🌸 mammoinsight-math

Deterministic Biomedical Image Processing & Clinical Feature Extraction for Mammography The Transparent Box Approach to Biomedical Image Analysis

📌 Overview

mammoinsight-math is a specialized, open-source Python library providing deterministic, non-black-box algorithms for mammographic image analysis. Developed as part of the MammoInsight Project, it provides robust algorithms for frequency-domain enhancement, BI-RADS lesion morphometry, radiomic texture extraction (Haralick GLCM), and classical segmentation.

• 🌐 Official Website: https://www.mammoinsight.org/
• 📦 Source Code Repository: https://github.com/mammoinsight/mammoinsight-math

⚡ Quickstart & Installation

Install the package via pip:

pip install mammoinsight-math

Quick Usage Example

import cv2
import mammo_math as mm

# Load ROI image and lesion mask
image = cv2.imread("mammo_roi.png", cv2.IMREAD_GRAYSCALE)
mask = cv2.imread("lesion_mask.png", cv2.IMREAD_GRAYSCALE)

# 1. Image Enhancement (Homomorphic Filter in Fourier Domain)
enhanced_img = mm.homomorphic_filter(image, cutoff=15)

# 2. Deterministic Morphometry (BI-RADS Lesion Geometry)
geom = mm.LesionGeometry(mask)
print(f"Circularidad:  {geom.calcular_circularidad():.4f}")
print(f"Diámetro Feret: {geom.calcular_diametro_feret():.2f} px")

# 3. Radiomic Texture Extraction (Haralick GLCM Descriptors)
texture_vector = mm.extract_glcm_features(enhanced_img)
print("Firma GLCM (Contraste, Disimilitud, Homogeneidad, Energía, Correlación, ASM):")
print(texture_vector)

🧪 Included Core Modules

1. Preprocessing (mammo_math.preprocessing):

   • homomorphic_filter: Frequency-domain illumination correction via 2D FFT.
   • apply_clahe: Contrast Limited Adaptive Histogram Equalization.
   • apply_unsharp_masking: Edge enhancement filter.
   • segment_with_frangi: Frangi vesselness filter for vascular structure suppression.
   • segment_with_gabor_kmeans: Multi-scale Gabor filter bank clustering.

2. Lesion Geometry & Morphometry (mammo_math.core):

   • LesionGeometry: Deterministic calculations of Circularity, Feret Diameter, Eccentricity, and Elongation for BI-RADS mass shape characterization.
   • novoas_core: Boundary roughness and shape indices.

3. Radiomic Texture (mammo_math.features):

   • extract_glcm_features: Gray-Level Co-occurrence Matrix (Haralick 6-descriptor texture vector).
   • extract_feature_vector: Combined 6 and 12-dimensional feature extractors.

4. Segmentation & Clustering (mammo_math.segmentation):

   • novoas_segmentation: Contour-based lesion segmentation.
   • apply_gdc_filter: Grid-Density Clustering for noise reduction.
   • QuantileSegmenter: Dynamic quantile thresholding.
   • crecimiento_de_regiones & segmentacion_otsu_local: Local Otsu & Region Growing.

📚 Academic References & Citations

1. Haralick, R. M., Shanmugam, K., & Dinstein, I. (1973). Textural features for image classification. IEEE Transactions on Systems, Man, and Cybernetics, SMC-3(6), 610–621. DOI: 10.1109/TSMC.1973.4309314 | IEEE Xplore
2. Suckling, J., et al. (1994). The Mammographic Image Analysis Society Digital Mammogram Database (MIAS). Excerpta Medica. International Congress Series, 1069, 375–378. Cambridge Repository | PMC 3092049

👤 Author & License

• Author: Ernesto Rafael Perez — Project Lead & Researcher (mammoinsight@gmail.com)
• Website: https://www.mammoinsight.org/
• License: MIT License. Free for academic, clinical, and open-source research use.