Monday, 10 December 2007

RSNA Breast Imaging - Computer Aided Detection

  1. CAD of malignant and benign clustered microcalcifications in temporal pairs of mammograms and it's effect on Radiologists characterization performance (Ann Arbor, MI) - increased diagnostic accuracy when using interval change analysis
  2. CAD review of false -ve FFDM exams on priors (Korea) - CAD correctly identified 34 0f 46 (73%) of retrospectively visible lesions and 14 of 14 (100%) of actionable lesions
  3. Critical analysis of studies measuring the effectiveness of CAD in screening (Chicago, IL) - while each individual study is small, the aggregate of studies indicate that performance is comparable to double reading by radiologists
  4. An inductive CAD system using genetic algorithms for detection of calcification on mammograms (Orinda, CA) - CAD teaching itself and learning from it's mistakes
  5. US CAD into clinical practice: and observational study (Chicago, IL) - cautious optimism using CADx
  6. Independent US CAD BI RADS assessment of breast lesions (Philadelphia, PA) - Initial study showed significant improvement in the hands of trained users
  7. MRI detected suspicious findings: comparison of kinetic feature measurement using CAD in benign and malignant lesions (Seattle, WA) - ONLY delayed kinetics categorized by the most suspicious type of enhancement were significantly different between B/M. In keeping with MRI BI RADS recommendation of reporting the most suspicious kinetic curve. Any washout enhancement was assoc with malignancy in nearly 50%, entirely persistent with 13.3%
  8. Application of CAD in MRM: do we really need whole lesion time curve distribution analysis? (Jena, Germany) - Higher diagnostic accuracy for curve distribution data after application of a statistical regression model compared to semi-automatic detected 'most suspect curve' type