Adapting DeepLabV3+ for biopsy cervical cancer lesion segmentation - Takeaways - MDSpire

Adapting DeepLabV3+ for biopsy cervical cancer lesion segmentation

  • By

  • Rose Nakasi

  • Cosmas Wamozo

  • Solomon Nsumba

  • Benjamin Rukundo

  • Tonny Okecha

  • Byron Mubiru

  • Chodrine Mutebi

  • May 11, 2026

  • 0 min

Share

  • 1

    Cervical cancer is a leading cause of mortality, particularly in resource-constrained settings with limited access to advanced diagnostic tools.

  • 2

    The study presents a smartphone-assisted microscopy approach combined with DeepLabV3+ for effective segmentation of cervical cancer lesions.

  • 3

    The DeepLabV3+ model achieved a mean Intersection over Union of 75.8% and a Dice coefficient of 93.1% on a validation set of 5,966 images.

  • 4

    This method demonstrates the feasibility of digital pathology in low-resource environments, addressing significant healthcare disparities.

  • 5

    Future evaluations across multiple institutions are necessary to ensure the model's generalizability and effectiveness in diverse clinical settings.

Original Source(s)

Adapting DeepLabV3+ for biopsy cervical cancer lesion segmentation

  • by Rose Nakasi, Cosmas Wamozo, Solomon Nsumba, Benjamin Rukundo, Tonny Okecha, Byron Mubiru, Chodrine Mutebi

  • May 11, 2026

Related Content

Validation of a cervical CDO1/CELF4 methylation test for endometrial cancer: a prospective paired-sample comparison with intrauterine specimen

Trametinib as second-line therapy for advanced KRAS G12C-mutant non-small cell lung cancer: a single-center clinical analysis of 20 cases

Icaritin plus TACE improves survival in advanced HCC with macrovascular invasion: a multicenter cohort study