Selected work

Representative Papers

Exploiting Test-Time Augmentation in Federated Learning for Brain Tumor MRI Classification

Jan 19, 2026

This work addresses the suboptimal classification performance in federated learning for brain tumor MRI images, which stems from lesion heterogeneity and image complexity. The authors propose a federated learning framework that integrates lightweight preprocessing with test-time augmentation (TTA). They present the first systematic validation of TTA’s effectiveness in federated medical image classification and demonstrate that combining TTA with lightweight preprocessing techniques—such as normalization and histogram equalization—yields significant and consistent improvements in classification accuracy (p<0.001). The proposed approach achieves reliable performance gains while maintaining computational efficiency, making it well-suited for resource-constrained, distributed healthcare settings.

1 citationsRead paper

PRISM: Perinuclear Ring-based Image Segmentation Method for Acute Lymphoblastic Leukemia Classification

May 12, 2026

This study addresses the challenges in classifying acute lymphoblastic leukemia (ALL) cells from blood smears, where low cytoplasmic contrast and high morphological variability render conventional membrane-based segmentation methods ineffective and limit the generalization of existing deep learning models. To overcome these limitations, the authors propose PRISM, a novel approach that eschews precise cytoplasmic segmentation and instead constructs adaptive concentric rings around the nucleus. This framework integrates color features with gray-level co-occurrence matrix–derived texture information and employs a calibrated stacked ensemble classifier for robust discrimination. By eliminating reliance on cell boundary delineation, PRISM demonstrates strong robustness and generalizability across diverse staining protocols and imaging conditions, achieving 98.46% accuracy and a PR-AUC of 0.9937 in ALL classification.

0 citationsRead paper

A chaotic flux cipher based on the random cubic family $f_{c_n}(z)=z^3+c_n z$

Mar 21, 2026

This work proposes a symmetric stream cipher scheme based on chaotic dynamics in the complex plane to address the demanding requirements of high security and noise resilience in complex communication environments such as 5G. The method innovatively introduces the chaotic behavior of random cubic polynomial maps into cryptography, leveraging a control parameter δ to toggle between stable and chaotic regimes. It generates pseudorandom keystreams by exploiting the structural stability of Julia sets and integrates HKDF key derivation, HMAC-SHA-256 authenticated encryption, and a warm-up iteration mechanism. Experimental results demonstrate that the generated keystreams pass the full NIST SP 800-22 statistical test suite, χ² tests, and entropy analysis, achieving high randomness, strong security, and robustness against noise while maintaining key consistency.

0 citationsRead paper

Real-Time 2D LiDAR Object Detection Using Three-Frame RGB Scan Encoding

Feb 02, 2026

This work addresses the urgent need for efficient, privacy-preserving real-time perception in indoor service robots on embedded platforms. To overcome the privacy risks of conventional RGB-based approaches and the high computational cost or temporal information neglect of existing LiDAR methods, the authors propose a purely 2D LiDAR-based object detection framework. By encoding three consecutive LiDAR scans into an image-like RGB representation that preserves angular structure and motion cues, the method directly feeds this input into a lightweight YOLOv8n model—bypassing occupancy grid construction entirely. Evaluated in Webots across 160 random indoor scenes, the approach achieves 98.4% mAP@0.5 and 77.8% mAP@0.5:0.95, with both precision and recall exceeding 94.7%. On a Raspberry Pi 5, it attains an end-to-end latency of only 47.8 ms, significantly outperforming existing grid-based methods.

0 citationsRead paper

Generalizable Hyperparameter Optimization for Federated Learning on Non-IID Cancer Images

Jan 19, 2026

This work addresses the challenge of limited model generalization and sensitivity to hyperparameter selection in federated learning with non-IID cancer histopathology images. The authors propose a cross-dataset hyperparameter transfer strategy: first, Bayesian optimization is applied under a centralized setting to identify optimal hyperparameters separately for ovarian and colorectal cancer datasets; then, a simple yet effective aggregation heuristic—averaging learning rates and selecting the mode for optimizer and batch size—is used to construct a universal configuration, which is subsequently transferred to the federated learning scenario. Experimental results demonstrate that this approach significantly improves classification performance under non-IID conditions, confirming the efficacy and practicality of cross-dataset hyperparameter transfer in federated medical image analysis.

0 citationsRead paper

Recent publications

Latest Papers

PRISM: Perinuclear Ring-based Image Segmentation Method for Acute Lymphoblastic Leukemia Classification

May 12, 2026

This study addresses the challenges in classifying acute lymphoblastic leukemia (ALL) cells from blood smears, where low cytoplasmic contrast and high morphological variability render conventional membrane-based segmentation methods ineffective and limit the generalization of existing deep learning models. To overcome these limitations, the authors propose PRISM, a novel approach that eschews precise cytoplasmic segmentation and instead constructs adaptive concentric rings around the nucleus. This framework integrates color features with gray-level co-occurrence matrix–derived texture information and employs a calibrated stacked ensemble classifier for robust discrimination. By eliminating reliance on cell boundary delineation, PRISM demonstrates strong robustness and generalizability across diverse staining protocols and imaging conditions, achieving 98.46% accuracy and a PR-AUC of 0.9937 in ALL classification.

0 citationsRead paper

A chaotic flux cipher based on the random cubic family $f_{c_n}(z)=z^3+c_n z$

Mar 21, 2026

This work proposes a symmetric stream cipher scheme based on chaotic dynamics in the complex plane to address the demanding requirements of high security and noise resilience in complex communication environments such as 5G. The method innovatively introduces the chaotic behavior of random cubic polynomial maps into cryptography, leveraging a control parameter δ to toggle between stable and chaotic regimes. It generates pseudorandom keystreams by exploiting the structural stability of Julia sets and integrates HKDF key derivation, HMAC-SHA-256 authenticated encryption, and a warm-up iteration mechanism. Experimental results demonstrate that the generated keystreams pass the full NIST SP 800-22 statistical test suite, χ² tests, and entropy analysis, achieving high randomness, strong security, and robustness against noise while maintaining key consistency.

0 citationsRead paper

Real-Time 2D LiDAR Object Detection Using Three-Frame RGB Scan Encoding

Feb 02, 2026

This work addresses the urgent need for efficient, privacy-preserving real-time perception in indoor service robots on embedded platforms. To overcome the privacy risks of conventional RGB-based approaches and the high computational cost or temporal information neglect of existing LiDAR methods, the authors propose a purely 2D LiDAR-based object detection framework. By encoding three consecutive LiDAR scans into an image-like RGB representation that preserves angular structure and motion cues, the method directly feeds this input into a lightweight YOLOv8n model—bypassing occupancy grid construction entirely. Evaluated in Webots across 160 random indoor scenes, the approach achieves 98.4% mAP@0.5 and 77.8% mAP@0.5:0.95, with both precision and recall exceeding 94.7%. On a Raspberry Pi 5, it attains an end-to-end latency of only 47.8 ms, significantly outperforming existing grid-based methods.

0 citationsRead paper

Exploiting Test-Time Augmentation in Federated Learning for Brain Tumor MRI Classification

Jan 19, 2026

This work addresses the suboptimal classification performance in federated learning for brain tumor MRI images, which stems from lesion heterogeneity and image complexity. The authors propose a federated learning framework that integrates lightweight preprocessing with test-time augmentation (TTA). They present the first systematic validation of TTA’s effectiveness in federated medical image classification and demonstrate that combining TTA with lightweight preprocessing techniques—such as normalization and histogram equalization—yields significant and consistent improvements in classification accuracy (p<0.001). The proposed approach achieves reliable performance gains while maintaining computational efficiency, making it well-suited for resource-constrained, distributed healthcare settings.

1 citationsRead paper

Generalizable Hyperparameter Optimization for Federated Learning on Non-IID Cancer Images

Jan 19, 2026

This work addresses the challenge of limited model generalization and sensitivity to hyperparameter selection in federated learning with non-IID cancer histopathology images. The authors propose a cross-dataset hyperparameter transfer strategy: first, Bayesian optimization is applied under a centralized setting to identify optimal hyperparameters separately for ovarian and colorectal cancer datasets; then, a simple yet effective aggregation heuristic—averaging learning rates and selecting the mode for optimizer and batch size—is used to construct a universal configuration, which is subsequently transferred to the federated learning scenario. Experimental results demonstrate that this approach significantly improves classification performance under non-IID conditions, confirming the efficacy and practicality of cross-dataset hyperparameter transfer in federated medical image analysis.

0 citationsRead paper