AI-Driven Collaboration Revolutionizes Breast Cancer Screening and Diagnosis

Credits By: Expat Guide Turkey

According to current research, radiologists exhibit improved breast cancer detection abilities with the aid of artificial intelligence (AI) compared to independent performance. Additionally, this cutting-edge AI technology demonstrates improved accuracy when collaborating with radiologists, outperforming its performance independently.

The study, carried out on a sizable scale and published this month in The Lancet Digital Health, is the first to compare the effectiveness of AI in solo breast cancer screening directly with that of AI as a supporting tool for human experts. Such AI systems could have a significant impact, promising to find tumors that escape clinicians’ notice, maximizing radiologists’ time for more patients, and relieving pressure in areas struggling with a shortage of specialists.

The study’s driving force, the German firm Vara, is the source of the software being evaluated. Vara’s AI is already in use in more than a fourth of Germany’s breast cancer screening facilities. It was also made available to medical facilities in Greece and Mexico recently.

The Vara team investigated two operating techniques in association with radiologists from the Memorial Sloan Kettering Cancer Centre in New York and Germany’s Essen University Hospital. In the first instance, the AI analyzed mammograms on its own. In the second instance, the AI distinguished between scans that raised red flags and those it deemed normal. Before the AI’s evaluation, suspicious scans were sent to a radiologist for inspection, enabling a combined diagnosis. If the AI discovered cancer and the doctor missed it, it sent alarms. Previous scans were reviewed in the study, and AI ratings were compared to the radiologists’ initial evaluations.

Compared to a radiologist operating alone, the collaborative strategy between AI and radiologists was found to be 2.6% more effective at detecting breast cancer. Additionally, this technique produced fewer false alarms. This method simplified radiologists’ work by automatically classifying images as “confident normal,” which accounted for 63% of all mammograms.

Patients who undergo post-screening treatments with average scan results are released, whereas abnormal or unclear scans call for extra testing. However, one in eight tumors is missed by radiologists during mammography analysis because of variables like weariness and the sheer amount of scans. Subtle visual cues and dense breast tissue further complicate accurate identification.

In Germany, radiologists are legally required to examine every mammogram, even ones that AI has deemed normal. However, AI assists radiologists by producing the first reports on routine scans. However, these can be disregarded.

Vara’s AI occasionally underperformed a single physician, even though AI normally excels at image categorization. This is partly explained by the fact that mammography cannot accurately diagnose cancer independently; tissue testing is necessary. Due to the visual resemblance between healthy and malignant breasts and the comparatively low occurrence of cancer in screenings, the AI that examines mammograms for suggestive signals encounters difficulties.

Although the study only considered historical mammography decisions and presupposed radiologists would concur with the AI’s choices, it nevertheless demonstrated how AI could revolutionize the industry. This AI-assisted strategy has the potential to address the global lack of radiologists.

Contact us for an Advertisement