Diagnostic     Screening     CAD methods    

New CAD Methods for the intelligent MammoWorkstation

Image Diagnost’s intelligent MammoWorkstation features an optimal presentation of the diagnostically relevant information to the radiologist. With its special tools it increases diagnostic quality and cuts down the time necessary to read the image. Based on its proprietary CAD algorithms Image Diagnost proudly presents advanced functions for mammogram reading:

Scale & Align: Automatic scaling and aligning

When opening a study at the workstation, you expect an optimal presentation of the image. To that end Scale & Align vertically aligns the images in pairs to achieve symmetry. Additionally, the images are scaled such that the entire breast is displayed while the black area is minimised, making the most out of precious screen real estate. All images of a study are scaled using the same scaling factor so that sizes are comparable between images. Scale & Align obviates the need to correct image positions manually and is a real time saver. A patent is pending for this technique.

Image presentation without Scale & Align

Image presentation with Scale & Align

[ go top ]

Backlight: Global contrast enhancement

Backlight enhances the contrast of an image in such a way that all areas of the breast are clearly visible. The reduced compression towards the skin border is compensated so that the skin line is crisply visible. In the denser gland tissue up to the pectoral muscle the local dynamic range is amplified so that this area no longer appears over-exposed. By its local contrast optimization Backlight increases the visibility of lesions. Instead of manually adjusting brightness and contrast region by region you get an all-encompassing impression at a single glance. Following the recommendations of PAS 1054, it is the goal of Image Diagnost to provide a most homogeneous, comparable image impression based on the raw data of each particular modality. Especially in inhomogeneous screening networks this normalized presentation of images greatly increases the comfort in reading mammograms. Backlight is work in progress.

Left halves: The vendor proprietary post-processing renders dense tissue with too little contrast and overemphasizes the pectoral muscle.

Right halves: Backlight renders the parenchyma as well as the pectoral muscle with highly contrasting detail. The skin line is presented crisp and clear

[ go top ]

Spotlight: Local contrast enhancement

Much like a spotlight Spotlight enhances the contrast in a circular area. And just like a magnifying glass it can be moved all over the image of the breast. Spotlight computes the optimal contrast for the respective area dynamically. The optimized area is blended into the overall image without causing artifacts. The dynamic can be finely adjusted by varying the size of the area. Most notably, Spotlight enhances the visibility of lesions. The locally optimized contrast contributes to the reliable evaluation of suspect areas, especially in dense tissue. It obviates time consuming manual windowing.

Left picture: Image without spotlight
Right picture: Image with Spotlight applied to a region within the parenchyma

[ go top ]

CrossLink: cross-reference between projections

CrossLink serves as aid in localizing a particular area in the respective other projection. It relates the ray paths MLO and CC of the same breast. For an area in one projection CrossLink shows the region of possible correspondence in the other projection. Subsidiary lines indicate an area that corresponds with a probability of 90%.

CrossLink indicates a region of a corresponding location in the respective other projection

[ go top ]

CalcEnhance: Optimized rendering of micro calcifications

CalcEnhance renders micro calcifications with notably higher contrast so that they stand out even in an overview display. CalcEnhance causes no artifacts since it does not paint over micro calcification clusters but renders calcifications in their original size and form with amplified contrast. CalcEnhance greatly speeds up the perception of micro calcifications. Other than with classic CAD the eye is not distracted by markers. The advantages of this technique have been clinically evaluated. CalcEnhance is part of the Image Diagnost CAD-Server. The technique is patented.

Top row: excerpts from mammograms containing micro calcification clusters
Bottom row: the same excerpts with CalcEnhance activated

[ go top ]

CAD: Computer assisted detection

The CAD system of Image Diagnost draws a contour around a micro calcification cluster and marks masses. The sensitivity for micro calcification clusters is 98%, for masses it is above 80%. A clinical study found that in a clinical database with 180 lesions 100% of the micro calcification clusters with BIRAD 3-5 ratings were marked correctly. Just in soft-copy reading there is a significant risk to overlook small calcifications, since reliable detection requires a costly scrutiny in 1:1 pixel magnification. The high degree of sensitivity for micro calcifications greatly increases certainty. The Image Diagnost CAD system can be used in connection with mammography devices from different vendors and it can serve in a location independent way for a number of MammoWorkstations.

Image Diagnost’s CAD draws a contour around micro calcification clusters and marks masses

[ go top ]

Conclusion

Image Diagnost’s MammoWorkstation and CAD server are an investment in a promising mammography technology. The increased efficiency it offers pays dividends day by day. During the recent years Image Diagnost has shaped the marked for reading systems for digital mammography. A number of patents covering CAD methods exemplify the innovative power of the company. Its strengths lie in mastering the technology of digital image processing and image analysis and in its high degree of competence in IT as well as in its flexibility.