POSTER SECTIONS Coverpage Learning objectives Background Findings and procedure details Conclusion Personal information References

ECR 2019 / C-2151

Case examples to demonstrate positive and negative impacts of a deep learning based concurrent artificial intelligence system for digital breast tomosynthesis

Congress:

ECR 2019

Poster Number:

C-2151

Type:

Educational Exhibit

Keywords:

Education and training, Computer Applications-Detection, diagnosis, Neural networks, Mammography, Computer applications, Breast, Artificial Intelligence

Authors:

E. F. Conant¹, A. Y. Toledano², S. Periaswamy³, S. Fotin³, J. Go³, J. Pike³, J. boatsman⁴, J. Hoffmeister³; ¹Philadelphia/US, ²Kensington/US, ³Nashua, NH/US, ⁴San Antonio, TX/US

DOI:

10.26044/ecr2019/C-2151

DOI-Link:

https://dx.doi.org/10.26044/ecr2019/C-2151

Findings and procedure details

Presentation of case examples from MRMC reader study with detection outlines along with calibrated lesion-level and case-level scores to demonstrate performance of radiologists with concurrent use of AI and without AI:

Cases that demonstrate an increase or decrease in detection of malignancies (AI’s effect on sensitivity)

Cases that demonstrate an increase or decrease in detection of non-cancer lesions (AI’s effect on specificity)

Case 1: Increased sensitivity with use of AI -- soft tissue lesion (Figures 5-8)

Case 2: Increased sensitivity with use of AI -- calcifications (Figures 9-13)

Case 3: Decreased sensitivity with use of AI (Figures 14-18)

Case 4: Increased specificity with use of AI (Figures 19-22)

Case 5: Decreased specificity with use of AI (Figures 23-29)

Fig. 5: Case 1 -- DM views

Fig. 6: Case 1 -- Left DBT slices

Fig. 7: Case 1 -- Zoomed left DBT slices and histopathology

Fig. 8: Case 1 -- Increased sensitivity with use of AI -- soft tissue lesion

Fig. 9: Case 2 -- DM views

Fig. 10: Case 2 -- Zoomed left DM views

Fig. 11: Case 2 -- Left DBT slices

Fig. 12: Case 2 -- Zoomed left DBT slices and histopathology

Fig. 13: Case 2 -- Increased sensitivity with use of AI -- calcifications

Fig. 14: Case 3 -- DM views

Fig. 15: Case 3 -- Right DBT slices

Fig. 16: Case 3 -- Zoomed right DBT slices and histopathology

Fig. 17: Case 3 -- Decreased sensitivity with use of AI

Fig. 18: Case 3 -- Mitigating decreased sensitivity with use of AI

Fig. 19: Case 4 -- DM views

Fig. 20: Case 4 -- Left DBT slices

Fig. 21: Case 4 -- Zoomed left DBT slices

Fig. 22: Case 4 -- Increased specificity with use of AI

Fig. 23: Case 5 -- DM views

Fig. 24: Case 5 -- DBT slices and case outcome

Fig. 25: Case 5 -- DBT slices of False Positives (FPs) by readers (both breasts) and by AI (left breast)

Fig. 26: Case 5 -- Reader FPs in zoomed right DBT slices

Fig. 27: Case 5 -- Reader and AI FPs in zoomed left DBT slices

Fig. 28: Case 5 -- Decreased specificity with use of AI

Fig. 29: Case 5 -- Mitigating decreased specificity with use of AI

Home | Browse posters | Help | Privacy policy | Disclamer | Contact | myESR

© 2003-2025 ESR - European Society of Radiology