Multiplexed wavelet transform technique for detection of microcalcification in digitized mammograms.

Area of Science:

• Medical Imaging
• Biomedical Signal Processing
• Computer-Aided Diagnosis

Background:

• Microcalcifications are early indicators of breast cancer.
• Accurate detection of microcalcifications in mammograms is crucial for early diagnosis.
• Wavelet Transform (WT) offers potential for microcalcification detection.

Purpose of the Study:

• To implement and evaluate two novel WT-based schemes for automatic detection of clustered microcalcifications.
• To assess the performance of Marr-Hildreth (M-H) and Canny detectors within the WT framework.
• To analyze the trade-offs between detection sensitivity and false positives.

Main Methods:

• Utilized a one-dimensional WT technique exploiting image sequence pseudo-periodicity.
• Employed parent-child relationships of WT coefficients at different decomposition levels for detection.
• Applied Marr-Hildreth (zero-crossings) and Canny (local extrema) detectors.
• Weighted detected pixels before inverse transform and segmented using global gray level thresholding.

Main Results:

• Both WT-based schemes achieved 95% detection sensitivity for microcalcifications.
• The M-H detector produced more false positives but preserved shape information.
• The M-H detector demonstrated better sensitivity for widely distributed calcifications.
• Algorithms exhibited high detection efficiency and low memory requirements.

Conclusions:

• WT-based schemes are effective for automatic microcalcification detection in mammograms.
• The M-H detector offers advantages in shape preservation and sensitivity for diffuse calcifications.
• These methods contribute to computer-aided diagnosis systems for breast cancer screening.