A Self Organizing Map approach to breast cancer detection

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Mammography is crucial for cancer tumor detection, but faces challenges due to tissue diversity and unclear boundaries.
• Existing methods for enhancing and segmenting mammographic images require further development.

Purpose of the Study:

• To evaluate the performance of the Self Organizing Map (SOM) for detecting cancer-suspicious regions in digitized mammograms.
• To investigate the impact of preprocessing algorithms on SOM-based tumor detection accuracy.

Main Methods:

• Digitized mammograms were preprocessed using specific algorithms.
• The Self Organizing Map (SOM) algorithm was applied to identify potential cancer areas.
• Performance was evaluated based on the detection of suspicious regions.

Main Results:

• The Self Organizing Map (SOM) demonstrated effectiveness in identifying cancer-suspicious regions within mammograms.
• Preprocessing significantly contributed to achieving optimal detection results.
• The study highlights SOM's potential in addressing the complexity of tumor detection.

Conclusions:

• The Self Organizing Map (SOM) is a promising tool for improving cancer detection in mammography.
• Further research into enhancement and segmentation algorithms can refine these AI-driven diagnostic tools.
• This approach aids in overcoming challenges associated with tumor shape and border definition.