&lt;title&gt;Classification of false-positive findings on computer-aided detection of breast microcalcifications&lt;/title&gt;

Freedman, Matthew T.; Lo, Shih-Chung Benedict; Artz, Dorothy Steller; Lau, Ivan; Mun, Seong Ki

doi:10.1117/12.274173

Cited by 1 publication

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A separate evaluation of the detection of artifacts and other structures is extremely valuable in algorithm optimization and further development, as shown by the results in Tables II and III and also suggested by Freedman et al 10 In the cases with masses, for example, high intensity objects such as BBs and large single calcifications were identified by the algorithm as ''masses,'' but three of the lymph nodes were missed. One could perhaps proceed development by adding an intensity threshold to eliminate the high intensity objects from the final output.…”

Section: Discussionmentioning

confidence: 76%

“…Films may often contain artifacts due to the grid, screen, film defects, processor, deodorant, etc. 10 The digitized images may further contain artifacts generated from the digitizer. Quality control tests and procedures are designed to eliminate as many of these artifacts as possible.…”

Section: Artifactsmentioning

confidence: 99%

“…Such detections are usually included in the FP rate. 10 But artifacts are true although artificial signals and including them in the FP or TP estimates of the objects of interest provides no useful information to the investigator or the evaluation of the methodology. They need to be considered separately leading to the definition of separate measures, e.g., TPA and FNA for true positive artifact and false negative artifact rates in the case they are correctly identified ͑TPA͒ or missed to be FIG.…”

Section: Artifactsmentioning

confidence: 99%

“…Some investigators report them as part of the FP rate. 10 Consider, however, that these are true mammographic elements and one can make a strong argument for including these detections in the TP rate instead. But should these detections enter the TP or FP performance estimates at all?…”

Section: Other Mammographic Structuresmentioning

confidence: 99%

“…This implies that the problem of direct comparison of similar algorithms cannot be resolved by simply using the same set of images ͑common database͒, a concept that has gained popularity in the last 5 years. 8 Examples are found in Nishikawa et al, 9 where two ways of clustering signals yielded different outcomes and in Freedman et al, 10 where the first detailed analysis is presented of false positive detections related to calcifications and their effect on performance. Hence, different performance indices were obtained by interpreting the results differently.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Evaluating the performance of detection algorithms in digital mammography

1999

View full text Add to dashboard Cite

The initial and relative evaluation of computer methodologies developed for assisting diagnosis in mammography is usually done by comparing the computer output to ground truth data provided by experts and/or biopsy. Reported studies, however, give little information on how the performance indices of computer assisted diagnosis (CAD) algorithms are determined in this initial stage of evaluation. Several strategies exist in the estimation of the true positive (TP) and false positive (FP) rates with respect to ground truth. Adopting one strategy over another yields different performance rates that can be over- or underestimates of the true performance. Furthermore, the estimation of pairs of TP and FP rates gives a partial picture of the performance of an algorithm. It is shown in this work that new performance indices are needed to fully describe the degree of detection (part or whole) and the type of detection (single calcification, cluster of calcifications, mass, or artifact). Several evaluation strategies were tested. The one that yielded the most realistic performances included the following criteria: The detected area should be at least 50% of the true area and no more than four times the true area in order to be considered TP. At least three true calcifications should be detected to within 1 cm2 with nearest neighbor distances of less than square root(2) cm for a cluster to be considered TP. Separate detection measures should be established and used for artifacts and naturally occurring structures to maximize the benefits of the evaluation. Finally, it is critical that CAD investigators provide information on the tested image set as well as the criteria used for the evaluation of the algorithms to allow comparisons and better understanding of their methodologies.

show abstract

Section: Discussionmentioning

confidence: 76%

Section: Artifactsmentioning

confidence: 99%

Section: Artifactsmentioning

confidence: 99%

Section: Other Mammographic Structuresmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations