JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Utilizing deep learning model for assessing melanocytic density in resection margins of lentigo maligna

  • 0Department of Laboratory Medicine, Institute of Biomedicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden. jan.siarov@vgregion.se.
Diagnostic Pathology +

|

August 3, 2024

|

August 3, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

A new deep learning model accurately detects lentigo maligna (LM) margins, aiding pathologists in assessing recurrence risk. This AI tool improves diagnostic accuracy for melanoma precursor lesions.

Area Of Science

  • Dermatopathology
  • Artificial Intelligence in Medicine
  • Oncology

Background

  • Surgical excision with clear margins is crucial for treating lentigo maligna (LM) to prevent invasive melanoma.
  • Assessing resection margins in sun-damaged skin for LM is diagnostically challenging.
  • A deep learning (AI) model was developed to detect melanocytes in LM resection margins.

Purpose Of The Study

  • To evaluate the accuracy of a deep learning model in identifying high-risk lentigo maligna (LM) resection margins.
  • To compare the AI model's performance against dermatopathologists and pathology residents.
  • To assess the impact of AI assistance on pathologist performance in margin evaluation.

Main Methods

  • Trained a deep learning algorithm on 3,973 pixel-wise annotations from 295 whole slide images (WSIs) of LM.
  • Validated and tested the AI model on 58 WSIs, comparing its performance to 3 dermatopathologists and 2 pathology residents.
  • Used immunohistochemistry (SOX10) as the reference standard and defined recurrence risk based on melanocyte counts (≤25 vs. >25 per 0.5 mm²).

Main Results

  • The AI model achieved an AUC of 0.84 for discriminating low vs. high recurrence risk margins.
  • Dermatopathologists' AUCs ranged from 0.72 to 0.90, and residents' AUCs ranged from 0.68 to 0.80.
  • AI assistance significantly improved the performance of 2 out of 5 participating pathologists.

Conclusions

  • Deep learning demonstrates significant accuracy in detecting LM resection margins associated with recurrence risk.
  • The AI tool has the potential to assist pathologists in pre-screening and assessing LM margins.
  • AI integration can enhance diagnostic performance in dermatopathology for challenging cases like LM.

Keywords: