An efficient CNN based algorithm for detecting melanoma cancer regions in H&E-stained images

Area of Science:

• Digital pathology
• Computational oncology
• Artificial intelligence in medicine

Background:

• Histopathological images are crucial for diagnosing diseases like skin cancer.
• Large digital histopathology image sizes pose challenges for automated analysis.
• Accurate segmentation of abnormal cell nuclei and their distribution is needed for efficient diagnostics.

Purpose of the Study:

• To develop a deep learning-based technique for segmenting melanoma regions in histopathology images.
• To enable rapid and comprehensive diagnostic assessment through automated analysis.
• To improve the accuracy and efficiency of melanoma detection in digital pathology.

Main Methods:

• A deep learning neural network was employed for initial segmentation of cell nuclei.
• Segmented nuclei were utilized to generate precise melanoma region masks.
• The technique was applied to Hematoxylin and Eosin-stained histopathological images.

Main Results:

• The proposed method achieved approximately 90% accuracy in nuclei segmentation.
• Melanoma region segmentation accuracy reached around 98%.
• The technique demonstrated low computational complexity, enabling efficient processing.

Conclusions:

• The developed deep learning technique effectively segments melanoma regions in histopathology images.
• High accuracy in nuclei and melanoma region segmentation facilitates diagnostic assessment.
• The method offers a computationally efficient solution for digital pathology applications.