Mohammed El Amine Bechar
Mohammed El Amine Bechar has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Marwa El Bouz
LightScatterLearning (LSL) group, L@bISEN, ISEN Ouest, Brest, France
7 shared publications - Nesma Settouti
LabISEN, LSL, ISEN Ouest, Caen, France
6 shared publications - Marie-Bérengère Troadec
Univ Brest, Inserm, EFS, UMR 1078, GGB Brest France
5 shared publications - Jean-Marie Guyader
LABISEN, ISEN Yncréa Ouest, Brest, France
4 shared publications - Nathalie Douet-Guilbert
Univ Brest, Inserm, EFS, UMR 1078, GGB Brest France
4 shared publications - Jean-Marie Guyader
LabISEN, VISION-AD, ISEN Ouest, Brest, France
3 shared publications
Latest Publications
- Open Access
- 3 October 2026
- Published by Elsevier in SSRN Electronic Journal
Phasor-nnU-Net: A Multimodal Architecture for Segmenting Overlapping Chromosomes in Hyperspectral Images
Accurate segmentation of overlapped chromosomes in hyperspectral imageryremains challenging due to complex spatial interactions and spectral redun-dancy. In this work, we propose a multimodal framework for segmentationthat integrates phasor-based spectral representations with a 3D nnU-Netarchitecture. The proposed framework exploits phasor features as a com-plementary modality processed through a dedicated 2D encoder and fusedwith volumetric hyperspectral representations within a unified 3D decodingpathway. To bridge the dimensional gap between 2D and 3D feature spaces,decoder-level fusion strategies are investigated to enable effective cross-modalinteraction. Experimental results demonstrate consistent improvements overRGB-based, hyperspectral imaging (HSI)-based, and hybrid multimodal ap-proaches. Compared with the Vanilla nnU-Net 3D, the proposed frameworkimproves the IoU and Dice scores by approximately 4 and 2 percentage points,respectively. Compared with the RGB-based baseline methods, it achievesIoU gains of up to 11 percentage points. Statistical analysis confirms thesignificance of the observed improvements (p ≪ 0.001). The results fur-ther suggest that phasor representations provide complementary informationto volumetric hyperspectral features, leading to more accurate and robustchromosome segmentation. These findings highlight the potential of phasor-guided multimodal learning for hyperspectral biomedical image analysis.
- 3 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPhasor-nnU-Net: A Multimodal Architecture for Segmenting Overlapping Chromosomes in Hyperspectral Images
Accurate segmentation of overlapped chromosomes in hyperspectral imageryremains challenging due to complex spatial interactions and spectral redun-dancy. In this work, we propose a multimodal framework for segmentationthat integrates phasor-based spectral representations with a 3D nnU-Netarchitecture. The proposed framework exploits phasor features as a com-plementary modality processed through a dedicated 2D encoder and fusedwith volumetric hyperspectral representations within a unified 3D decodingpathway. To bridge the dimensional gap between 2D and 3D feature spaces,decoder-level fusion strategies are investigated to enable effective cross-modalinteraction. Experimental results demonstrate consistent improvements overRGB-based, hyperspectral imaging (HSI)-based, and hybrid multimodal ap-proaches. Compared with the Vanilla nnU-Net 3D, the proposed frameworkimproves the IoU and Dice scores by approximately 4 and 2 percentage points,respectively. Compared with the RGB-based baseline methods, it achievesIoU gains of up to 11 percentage points. Statistical analysis confirms thesignificance of the observed improvements (p ≪ 0.001). The results fur-ther suggest that phasor representations provide complementary informationto volumetric hyperspectral features, leading to more accurate and robustchromosome segmentation. These findings highlight the potential of phasor-guided multimodal learning for hyperspectral biomedical image analysis.
- Open Access
- 15 July 2026
- Published by Springer Nature in Scientific Data
- Vol. 13 (1)
HSI-AgriFoodAnomaly, a hyperspectral dataset for foreign object detection in agri-food inspection
Ensuring food safety in agri-food production requires reliable inspection data for detecting foreign objects and other contamination risks. Hyperspectral imaging (HSI) is a relevant non-destructive modality for this purpose because it combines spatial information with detailed spectral responses. This Data Descriptor presents HSI-AgriFoodAnomaly, an open hyperspectral image dataset acquired under conveyor-based industrial-like conditions for foreign object annotation in an oat, white-chocolate and dark-chocolate mixture. The dataset contains 147 calibrated hyperspectral cubes, associated red-green-blue (RGB) renderings, pixel-level binary masks, and polygon annotations. Each hyperspectral cube was acquired in the visible to near-infrared range, with 300 contiguous spectral bands covering approximately 381–1016 nm. The annotated foreign objects (FOs) include textile and fibre-based materials, plastics, paper-based materials, metals, wood and plant residues, minerals, glass, mixed-object scenes, and anomaly-free scenes. The dataset is organised into training, validation and test subsets at the cube level, in order to prevent data leakage between splits. HSI-AgriFoodAnomaly can be reused for hyperspectral image analysis, foreign object localisation, binary classification, object detection, semantic segmentation, and benchmarking of data processing pipelines. The dataset and the accompanying code are publicly available.
- 15 July 2026
- Published by Springer Nature in Scientific Data
- Vol. 13 (1)
Open AccessHSI-AgriFoodAnomaly, a hyperspectral dataset for foreign object detection in agri-food inspection
Ensuring food safety in agri-food production requires reliable inspection data for detecting foreign objects and other contamination risks. Hyperspectral imaging (HSI) is a relevant non-destructive modality for this purpose because it combines spatial information with detailed spectral responses. This Data Descriptor presents HSI-AgriFoodAnomaly, an open hyperspectral image dataset acquired under conveyor-based industrial-like conditions for foreign object annotation in an oat, white-chocolate and dark-chocolate mixture. The dataset contains 147 calibrated hyperspectral cubes, associated red-green-blue (RGB) renderings, pixel-level binary masks, and polygon annotations. Each hyperspectral cube was acquired in the visible to near-infrared range, with 300 contiguous spectral bands covering approximately 381–1016 nm. The annotated foreign objects (FOs) include textile and fibre-based materials, plastics, paper-based materials, metals, wood and plant residues, minerals, glass, mixed-object scenes, and anomaly-free scenes. The dataset is organised into training, validation and test subsets at the cube level, in order to prevent data leakage between splits. HSI-AgriFoodAnomaly can be reused for hyperspectral image analysis, foreign object localisation, binary classification, object detection, semantic segmentation, and benchmarking of data processing pipelines. The dataset and the accompanying code are publicly available.
- 7 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 23 (1545598X) , 5504405
MASCA: Multiscale Adaptive Spectral Component Analysis for Hyperspectral Image Classification
In this paper, we propose Multi-Scale Adaptive Spectral Component Analysis (MASCA), a frequency-domain preprocessing method that adaptively identifies discriminative Discrete Cosine Transform (DCT) components through multi-scale peak analysis. MASCA assigns a relevance score to each DCT component by detecting magnitude transitions across multiple scales while integrating spatial coherence and spectral stability constraints, with hyperparameters tuned via Bayesian optimization. Experimental results on four benchmark hyperspectral datasets (Indian Pines, Pavia University, Botswana, and Salinas) demonstrate the effectiveness of the proposed method.
- 7 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 23 (1545598X) , 5504405
MASCA: Multiscale Adaptive Spectral Component Analysis for Hyperspectral Image Classification
In this paper, we propose Multi-Scale Adaptive Spectral Component Analysis (MASCA), a frequency-domain preprocessing method that adaptively identifies discriminative Discrete Cosine Transform (DCT) components through multi-scale peak analysis. MASCA assigns a relevance score to each DCT component by detecting magnitude transitions across multiple scales while integrating spatial coherence and spectral stability constraints, with hyperparameters tuned via Bayesian optimization. Experimental results on four benchmark hyperspectral datasets (Indian Pines, Pavia University, Botswana, and Salinas) demonstrate the effectiveness of the proposed method.
- Open Access
- 24 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Scaling AI for Embryo Transition Detection: Benchmarking Spatio-Temporal Deep Models
- 24 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessScaling AI for Embryo Transition Detection: Benchmarking Spatio-Temporal Deep Models
- 12 December 2025
- Published by Springer Nature in Chromosome Research
- Vol. 33 (1) , 30
Exploring HSI imagery for overlapped chromosomes segmentation
The segmentation of overlapping chromosomes in metaphase images is a longstanding challenge in cytogenetics, where limited spectral contrast in conventional RGB imaging. In this study, we explore hyperspectral imaging (HSI) as a promising alternative modality to enhance segmentation accuracy by exploiting information from a narrow band of the spectrum. To evaluate the benefit of spectral data, we acquired and annotated a dedicated dataset of overlapping chromosomes, using a hyperspectral microscopy system. Multiple segmentation baseline models were trained and benchmarked on both RGB and HSI data. Experimental results show that HSI-based models consistently outperform RGB-based counterparts, achieving up to 92% Dice score. These findings underline the potential of HSI for improving cytogenetic image analysis and provide a new benchmark dataset for future research in spectral segmentation.
- 12 December 2025
- Published by Springer Nature in Chromosome Research
- Vol. 33 (1) , 30
Exploring HSI imagery for overlapped chromosomes segmentation
The segmentation of overlapping chromosomes in metaphase images is a longstanding challenge in cytogenetics, where limited spectral contrast in conventional RGB imaging. In this study, we explore hyperspectral imaging (HSI) as a promising alternative modality to enhance segmentation accuracy by exploiting information from a narrow band of the spectrum. To evaluate the benefit of spectral data, we acquired and annotated a dedicated dataset of overlapping chromosomes, using a hyperspectral microscopy system. Multiple segmentation baseline models were trained and benchmarked on both RGB and HSI data. Experimental results show that HSI-based models consistently outperform RGB-based counterparts, achieving up to 92% Dice score. These findings underline the potential of HSI for improving cytogenetic image analysis and provide a new benchmark dataset for future research in spectral segmentation.
- 12 December 2025
- Published by SAGE Publications in Applied Spectroscopy
- Vol. 80 (2) , 133-145
Exploring Band Selection Methods for Enhanced Chromosomal Analysis in Hyperspectral Imaging
Chromosome characterization is crucial in cytogenetic research and diagnostics, necessitating precise imaging methods to ensure proper analyses. The aim of this project is to identify a reliable method for chromosomal characterization that uses hyperspectral imagery of stained metaphase chromosomes using bright-field microscopy. We analyzed four hyperspectral images of stained chromosomes acquired under bright-field illumination. To address the high dimensionality of the hyperspectral hypercubes, we applied five dimension reduction algorithms based on spectral band selection to determine the most effective approach. A comparative study was conducted between five band selection methods to assess their effectiveness in chromosome characterization. The results indicate that sparse subspace clustering and multi-objective band selection are the most effective methods, outperforming the others in reducing the spectral dimensionality of the hyperspectral data, while preserving key properties essential for accurate chromosomes characterization. This study demonstrates that careful selection of spectral bands can enhance the analysis of spectral hypercubes for chromosome characterization.
- 12 December 2025
- Published by SAGE Publications in Applied Spectroscopy
- Vol. 80 (2) , 133-145
Exploring Band Selection Methods for Enhanced Chromosomal Analysis in Hyperspectral Imaging
Chromosome characterization is crucial in cytogenetic research and diagnostics, necessitating precise imaging methods to ensure proper analyses. The aim of this project is to identify a reliable method for chromosomal characterization that uses hyperspectral imagery of stained metaphase chromosomes using bright-field microscopy. We analyzed four hyperspectral images of stained chromosomes acquired under bright-field illumination. To address the high dimensionality of the hyperspectral hypercubes, we applied five dimension reduction algorithms based on spectral band selection to determine the most effective approach. A comparative study was conducted between five band selection methods to assess their effectiveness in chromosome characterization. The results indicate that sparse subspace clustering and multi-objective band selection are the most effective methods, outperforming the others in reducing the spectral dimensionality of the hyperspectral data, while preserving key properties essential for accurate chromosomes characterization. This study demonstrates that careful selection of spectral bands can enhance the analysis of spectral hypercubes for chromosome characterization.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 4469-4472
SPTPCA: Structure-Preserving Tensor Principal Component Analysis for Hyperspectral Dimensionality Reduction
Hyperspectral imaging generates high-dimensional data with complex spatial-spectral correlations that pose significant dimensionality reduction challenges. Principal Component Analysis (PCA) flattens the natural multidimensional tensor structure into vectors, causing loss of critical spatial relationships. Existing tensor methods including Tucker decomposition and Tensor Train (TT) provide low-rank approximations but do not extend PCA's variance optimization framework to tensor domains. In this paper, we present Structure-Preserving Tensor Principal Component Analysis (SPTPCA), a dimensionality reduction method based on the generalized tensor $\ast _{\mathcal {L}}$ -product framework that addresses this gap. Unlike standard PCA, SPTPCA operates directly on tensor representations, preserving natural structure and spatial-spectral correlations while maintaining variance optimization properties. Experimental validation on the Indian Pines dataset demonstrates MSE reductions of 7.9–50.0% and PSNR improvements of 0.35–2.59dB across different numbers of components, establishing a mathematically rigorous framework for structure-preserving hyperspectral dimensionality reduction.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 4469-4472
SPTPCA: Structure-Preserving Tensor Principal Component Analysis for Hyperspectral Dimensionality Reduction
Hyperspectral imaging generates high-dimensional data with complex spatial-spectral correlations that pose significant dimensionality reduction challenges. Principal Component Analysis (PCA) flattens the natural multidimensional tensor structure into vectors, causing loss of critical spatial relationships. Existing tensor methods including Tucker decomposition and Tensor Train (TT) provide low-rank approximations but do not extend PCA's variance optimization framework to tensor domains. In this paper, we present Structure-Preserving Tensor Principal Component Analysis (SPTPCA), a dimensionality reduction method based on the generalized tensor $\ast _{\mathcal {L}}$ -product framework that addresses this gap. Unlike standard PCA, SPTPCA operates directly on tensor representations, preserving natural structure and spatial-spectral correlations while maintaining variance optimization properties. Experimental validation on the Indian Pines dataset demonstrates MSE reductions of 7.9–50.0% and PSNR improvements of 0.35–2.59dB across different numbers of components, establishing a mathematically rigorous framework for structure-preserving hyperspectral dimensionality reduction.
- Open Access
- 11 September 2025
- Published by Springer Nature in Research Square
Exploring HSI imagery for overlapped chromosomes segmentation
The segmentation of overlapping chromosomes in metaphase images is alongstanding challenge in cytogenetics, where limited spectral contrast inconventional RGB imaging. In this study, we explore hyperspectral imaging(HSI) as a promising alternative modality to enhance segmentation accuracyby exploiting information from a narrow band of the spectrum. To evaluatethe benefit of spectral data, we acquired and annotated a dedicated dataset ofoverlapping chromosomes, using a hyperspectral microscopy system. Multiplesegmentation baseline models were trained and benchmarked on both RGB andHSI data. Experimental results show that HSI-based models consistently outper-form RGB-based counterparts, achieving up to 92% Dice score. These findingsunderline the potential of HSI for improving cytogenetic image analysis andprovide a new benchmark dataset for future research in spectral segmentation.
- 11 September 2025
- Published by Springer Nature in Research Square
Open AccessExploring HSI imagery for overlapped chromosomes segmentation
The segmentation of overlapping chromosomes in metaphase images is alongstanding challenge in cytogenetics, where limited spectral contrast inconventional RGB imaging. In this study, we explore hyperspectral imaging(HSI) as a promising alternative modality to enhance segmentation accuracyby exploiting information from a narrow band of the spectrum. To evaluatethe benefit of spectral data, we acquired and annotated a dedicated dataset ofoverlapping chromosomes, using a hyperspectral microscopy system. Multiplesegmentation baseline models were trained and benchmarked on both RGB andHSI data. Experimental results show that HSI-based models consistently outper-form RGB-based counterparts, achieving up to 92% Dice score. These findingsunderline the potential of HSI for improving cytogenetic image analysis andprovide a new benchmark dataset for future research in spectral segmentation.
- Open Access
Enhancing Embryo Stage Classification with Multi-Focal Plane Imaging and Deep Learning
Purpose: Accurately identifying embryo developmental stages is crucial for improving success rates in in vitro fertilization (IVF). Traditional embryo assessment relies on 2D imaging with a single focal plane, which may overlook critical morphological details and lead to misclassification. This study investigates whether incorporating depth information through multi-focal plane imaging can enhance the accuracy of embryo stage classification. Methods: We compared 2D and 3D convolutional neural network (CNN) architectures trained on a time-lapse embryo dataset containing seven focal planes. The models were evaluated based on their classification performance across different embryo developmental stages. Additionally, we used Gradient-weighted Class Activation Mapping (Grad-CAM) to analyze model attention and interpretability. Results: The findings demonstrate that 3D CNN models leveraging multi-focal plane data significantly outperform their 2D counterparts, particularly in complex stages such as Morula formation (tM) and expanded blastocyst (tSB, tB, and tEB). Grad-CAM visualization confirms that the models focus on relevant morphological structures, further validating the advantages of volumetric modeling. Conclusion: Multi-focal plane imaging enhances embryo stage classification accuracy by providing richer morphological information. This study underscores the potential of volumetric modeling in embryo assessment and paves the way for extending this approach to full 3D time-lapse analysis.
Open AccessEnhancing Embryo Stage Classification with Multi-Focal Plane Imaging and Deep Learning
Purpose: Accurately identifying embryo developmental stages is crucial for improving success rates in in vitro fertilization (IVF). Traditional embryo assessment relies on 2D imaging with a single focal plane, which may overlook critical morphological details and lead to misclassification. This study investigates whether incorporating depth information through multi-focal plane imaging can enhance the accuracy of embryo stage classification. Methods: We compared 2D and 3D convolutional neural network (CNN) architectures trained on a time-lapse embryo dataset containing seven focal planes. The models were evaluated based on their classification performance across different embryo developmental stages. Additionally, we used Gradient-weighted Class Activation Mapping (Grad-CAM) to analyze model attention and interpretability. Results: The findings demonstrate that 3D CNN models leveraging multi-focal plane data significantly outperform their 2D counterparts, particularly in complex stages such as Morula formation (tM) and expanded blastocyst (tSB, tB, and tEB). Grad-CAM visualization confirms that the models focus on relevant morphological structures, further validating the advantages of volumetric modeling. Conclusion: Multi-focal plane imaging enhances embryo stage classification accuracy by providing richer morphological information. This study underscores the potential of volumetric modeling in embryo assessment and paves the way for extending this approach to full 3D time-lapse analysis.