Premium
SU‐FF‐I‐22: An Inward Mammilla Detection Algorithm for Analysis of Skin‐Line Retraction
Author(s) -
Sun Y,
Suri J,
Rangayyan R
Publication year - 2005
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.1997502
Subject(s) - pixel , physics , position (finance) , gray level , line (geometry) , mathematics , nuclear medicine , anatomy , medicine , geometry , optics , finance , economics
Purpose: A specific algorithm is designed for the detection of the mammilla to locate the inward mammilla position along the breast skin‐line in mammograms. The position of the inward mammilla can assist in the analysis of focal retraction near the nipple. Method and Materials: Between the breast skin‐line and the fibro‐glandular tissue is a zone of fatty peripheral tissue, which appears with low gray‐levels on mammograms. Due to the mammary glands connecting to the mammilla, the gray‐level in the fatty zone near the mammilla will be higher. We define a fatty peripheral zone ( Z f ) of 40 pixels width (8mm) parallel to the skin‐line on mammograms. A disk mask of diameter 40 pixels, K P , tangential to the skin‐line boundary point P and rolling in the zone Z f , is used to obtain a mammilla index value for P . A mammilla index ( I P ) for P is defined as the average gray‐level of the pixels in both Z f and the current mask K p . Then, three highest values, I t1 (highest), I t2 (second highest), and I t3 (third highest), corresponding to position indexes P t1 , P t2 , and P t3 on the skin‐line, are found on the curve of I P . If the differences between P t1 and P t2 , as well as between P t1 and P t3 , are larger than a threshold T 1 , and the difference between P t2 and P t3 is less than another threshold T 2 , the mammilla position is defined as the average of P t2 and P t3 ; otherwise the mammilla position is defined as P t1 . Empirically, we selected T 1 =90 and T 2 =36. Results: We have tested our algorithm on 40 mammograms from the MiniMIAS database with inward nipples, and our method achieved accurate detection of the mammilla position on each image. Conclusion: The proposed algorithm for the detection of the inward mammilla position gave accurate results on the mammograms tested.