Automated proliferation index calculation for skin melanoma biopsy images using machine learning

Area of Science:

• Oncology
• Computational Pathology
• Medical Imaging

Background:

• The Proliferation Index (PI) is crucial for cancer evaluation.
• Accurate PI measurement is vital for diagnosis, prediction, and prognosis.
• Automated methods can improve efficiency and consistency in PI assessment.

Purpose of the Study:

• To develop an automated machine learning technique for measuring the Proliferation Index (PI) in skin melanoma.
• To accurately segment and classify tumor cell nuclei in histology images.
• To reduce the error and computational complexity associated with manual PI assessment.

Main Methods:

• Utilized machine learning algorithms, specifically Convolutional Neural Networks (CNNs).
• Developed a technique involving Mart-1 and Ki-67 stained histology images.
• Automated tumor region identification using region of interest (ROI) masks and nuclei segmentation/classification.

Main Results:

• Achieved 94% accuracy in robust nuclei segmentation and classification.
• Demonstrated low computational complexity for the automated technique.
• Calculated PI values exhibited an average error of less than 4%.

Conclusions:

• The proposed automated technique accurately measures the Proliferation Index in skin melanoma.
• This method offers a robust, efficient, and low-error alternative to manual PI assessment.
• The findings support the use of AI in digital pathology for cancer diagnostics.