Are You MAD (Mammographic Architectural Distortion)? Introducing Subtypes of Synthetic Mammography/Tomosynthesis-Detected Architectural Distortion: Radiating Distortion and Nonradiating Tissue Alteration.

OBJECTIVE

The goals of this study are to introduce subtypes of mammographic architectural distortion (MAD), designated as radiating distortion (RD) and nonradiating tissue alteration (TA), and evaluate the clinical outcomes, interobserver agreement, and predictive capability of these newly defined subtypes.

METHODS

Radiating distortion, TA, and 7 mammographic features of MAD were prospectively defined. Three breast radiologists reviewed consecutive screening synthetic mammography/tomosynthesis examinations recalled for MAD between January 1, 2021, and December 31, 2021. Reviewers reclassified MAD as RD, TA, or Neither (RD nor TA) and rated mammographic features. Majority classification was used for analyses. Clinical outcomes were established with ≥2 year follow-up and compared using Fisher's exact test. Logistic regression was performed for feature prediction of MAD subtypes and malignancy. Interobserver agreement was assessed using Gwet's agreement coefficient (AC).

RESULTS

Eighty-nine eligible cases were majority classified as RD in 22.5% (20/89), TA in 48.3% (43/89), and Neither in 29.2% (26/89). Malignancy was found in 45% (9/20) of RD, 11.6% (5/43) of TA, and 0% (0/26) of Neither. There were significant differences in diagnostic imaging findings, assessment, and recommendation between subtypes. In all, 80.8% (21/26) of Neither designations represented superimposition. There was substantial interobserver agreement for MAD subtype, Gwet's AC = 0.65. The following features were significantly predictive of RD: ≥7 dense lines, straight-line type, and point convergence. Radiating distortion was predictive of malignancy, and nondense center was associated with benignity in this limited sample.

CONCLUSION

Synthetic mammography/tomosynthesis screen-detected MAD classified as RD and TA demonstrated substantial interobserver agreement and significantly different clinical outcomes.