An interactive dynamic analysis and decision support software for MR mammography

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Biomedical Signal Processing

Background:

• Subjectivity in MR mammography (MRM) diagnosis can impact accuracy.
• Existing methods may struggle with normal tissue enhancement and artifacts.
• Automated analysis tools are needed to improve consistency and efficiency.

Purpose of the Study:

• To introduce a fully automated software, DynaMammoAnalyst, for MR mammography.
• To overcome diagnostic subjectiveness using normalized maximum intensity-time ratio (nMITR) maps.
• To improve the accuracy and efficiency of MRM examinations.

Main Methods:

• Development of automated software incorporating nMITR maps.
• nMITR maps designed to suppress normal parenchyma and blood vessel enhancement.
• A classifier trained on normalized complexity and maximum nMITR features from 22 lesions and tested on 22 others.

Main Results:

• The software achieved high diagnostic performance: 92% sensitivity, 90% specificity, 91% accuracy, 92% positive predictive value, and 90% negative predictive value.
• nMITR maps demonstrated tolerance to field inhomogeneities and motion artifacts.
• The automated system significantly reduced evaluation time and inter/intra-observer variability.

Conclusions:

• DynaMammoAnalyst offers an objective and automated approach to MR mammography interpretation.
• The software enhances diagnostic decision support, improving accuracy and consistency.
• This automated tool has the potential to streamline MRM workflows and improve patient care.