ToxoSegFusion: Attention-enhanced Dual-backbone Neural Architecture for Retinal Lesion Segmentation

Md Mominul Haque; Saipunidzam Mahamad; Suziah Sulaiman; Abdullateef Oluwabemiga Balogun; Hussaini Mamman

doi:10.31436/iiumej.v27i1.3806

Authors

Md Mominul Haque Universiti Teknologi Petronas https://orcid.org/0009-0005-0596-2406
Saipunidzam Mahamad Universiti Teknologi Petronas https://orcid.org/0000-0002-9300-4363
Suziah Sulaiman Universiti Teknologi Petronas https://orcid.org/0000-0003-0956-9853
Abdullateef Oluwabemiga Balogun Universiti Teknologi Petronas https://orcid.org/0000-0001-7411-3639
Hussaini Mamman Universiti Teknologi Petronas https://orcid.org/0009-0005-0404-3125

DOI:

https://doi.org/10.31436/iiumej.v27i1.3806

Keywords:

Retinal Image Segmentation, Deep Learning, ToxoSegFusion

Abstract

Ocular toxoplasmosis (OT) often presents a diagnostic dilemma in clinics, with retinal lesions that are not only varied in appearance but also frequently subtle and underrepresented in fundus images. Current automated segmentation tools, though promising, are often hampered by class imbalance and a lack of robust testing across real-world scenarios. To address these gaps, we developed ToxoSegFusion, a dual-backbone deep learning framework that capitalizes on the complementary strengths of DenseNet121 and ResNet101, enhanced with attention modules. Unlike typical single-backbone models, this hybrid approach was specifically tuned for the intricate challenges of OT lesion segmentation, using a combined Dice and binary cross-entropy loss to better balance rare lesion pixels. We trained and validated on 149 image-mask pairs from the OTFID-Version 3 dataset, achieving an intersection over union of 0.858 and a Dice coefficient of 0.795, both exceeding the current MobileNetV2/U-Net baseline. The model also demonstrated reliable performance on the DRIVE dataset for vessel segmentation, indicating practical flexibility. By facilitating accurate lesion localization, ToxoSegFusion enables more timely interventions in ophthalmology. Future directions include larger multi-center trials and streamlined models for routine deployment.

ABSTRAK: Toksoplasmosis okular (OT) sering menimbulkan cabaran diagnostik di klinik, dengan lesi retina halus pelbagai rupa dan kurang terwakili pada imej fundus. Alat segmentasi automatik semasa, walaupun memberi harapan, sering terhad pada ketidakseimbangan kelas dan kekurangan ujian di peringkat perubatan. Bagi mengatasi kekurangan ini, kajian ini membangunkan ToxoSegFusion, sebuah rangka kerja pembelajaran mendalam berkomponen dua yang memanfaatkan kekuatan saling melengkapi DenseNet121 dan ResNet101, diperkaya dengan mekanisme perhatian. Tidak seperti model komponen tunggal biasa, pendekatan hibrid ini dirancang khusus bagi cabaran kompleks segmentasi lesi OT, menggunakan kehilangan Dice dan entropi silang binari gabungan bagi keseimbangan terbaik antara piksel lesi yang jarang. Kajian ini melatih dan mengesahkan 149 pasangan imej-topeng dari set data OTFID-Versi 3, mencapai persilangan atas kesatuan 0.858 dan pekali Dice 0.795, keduanya melebihi garis dasar MobileNetV2/U-Net semasa. Model juga menunjukkan prestasi terbaik pada DRIVE bagi segmentasi salur darah, mencadangkan fleksibiliti praktis. Melalui pengesanan lokasi lesi yang tepat, ToxoSegFusion membuka jalan bagi intervensi lebih tepat pada masa oftalmologi. Pada masa hadapan, cadangan bagi penyebaran rutin adalah melalui ujian berbilang pusat yang lebih besar dan perkemasan model.

Downloads

Download data is not yet available.

Metrics

Metrics Loading ...

References

D. Etya’ale, “Blindness and vision impairment,” in Global health essentials, pp. 209–213, Springer, 2023.

C. Brandão-de Resende, M. B. Balasundaram, S. Narain, P. Mahendradas, and D. V. Vasconcelos-Santos, “Multimodal imaging in ocular toxoplasmosis,” Ocular immunology and inflammation, vol. 28, no. 8, pp. 1196–1204, 2020.

A. Gupta, R. Bansal, A. Sharma, and A. Kapil, “Retinal and choroidal infections and inflammation,” in Ophthalmic Signs in Practice of Medicine, pp. 205–270, Springer, 2024.

G. N. Holland, “Ocular toxoplasmosis: A global reassessment. part i: Epidemiology and course of disease,” American Journal of Ophthalmology, vol. 136, no. 6, pp. 973–988, 2003.

N. J. Butler, R. B. Furtado, K. L. Winthrop, and J. R. Smith, “Ocular toxoplasmosis: Vari- ability in clinical presentation and diagnostic challenges,” Expert Review of Ophthalmology, vol. 8, no. 4, pp. 349–357, 2013.

M. Gupta, S. Gupta, G. Palanisamy, J. Nisha, V. Goutham, S. A. Kumar, K. Gavaskar, and G. R. Naik, “A comprehensive survey on detection of ocular and non-ocular diseases using color fundus images,” IEEE Access, 2024.

W. Lai and H. Menghan, “A review of medical ocular image segmentation,” Virtual Reality & Intelligent Hardware, vol. 6, no. 3, pp. 181–202, 2024.

N. Patton, T. M. Aslam, T. MacGillivray, et al., “Retinal image analysis: Concepts, appli-cations and potential,” Progress in Retinal and Eye Research, vol. 25, no. 1, pp. 99–127, 2006.

U. R. Acharya, C. M. Lim, E. Y. K. Ng, et al., “Computer-based detection of diabetes retinopathy stages using digital fundus images,” Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine, vol. 223, no. 5, pp. 545–553, 2008.

H. Jiang, Z. Diao, T. Shi, Y. Zhou, F. Wang, W. Hu, X. Zhu, S. Luo, G. Tong, and Y.-

D. Yao, “A review of deep learning-based multiple-lesion recognition from medical images: classification, detection and segmentation,” Computers in Biology and Medicine, vol. 157,p. 106726, 2023.

S. S. Alam, S. B. Shuvo, S. N. Ali, F. Ahmed, A. Chakma, and Y. M. Jang, “Benchmarking deep learning frameworks for automated diagnosis of ocular toxoplasmosis: A comprehen- sive approach to classification and segmentation,” IEEE Access, vol. 12, pp. 22759–22777, 2024.

D. Cardozo et al., “Multiclass classification of ocular toxoplasmosis from fundus images with residual neural networks,” Computational Science and Its Applications, 2023.

M. Aziz et al., “Optimizing few-shot learning via reptile meta-learning approach for toxo- plasmosis chorioretinitis detection,” arXiv preprint, 2024.

K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in

Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016.

G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, “Densely connected convolu- tional networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4700–4708, 2017.

M. Maenz, U. Schluter, S. Nagel, et al., “Reliability of standardized protocols for the diag- nosis of ocular toxoplasmosis,” Acta Ophthalmologica, vol. 92, no. S253, 2014.

J. Lowell, A. Hunter, D. Steel, et al., “Optic nerve head segmentation,” IEEE Transactions on Medical Imaging, vol. 23, no. 2, pp. 256–264, 2004.

F. Li, H. Chen, Z. Liu, X. Zhang, and Z. Wu, “Fully automated detection of retinal disorders by image-based deep learning,” Graefe’s Archive for Clinical and Experimental Ophthalmol- ogy, vol. 257, pp. 495–505, 2019.

O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” in Medical Image Computing and Computer-Assisted Intervention– MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Pro- ceedings, Part III 18, pp. 234–241, Springer, 2015.

L.-C. Chen, Y. Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder-decoder with atrous separable convolution for semantic image segmentation,” in Proceedings of the European conference on computer vision (ECCV), pp. 801–818, 2018.

P. Porwal, S. Pachade, R. Kamble, M. Kokare, G. Deshmukh, V. Sahasrabuddhe, and

F. Meriaudeau, “Indian diabetic retinopathy image dataset (idrid),” 2018.

M. M. Haque, S. Akter, and A. F. Ashrafi, “Swinmednet: Leveraging swin transformer for robust diabetic retinopathy classification from the retinamnist2d dataset,” in 2024 6th In- ternational Conference on Electrical Engineering and Information & Communication Tech- nology (ICEEICT), pp. 1286–1291, 2024.

J. Chen, Y. Lu, Q. Yu, X. Luo, E. Adeli, Y. Wang, L. Lu, A. L. Yuille, and Y. Zhou, “Transunet: Transformers make strong encoders for medical image segmentation,” arXiv preprint arXiv:2102.04306, 2021.

[24] S. Karkuzhali, P. Thendal, and S. Senthilkumar, “Medical image analysis based on deep learning approach and internet of medical things (iomt) for early diagnosis of retinal dis- ease,” in Internet of Things enabled Machine Learning for Biomedical Applications, pp. 188– 201, CRC Press, 2024.

M. Ragab, E. Eldele, M. Chen, et al., “Deep learning for retinal image analysis: A review,” Artificial Intelligence in Medicine, vol. 145, p. 102659, 2023.

M. Biglarbeiki, “Improving classification and segmentation of choroidal lesions by addressing data limitations with patch-based approaches,” 2024.

C. Iriondo, Characterizing Phenotypes of Musculoskeletal Degeneration Using Medical Imag- ing and Deep Learning. University of California, San Francisco, 2021.

Y. Zhou, B. Wang, L. He, et al., “Generative adversarial network for retinal image synthesis,” 2021.

C. Luo et al., “Universal medical imaging model for domain generalization with data pri- vacy,” arXiv preprint, 2024.

A. Buslaev, V. I. Iglovikov, E. Khvedchenya, A. Parinov, M. Druzhinin, and A. A. Kalinin, “Albumentations: Fast and flexible image augmentations,” Information, vol. 11, no. 2, p. 125, 2020.

C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learn- ing,” Journal of Big Data, vol. 6, no. 1, pp. 1–48, 2019.

K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” arXiv preprint arXiv:1409.1556, 2014.

A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convo- lutional neural networks,” in Advances in Neural Information Processing Systems, vol. 25, pp. 1097–1105, 2012.

L. Perez and J. Wang, “The effectiveness of data augmentation in image classification using deep learning,” arXiv preprint arXiv:1712.04621, 2017.

I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. Cambridge, MA: MIT Press, 2016.

P. Y. Simard, D. Steinkraus, and J. C. Platt, “Best practices for convolutional neural networks applied to visual document analysis,” in Seventh International Conference on Document Analysis and Recognition (ICDAR), pp. 958–963, IEEE, 2003.

N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: A simple way to prevent neural networks from overfitting,” The Journal of Machine Learning Research, vol. 15, no. 1, pp. 1929–1958, 2014.

N. Abraham and N. M. Khan, “A novel focal tversky loss function with improved attention u-net for lesion segmentation,” in 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019), pp. 683–687, IEEE, 2019.

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in International Conference on Learning Representations (ICLR), 2015.

O. Cardozo, V. Ojeda, R. Parra, J. C. Mello-Román, J. L. V. Noguera, M. García-Torres, F. Divina, S. A. Grillo, C. Villalba, J. Facon, V. E. C. Benítez, I. C. Matto, and D. Aquino- Brítez, “Dataset of fundus images for the diagnosis of ocular toxoplasmosis,” Data in Brief, vol. 48, 6 2023.

J. Staal, M. Abramoff, M. Niemeijer, M. Viergever, and B. van Ginneken, “Ridge-based vessel segmentation in color images of the retina,” IEEE Transactions on Medical Imaging, vol. 23, no. 4, pp. 501–509, 2004.

S. R. Ferdous, M. R. A. Rifat, M. J. Ayan, and R. Rahman, “An approach to classify ocular toxoplasmosis images using deep learning models,” in 2023 26th International Conference on Computer and Information Technology, ICCIT 2023, Institute of Electrical and Electronics Engineers Inc., 2023.

R. Field, “Deepvessel: Retinal vessel segmentation,” in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, October 17-21, 2016, Proceedings, Part II, vol. 9901, p. 132, Springer, 2016.

S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, “Cbam: Convolutional block attention module,” in Proceedings of the European Conference on Computer Vision (ECCV), pp. 3–19, 2018.