001bb4017bf953},
doi = {10.1016/j.knosys.2025.114127},
issn = {09507051 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Knowledge-Based Systems},
volume = {327},
abstract = {3D medical image segmentation is essential for disease diagnosis and treatment planning across a wide range of imaging modalities (e.g., MRI, CT, ultrasound, and PET). However, modality-specific challenges, such as noise, artifacts, low contrast, and anatomical variability, along with the presence of small lesions and fuzzy boundaries, hinder the generalization capability of existing segmentation models. In this work, we present HA-U3Net, a novel 3D U-Net-based model designed to address these limitations through a stepwise approach. First, we introduce a deeply nested U3-shaped structure built upon 3D V-Net modules, enabling multi-scale hierarchical representation learning. Second, we integrate a hybrid attention mechanism combining spatial and channel-wise attention to enhance salient features extraction and the delineation of small or poorly defined structures. Third, we demonstrate the cross-modality generalization capabilities of HA-U3Net through extensive evaluations on several datasets, where our model consistently outperforms baseline methods. Finally, we propose a lightweight variant, U3Mamba, reducing computational complexity while maintaining high performance. © 2025 Elsevier B.V.},
keywords = {3D medical image, 3D medical image segmentation, Diagnosis, Diagnosis planning, Disease diagnosis, Disease treatment, Generalization capability, Image segmentation, Magnetic resonance imaging, Medical image processing, Medical image segmentation, Nested volume-structure, Net structures, Self hybrid attention, Structures (built objects)},
pubstate = {published},
tppubtype = {article}
}
3D medical image segmentation is essential for disease diagnosis and treatment planning across a wide range of imaging modalities (e.g., MRI, CT, ultrasound, and PET). However, modality-specific challenges, such as noise, artifacts, low contrast, and anatomical variability, along with the presence of small lesions and fuzzy boundaries, hinder the generalization capability of existing segmentation models. In this work, we present HA-U3Net, a novel 3D U-Net-based model designed to address these limitations through a stepwise approach. First, we introduce a deeply nested U3-shaped structure built upon 3D V-Net modules, enabling multi-scale hierarchical representation learning. Second, we integrate a hybrid attention mechanism combining spatial and channel-wise attention to enhance salient features extraction and the delineation of small or poorly defined structures. Third, we demonstrate the cross-modality generalization capabilities of HA-U3Net through extensive evaluations on several datasets, where our model consistently outperforms baseline methods. Finally, we propose a lightweight variant, U3Mamba, reducing computational complexity while maintaining high performance. © 2025 Elsevier B.V.
3.
Allaoui, M. L.; Allili, M. S.; Belaid, A.
HA-U3Net: A modality-agnostic framework for 3D medical image segmentation using nested V-Net structure and hybrid attention Journal Article
In: Knowledge-Based Systems, vol. 327, 2025, ISSN: 09507051 (ISSN).
Abstract | Links | BibTeX | Tags: 3D medical image, 3D medical image segmentation, Diagnosis, Diagnosis planning, Disease diagnosis, Disease treatment, Generalization capability, Image segmentation, Magnetic resonance imaging, Medical image processing, Medical image segmentation, Nested volume-structure, Net structures, Self hybrid attention, Structures (built objects)
@article{allaouiHAU3NetModalityagnosticFramework2025,
title = {HA-U3Net: A modality-agnostic framework for 3D medical image segmentation using nested V-Net structure and hybrid attention},
author = {M. L. Allaoui and M. S. Allili and A. Belaid},
url = {https://www.scopus.com/pages/publications/105011370963?origin=resultslist},
doi = {10.1016/j.knosys.2025.114127},
issn = {09507051 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Knowledge-Based Systems},
volume = {327},
publisher = {Elsevier B.V.},
abstract = {3D medical image segmentation is essential for disease diagnosis and treatment planning across a wide range of imaging modalities (e.g., MRI, CT, ultrasound, and PET). However, modality-specific challenges, such as noise, artifacts, low contrast, and anatomical variability, along with the presence of small lesions and fuzzy boundaries, hinder the generalization capability of existing segmentation models. In this work, we present HA-U3Net, a novel 3D U-Net-based model designed to address these limitations through a stepwise approach. First, we introduce a deeply nested U3-shaped structure built upon 3D V-Net modules, enabling multi-scale hierarchical representation learning. Second, we integrate a hybrid attention mechanism combining spatial and channel-wise attention to enhance salient features extraction and the delineation of small or poorly defined structures. Third, we demonstrate the cross-modality generalization capabilities of HA-U3Net through extensive evaluations on several datasets, where our model consistently outperforms baseline methods. Finally, we propose a lightweight variant, U3Mamba, reducing computational complexity while maintaining high performance. © 2025 Elsevier B.V.},
keywords = {3D medical image, 3D medical image segmentation, Diagnosis, Diagnosis planning, Disease diagnosis, Disease treatment, Generalization capability, Image segmentation, Magnetic resonance imaging, Medical image processing, Medical image segmentation, Nested volume-structure, Net structures, Self hybrid attention, Structures (built objects)},
pubstate = {published},
tppubtype = {article}
}
3D medical image segmentation is essential for disease diagnosis and treatment planning across a wide range of imaging modalities (e.g., MRI, CT, ultrasound, and PET). However, modality-specific challenges, such as noise, artifacts, low contrast, and anatomical variability, along with the presence of small lesions and fuzzy boundaries, hinder the generalization capability of existing segmentation models. In this work, we present HA-U3Net, a novel 3D U-Net-based model designed to address these limitations through a stepwise approach. First, we introduce a deeply nested U3-shaped structure built upon 3D V-Net modules, enabling multi-scale hierarchical representation learning. Second, we integrate a hybrid attention mechanism combining spatial and channel-wise attention to enhance salient features extraction and the delineation of small or poorly defined structures. Third, we demonstrate the cross-modality generalization capabilities of HA-U3Net through extensive evaluations on several datasets, where our model consistently outperforms baseline methods. Finally, we propose a lightweight variant, U3Mamba, reducing computational complexity while maintaining high performance. © 2025 Elsevier B.V.
4.
Allaoui, M. L.; Allili, M. S.; Belaid, A.
HA-U3Net: A modality-agnostic framework for 3D medical image segmentation using nested V-Net structure and hybrid attention Journal Article
In: Knowledge-Based Systems, vol. 327, 2025, ISSN: 09507051 (ISSN).
Abstract | Links | BibTeX | Tags: 3D medical image, 3D medical image segmentation, Diagnosis, Diagnosis planning, Disease diagnosis, Disease treatment, Generalization capability, Image segmentation, Magnetic resonance imaging, Medical image processing, Medical image segmentation, Nested volume-structure, Net structures, Self hybrid attention, Structures (built objects)
@article{allaouiHAU3NetModalityagnosticFramework2025a,
title = {HA-U3Net: A modality-agnostic framework for 3D medical image segmentation using nested V-Net structure and hybrid attention},
author = {M. L. Allaoui and M. S. Allili and A. Belaid},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-105011370963&doi=10.1016%2Fj.knosys.2025.114127&partnerID=40&md5=d98a109f015445adb3001bb4017bf953},
doi = {10.1016/j.knosys.2025.114127},
issn = {09507051 (ISSN)},
year = {2025},
date = {2025-01-01},
journal = {Knowledge-Based Systems},
volume = {327},
abstract = {3D medical image segmentation is essential for disease diagnosis and treatment planning across a wide range of imaging modalities (e.g., MRI, CT, ultrasound, and PET). However, modality-specific challenges, such as noise, artifacts, low contrast, and anatomical variability, along with the presence of small lesions and fuzzy boundaries, hinder the generalization capability of existing segmentation models. In this work, we present HA-U3Net, a novel 3D U-Net-based model designed to address these limitations through a stepwise approach. First, we introduce a deeply nested U3-shaped structure built upon 3D V-Net modules, enabling multi-scale hierarchical representation learning. Second, we integrate a hybrid attention mechanism combining spatial and channel-wise attention to enhance salient features extraction and the delineation of small or poorly defined structures. Third, we demonstrate the cross-modality generalization capabilities of HA-U3Net through extensive evaluations on several datasets, where our model consistently outperforms baseline methods. Finally, we propose a lightweight variant, U3Mamba, reducing computational complexity while maintaining high performance. © 2025 Elsevier B.V.},
keywords = {3D medical image, 3D medical image segmentation, Diagnosis, Diagnosis planning, Disease diagnosis, Disease treatment, Generalization capability, Image segmentation, Magnetic resonance imaging, Medical image processing, Medical image segmentation, Nested volume-structure, Net structures, Self hybrid attention, Structures (built objects)},
pubstate = {published},
tppubtype = {article}
}
3D medical image segmentation is essential for disease diagnosis and treatment planning across a wide range of imaging modalities (e.g., MRI, CT, ultrasound, and PET). However, modality-specific challenges, such as noise, artifacts, low contrast, and anatomical variability, along with the presence of small lesions and fuzzy boundaries, hinder the generalization capability of existing segmentation models. In this work, we present HA-U3Net, a novel 3D U-Net-based model designed to address these limitations through a stepwise approach. First, we introduce a deeply nested U3-shaped structure built upon 3D V-Net modules, enabling multi-scale hierarchical representation learning. Second, we integrate a hybrid attention mechanism combining spatial and channel-wise attention to enhance salient features extraction and the delineation of small or poorly defined structures. Third, we demonstrate the cross-modality generalization capabilities of HA-U3Net through extensive evaluations on several datasets, where our model consistently outperforms baseline methods. Finally, we propose a lightweight variant, U3Mamba, reducing computational complexity while maintaining high performance. © 2025 Elsevier B.V.