![]() | Only 14 pages are availabe for public view |
Abstract Background: Women with dense breasts are doubly disadvantaged as they are both at higher risk of developing breast cancer and at greater risk that cancer will not be detected because of masking of the radiological signs of cancer by increased density leading to high recall rate and false positive mammography findings, thus decreasing its specificity.Supplementary screening with ultrasound has been shown to reduce interval cancer rate, but at the expense of false positives and increased cost. Our aim was to study the role of AI in assessment of mammographically dense breasts. Results: This study included 110 patients; 148 breast lesions. All the patients were subjected to full field digital mammography, breast ultrasound and processing of the mammographic images by AI software.Combined mammography and breast ultrasound had had a sensitivity of 98.73%(95% CI: 93.15% to 99.97%), a specificity of 71.01%(95%CI: 58.84% to 81.31%) , a positive likelihood ratio of 3.41(95%CI: 2.35 to 4.93), a negative likelihood ratio of 0.02(95%CI: 0.00-0.13), a positive predictive value of 79.59%(95%CI: 72.92% to 84.96%), a negative predictive value of 98%( 95%CI: 87.42% to 99.71%) and a diagnostic accuracy 85.91%(95% CI: 79.13% to 91.00%). AI had a sensitivity of 86.08%(95%CI: 76.42%- 92.84%), a specificity of 91.3%(95%CI:82.03%- 96.74%) , a positive likelihood ratio of 9.9(95%CI: 4.58-21.37), a negative likelihood ratio of 0.15(95%CI: 0.09-0.27), a positive predictive value of 91.89%(84%-96.07%), a negative predictive value of 85.14%(95%CI: 76.71% - 90.87%) and a diagnostic accuracy 88.51%(95%CI: 82.25%-93.16%). Conclusion: The performance of artificial intelligence is comparable to sono-mammographic diagnosis and has the potential to substantially reduce missed diagnosis.AI can provide objective and effective information to radiologists and reduce the workload and rates of missed diagnosis and misdiagnosis |