Impact of breast cancer characteristics on reader performance with concurrent use of artificial intelligence with digital breast tomosynthesis

Congress:

ECR 2019

Poster Number:

C-1648

Type:

Scientific Exhibit

Keywords:

Breast, Computer applications, Mammography, Neural networks, Computer Applications-Detection, diagnosis, Efficacy studies

Authors:

E. F. Conant¹, S. Periaswamy², S. Fotin², J. Go², J. Pike², J. boatsman³, J. Hoffmeister²; ¹Philadelphia, PA/US, ²Nashua, NH/US, ³San Antonio, TX/US

DOI:

10.26044/ecr2019/C-1648

DOI-Link:

https://dx.doi.org/10.26044/ecr2019/C-1648

Table 1: Average radiologist sensitivity based on mammographic appearance and...

Table 2: Average radiologist reading time in cancer cases based on mammographic...

Table 3: Average radiologist specificity based on mammographic appearance

Table 4: Average radiologist reading time in non-cancer cases based on mammographic...

Results

The sensitivity subgroup analysis in Table 1 shows that average radiologist sensitivity improved 6.8% with AI for soft tissue and 6.2% for invasive carcinomas, with larger improvements of 12.0% for calcifications-only and 14.6% for DCIS. Of note, all 95% CIs were above zero, indicating strength of these improvements. For all cancer cases, average radiologist sensitivity improved 8.0% (95% CI: 2.6%, 13.5%) from 77.0% without AI to 85.0% with AI.

Subgroup analysis of reading time in cancer cases (Table 2) shows that average radiologist reading time was reduced 39.5% with AI for soft tissue and 41.5% for invasive carcinomas, with a similar 40.3% reduction for DCIS and a possibly larger 46.8% reduction with calcifications-only. All reductions in reading time had 95% CIs well below zero, indicating strong reductions. For all cancer cases, average radiologist reading time was reduced 41.3% (95% CI: 28.0%, 52.1%) from 61.1 seconds without AI to 35.9 seconds with AI.

The specificity subgroup analysis in Table 3 shows that average radiologist specificity improved 7.9% with AI for soft tissue and 8.4% for BI-RADS 1 or 2, but did not for calcifications-only at -2.7%. The strength of the specificity improvements for soft tissue and BI-RADS 1 or 2 were indicated by 95% CIs that were above zero, but the 95% CIs included zero for calcifications-only, indicating little change in specificity. For all non-cancer cases, average radiologist specificity improved 6.9% (95% CI: 3.0%, 10.8%) from 62.7% without AI to 69.6% with AI.

Subgroup analysis of reading time in non-cancer cases (Table 4) shows that average radiologist reading time was reduced 54.4% with AI for soft tissue with a similar 59.1% reduction for BI-RADS 1 or 2 and a possibly smaller 43.6% reduction with calcifications-only. All reductions in reading time had 95% CIs well below zero, indicating strong reductions. For all non-cancer cases, average radiologist reading time was reduced 55.9% (95% CI: 45.5%, 64.4%) from 65.1 seconds without AI to 28.7 seconds with AI.