POSTER SECTIONS Coverpage Purpose Methods and materials Results Conclusion Personal information and conflict of interest References

ECR 2020 / C-07603

How does a prototype synthetic mammography version compare to current synthetic mammography and to full field digital mammography?

Congress:

ECR 2020

Poster Number:

C-07603

Type:

Scientific Exhibit

Keywords:

Performed at one institution, Not applicable, Retrospective, Tissue characterisation, Quality assurance, Cancer, Physics, Observer performance, Comparative studies, Mammography, Image manipulation / Reconstruction, Digital radiography, Radiation physics, Oncology, Breast

Authors:

C. Van Ongeval¹, L. Cockmartin², S. Postema², R. Prevos², M. Keupers², T. Thywissen², G. Zhang², H. Bosmans²; ¹Leuven/BE , ²Leuven/BE

DOI:

10.26044/ecr2020/C-07603

DOI-Link:

https://dx.doi.org/10.26044/ecr2020/C-07603

Fig. 5: Example of a aide-by-side evaluation during VGC analysis. Magnified part of an...

Fig. 4: Significant better representation of opacities in InsightV2 compared to...

Fig. 6: Reader 1 scores the representation of microcalcifications better in InsightV2...

Fig. 7: Significant better image contrast in InsightV2 compared to InsightV1 for both...

Fig. 8: Image Sharpness: Slight improvement in image sharpness in InsightV2 compared...

Fig. 9: All IQ criteria InsightV2 scores significantly better than InsightV1 both in...

Fig. 10: AUC results of reader 1 and reader 2 for InsightV1, InsightV2 and FFDM

Results

1. Visual grading characteristics

Significant better representation of opacities in InsightV2

AUC_R1 = 0.70 (0.65-0.76)

AUC_R2 = 0.60 (0.54-0.67)

Fig. 4: Significant better representation of opacities in InsightV2 compared to InsightV1 for both readers

Better representation of microcalcifications in InsightV1 for R1 but similar to FFDM for R2

AUC_R1 = 0.66 (0.61-0.72)

AUC_R2 = 0.52 (0.46-0.58)

Fig. 6: Reader 1 scores the representation of microcalcifications better in InsightV2 compared to InsightV1 but this difference is lower than for the opacities. Reader 2 finds no difference between microcalcification representation for InsightV1 and InsightV2

Significant better image contrast in InsightV2 compared to InsightV1 for both readers.

AUC_R1 = 0.89 (0.86-0.93)

AUC_R2 = 0.94 (0.91-0.96)

Fig. 7: Significant better image contrast in InsightV2 compared to InsightV1 for both readers

No significant difference in image sharpness.

AUC_R1= 0.60 (0.54-0.65)

AUC_R2 = 0.69 (0.63-0.74)

Fig. 8: Image Sharpness: Slight improvement in image sharpness in InsightV2 compared to InsightV1 for reader 1, reader 2 scores sharpness in InsightV2 significantly better compared to InsightV1

Significant lower noise level.

AUC_R1 = 0.88 (0.84-0.91)

AUC_R2 = 0.78 (0.73-0.82)

For all image quality criteria: InsightV2 scores significantly better compared to InsightV1, both in comparison with FFDM.

AUC_R1 = 0.77 (0.75-0.79)

AUC_R2 = 0.73 (0.70-0.75)

Fig. 9: All IQ criteria InsightV2 scores significantly better than InsightV1 both in comparison with FFDM

2. Accuracy - ROC result

R1 AUC_InsightV1 = 0.74

R1 AUC_InsightV2 = 0.78

R1 AUC_FFDM = 0.83

R2 AUC_InsightV1 = 0.71

R2 AUC_InsightV2 = 0.69

R2 AUC_FFDM = 0.77

Fig. 10: AUC results of reader 1 and reader 2 for InsightV1, InsightV2 and FFDM

3. Breast Density

InsightV1 versus FFDM

R1: 70% agreement, kappa = 0.53
R2: 68% agreement, kappa = 0.55

InsightV2 versus FFDM

R1: 89% agreement, kappa = 0.82
R2: 79% agreement, kappa = 0.70

4. FFDM likeliness score (/10)

Median score for InsightV1:3

Median score for InsightV2: 8

Median score for FFDM: 10