DualMiT-Net: Local-Global Transformer-Convolutional Fusion for Breast Mass Segmentation in Mammographic Regions of Interest

Alibek Kamiluly; Milana Muratova; Yash Patel; Fan Li

cs.CVAug 15, 2026

DualMiT-Net: Local-Global Transformer-Convolutional Fusion for Breast Mass Segmentation in Mammographic Regions of Interest

Authors: Alibek Kamiluly, Milana Muratova, Yash Patel, Fan Li

Organizations: Department of Mathematics and Computer Science, Lawrence Technological University · College of Business and Information Technology, Lawrence Technological University

Abstract

Breast mass segmentation is an important step in computer-aided mammography, but it remains difficult because masses can have low contrast, irregular shapes, and boundaries that blend with surrounding breast tissue. To address this problem, we present DualMiT-Net, a dual-branch network that uses both a focused view of the mass and a wider view of the surrounding tissue. The local branch uses a Mix Transformer (MiT-B5) encoder to learn mass shape, texture, and boundary information, while the global branch uses an EfficientNet-B5 encoder to learn surrounding breast context. Features from the two branches are shared at the deeper encoder levels and are then progressively fused in a single decoder. A spatial gate controls how much global information is added during decoding. We also evaluated four input representations and selected a percentile-windowed mammogram combined with a Gabor texture response. The model was trained and evaluated on the mass subset of the Curated Breast Imaging Subset of the Digital Database for Screening Mammography (CBIS-DDSM) using a patient-level split. Across three training runs, DualMiT-Net with exponential moving average weights achieved a mean Dice coefficient of 0.9375 and a mean Intersection over Union of 0.8834. It also achieved better Dice and IoU scores than six standard encoder-decoder baselines trained using the same data and training settings. These results show that combining local mass information with wider breast context can provide accurate and consistent breast mass segmentation.

Figures & tables

Study	Method	Main idea	Dice	IoU
AUNet [ 18 ]	Attention-based encoder-decoder	Attention-guided dense upsampling for whole mammograms	0.8180	-
AM-MSP-cGAN [ 19 ]	Conditional GAN	Attention and multi-scale pooling with adversarial learning	0.8449	-
ARF-Net [ 22 ]	Adaptive receptive-field network	Selects receptive fields for masses with different sizes	0.8575	-
MTLNet [ 20 ]	Attentive multi-task network	Joint segmentation, localization, and classification	0.8630	-
Connected-ResUNets [ 21 ]	Connected residual U-Net	Connected U-Net stages with residual learning	0.8952	0.8002
YOLO-LOGO [ 15 ]	Local-global transformer	Detection followed by local and global transformer segmentation	0.7450	0.6400

Table 1: Summary of selected breast mass segmentation methods evaluated on CBIS-DDSM.

Figure 1: Overview of the proposed segmentation pipeline.

Subset	Patients	Paired cases	Image files	Mask files
Training	480	753	753	753
Validation	62	91	91	91
Test	64	92	92	92

Table 2: Patient-level distribution of the CBIS-DDSM mass cases used in this study.

Figure 2: Examples from the CBIS-DDSM mass subset. The top row shows mammograms and the bottom row shows the corresponding expert mass masks.

Figure 3: DualMiT-Net architecture. The local and global branches are shown in blue and green, respectively, and their features are combined progressively in the decoder.

Architecture	Weights	Dice	IoU	Precision	Recall	Specificity	Accuracy	HD95 $\downarrow$
Full DualMiT-Net	Base	$0.9369\pm 0.0012$	$0.8824\pm 0.0020$	$0.9333\pm 0.0024$	$0.9418\pm 0.0040$	$0.9102\pm 0.0039$	$0.9278\pm 0.0012$	$6.85\pm 0.25$
Full DualMiT-Net	EMA	$\mathbf{0.9375\pm 0.0011}$	$\mathbf{0.8834\pm 0.0020}$	$0.9282\pm 0.0052$	$0.9482\pm 0.0037$	$0.9022\pm 0.0080$	$0.9280\pm 0.0015$	$6.88\pm 0.23$
Without scSE	Base	$0.9368\pm 0.0011$	$0.8822\pm 0.0019$	$0.9368\pm 0.0035$	$0.9380\pm 0.0050$	$0.9158\pm 0.0052$	$0.9280\pm 0.0010$	$6.82\pm 0.14$
Without scSE	EMA	$0.9372\pm 0.0002$	$0.8829\pm 0.0004$	$0.9238\pm 0.0012$	$0.9522\pm 0.0014$	$0.8953\pm 0.0018$	$0.9273\pm 0.0002$	$6.95\pm 0.09$
Without G2L sharing	Base	$0.9363\pm 0.0003$	$0.8814\pm 0.0006$	$0.9267\pm 0.0065$	$0.9474\pm 0.0065$	$0.8998\pm 0.0103$	$0.9266\pm 0.0007$	$6.97\pm 0.17$
Without G2L sharing	EMA	$0.9371\pm 0.0006$	$0.8829\pm 0.0011$	$0.9273\pm 0.0026$	$0.9484\pm 0.0018$	$0.9008\pm 0.0038$	$0.9275\pm 0.0008$	$6.93\pm 0.06$

Table 3: Segmentation performance of DualMiT-Net and the two ablation variants.

Input representation	Base Dice	Base IoU	Base HD95	EMA Dice	EMA IoU	EMA HD95
Window only	$0.9364\pm 0.0278$	$0.8817\pm 0.0480$	$6.88\pm 3.70$	$0.9363\pm 0.0277$	$0.8815\pm 0.0478$	$7.00\pm 3.93$
Window + CLAHE	$0.9361\pm 0.0268$	$0.8811\pm 0.0465$	$7.09\pm 4.13$	$0.9363\pm 0.0266$	$0.8814\pm 0.0462$	$7.21\pm 4.20$
Window + Gabor	$\mathbf{0.9372\pm 0.0248}$	$\mathbf{0.8829\pm 0.0433}$	$\mathbf{6.70\pm 2.90}$	$\mathbf{0.9376\pm 0.0250}$	$\mathbf{0.8836\pm 0.0438}$	$\mathbf{6.85\pm 3.43}$
Window + CLAHE + Gabor	$0.9356\pm 0.0276$	$0.8802\pm 0.0477$	$7.01\pm 3.97$	$0.9356\pm 0.0272$	$0.8802\pm 0.0470$	$7.14\pm 3.96$

Table 4: Comparison of the four input representations.

Architecture	Weights	Dice	IoU	Precision	Recall	Specificity	Accuracy	HD95 $\downarrow$
Full DualMiT-Net	Base	0.9369 [0.9336, 0.9401]	0.8824 [0.8767, 0.8879]	0.9333 [0.9280, 0.9383]	0.9418 [0.9364, 0.9467]	0.9102 [0.9029, 0.9168]	0.9278 [0.9242, 0.9312]	6.85 [6.42, 7.33]
Full DualMiT-Net	EMA	0.9375 [0.9342, 0.9406]	0.8834 [0.8777, 0.8889]	0.9282 [0.9212, 0.9344]	0.9482 [0.9436, 0.9529]	0.9022 [0.8924, 0.9111]	0.9280 [0.9242, 0.9316]	6.88 [6.45, 7.39]
Without G2L sharing	Base	0.9363 [0.9331, 0.9393]	0.8814 [0.8758, 0.8867]	0.9267 [0.9194, 0.9345]	0.9474 [0.9400, 0.9538]	0.8998 [0.8892, 0.9115]	0.9266 [0.9230, 0.9300]	6.97 [6.57, 7.42]
Without G2L sharing	EMA	0.9371 [0.9339, 0.9402]	0.8829 [0.8772, 0.8883]	0.9273 [0.9219, 0.9326]	0.9484 [0.9445, 0.9522]	0.9008 [0.8936, 0.9076]	0.9275 [0.9239, 0.9310]	6.93 [6.56, 7.33]
Without scSE	Base	0.9368 [0.9334, 0.9400]	0.8822 [0.8764, 0.8878]	0.9368 [0.9310, 0.9422]	0.9380 [0.9320, 0.9437]	0.9158 [0.9081, 0.9230]	0.9280 [0.9244, 0.9314]	6.82 [6.44, 7.26]
Without scSE	EMA	0.9372 [0.9340, 0.9402]	0.8829 [0.8773, 0.8883]	0.9238 [0.9187, 0.9287]	0.9522 [0.9488, 0.9555]	0.8953 [0.8889, 0.9016]	0.9273 [0.9237, 0.9307]	6.95 [6.54, 7.43]

Table 5: Seed-averaged results with 95% bootstrap confidence intervals.

Panel of Figure 4 — Figure 4: Per-image Dice, IoU, and HD95 distributions for the three architectures using EMA weights.

Comparison	Metric	Mean difference	95% CI	$p$ -value	Significance
Full vs. without scSE	Dice	$0.0003$	$[-0.0005,0.0011]$	$0.501$	n.s.
Full vs. without G2L sharing	Dice	$0.0004$	$[-0.0003,0.0010]$	$0.291$	n.s.
Full vs. without scSE	IoU	$0.0005$	$[-0.0009,0.0019]$	$0.513$	n.s.
Full vs. without G2L sharing	IoU	$0.0006$	$[-0.0005,0.0017]$	$0.300$	n.s.
Full vs. without scSE	HD95	$-0.07$	$[-0.23,0.08]$	$0.507$	n.s.
Full vs. without G2L sharing	HD95	$-0.05$	$[-0.19,0.10]$	$0.315$	n.s.

Table 6: Paired Wilcoxon signed-rank tests for the two ablation comparisons.

Figure 5: Representative segmentation results for DualMiT-Net.

Figure 6: Test cases with the lowest Dice scores.

Model	Dice	IoU	Precision	Recall
U-Net (ResNet-50)	$0.9025\pm 0.0339$	$0.8240\pm 0.0548$	$0.8846\pm 0.0524$	$0.9233\pm 0.0341$
U-Net++ (ResNet-50)	$0.9023\pm 0.0321$	$0.8235\pm 0.0523$	$0.8806\pm 0.0530$	$0.9276\pm 0.0337$
U-Net++ (ResNet-34)	$0.9009\pm 0.0315$	$0.8211\pm 0.0514$	$0.8870\pm 0.0527$	$0.9174\pm 0.0301$
U-Net (ResNet-34)	$0.9005\pm 0.0344$	$0.8207\pm 0.0555$	$0.8772\pm 0.0544$	$0.9271\pm 0.0290$
FPN (ResNet-34)	$0.9002\pm 0.0318$	$0.8201\pm 0.0514$	$0.8770\pm 0.0533$	$0.9269\pm 0.0274$
PSPNet (ResNet-34)	$0.8941\pm 0.0363$	$0.8103\pm 0.0582$	$0.8787\pm 0.0551$	$0.9126\pm 0.0402$

Table 7: Comparison of DualMiT-Net with standard segmentation baselines on CBIS-DDSM.

Method	Dice	IoU
YOLO-LOGO [ 15 ]	0.7450	0.6400
AUNet [ 18 ]	0.8180	-
AM-MSP-cGAN [ 19 ]	0.8449	-
ARF-Net [ 22 ]	0.8575	-
MTLNet [ 20 ]	0.8630	-
Connected-ResUNets [ 21 ]	0.8952	0.8002

Table 8: Comparison of breast mass segmentation methods on CBIS-DDSM.

Medical Image Segmentation

Medical image segmentation delineates organs, lesions, and other structures in scans to support measurement, diagnosis, treatment planning, and research.

See all 788 papers

BiLoG-Net: A Bi-Context Location-Guided Network for Breast Mass Segmentation and Malignancy Classification in Mammography
Jul 11, 2026Abu Fatema Mohammad Abdun Noor, Md Imam Ahasan, Md Samiul Ahasan +3Multi-Task Learning Medical Image Classification
Department of Software Engineering, Daffodil International University, Dhaka, Bangladesh · BricksCloud AI, Uttara, Dhaka, Bangladesh · College of Computer Science, Chongqing University, Chongqing, China +3
M3D-Net: Hierarchical Coordination of Spatial Context, Feature Reuse, and Differential Attention for Mammography Classification
Sep 23, 2026Zheng Yu, Xinhang Li, Jiabao Gao +2Multi-Scale Feature Fusion Medical Image Classification
Shenzhen Loop Area Institute, Shenzhen, China · The Chinese University of Hong Kong, Shenzhen, China · Shenzhen Research Institute of Big Data, Shenzhen, China
Subtraction-Based Tumor Segmentation and Lesion-Centered pCR Prediction for the MAMA-MIA Challenge
Aug 29, 2026Kai Geissler, Raphael SchäferClinical Outcome Prediction Medical Image Segmentation
Fraunhofer Institute for Digital Medicine MEVIS Bremen, Germany

Jul 11, 2026J/K move · Enter open · S save

BiLoG-Net: A Bi-Context Location-Guided Network for Breast Mass Segmentation and Malignancy Classification in Mammography

Abu Fatema Mohammad Abdun Noor, Md Imam Ahasan, Md Samiul Ahasan, Kah Ong Michael Goh+2

Department of Software Engineering, Daffodil International University, Dhaka, Bangladesh · BricksCloud AI, Uttara, Dhaka, Bangladesh · College of Computer Science, Chongqing University, Chongqing, China+3

Multi-Task Learning Medical Image Classification Medical Image Segmentation Mammography

Open paper AskPDF arXiv

Breast cancer remains the most commonly diagnosed malignancy among women worldwide, yet accurate detection and characterization of breast masses in mammography remain challenging due to subtle intensity variations, heterogeneous tissue densities, and indistinct lesion boundaries that complicate radiological interpretation. To address these limitations, we propose BiLoG-Net, a deep learning framework that jointly performs breast mass segmentation and malignancy classification through bi-context location-aware feature modeling and segmentation-guided attention mechanisms. Our architecture integrates a novel encoder-decoder paradigm with Fire-based feature extraction, lightweight global and local feature enhancement modules, and adaptive location-aware gating to simultaneously capture long-range contextual dependencies and fine-grained boundary-sensitive details. Unlike conventional multi-stage pipelines, our tightly coupled multi-task design enables mutual reinforcement between pixel-level localization and image-level diagnosis, reducing error propagation while producing spatially grounded malignancy predictions. Evaluated on CBIS-DDSM and INBreast benchmarks, BiLoG-Net achieves state-of-the-art performance with Dice scores of 94.20% and 93.10%, classification accuracies of 95.20% and 93.60%, and AUC values of 97.10% and 96.00%, respectively, substantially outperforming existing CNN and transformer-based baselines. By combining precise boundary delineation with reliable malignancy assessment in a single end-to-end model, this work holds strong potential for clinical computer-aided detection systems, helping radiologists prioritize suspicious cases and improve screening efficiency in busy clinical settings.

Abstract

Medical Image Segmentation

Explore similar work

BiLoG-Net: A Bi-Context Location-Guided Network for Breast Mass Segmentation and Malignancy Classification in Mammography

M3D-Net: Hierarchical Coordination of Spatial Context, Feature Reuse, and Differential Attention for Mammography Classification

Subtraction-Based Tumor Segmentation and Lesion-Centered pCR Prediction for the MAMA-MIA Challenge