Project Funding Details

Breast Cancer kills 3,087 Australian women per year, and whilst mammography has dramatically improved survival rates, a high error rate (30-67%) in disease detection remains. The basis of detection error appears to be a combination of human and technological limitations, however this is not well understood. We urgently need to address this gap in knowledge so that reliable and accurate screening strategies can be developed using more effective technology and education. ‍ When a woman is diagnosed with breast cancer at screening and images from her previous screens are examined, one of two things is evident: ‍ •Visible signs of a cancer that was not reported (i.e. missed); •No visible signs of a cancer. ‍ The reasons for these “missed” cancers are unclear but need careful evaluation to understand why they were missed. With respect to the mammograms with no visible signs, a new study shows that there may be covert information within the image of impending disease: expert radiologists do better than chance at recognising women (with no visible signs of cancer) who will develop cancer within 3 years from those who will not. This finding is linked to a ‘gist’ signal and is supported by evidence that women recalled for additional examination but whose images are eventually considered to be normal have a 1.4–6.6 fold increased risk of developing cancer within 3–10 years. ‍ This work will therefore: ‍ •identify the basis of “missed” cancers so that technological and educational interventions can be transformed; ‍ •establish whether previous mammograms contain information that can predict future malignant events. ‍ We will combine expert radiologists’ assessment and innovative image analysis algorithms to exploit information contained within mammograms. The work will transform breast cancer screening by informing technological and educational interventions to diagnose “missed” cancers and identify women of higher than average risk of future malignancy.

An estimated 17,586 females and 144 males will develop breast cancer in Australia in 2017, a 13.2% increase from 2016. Of these, only 6.5% deaths are estimated to result from breast cancer when early detection and effective treatment strategies are in place. However 5-50% of cancers are missed on screening, with 30-67% of the missed cancers that were identified at subsequent screens, being visible in previous screens. Also, there is another subset of women with cancers whose previous screens had no visible signs of cancer, yet a radiologist’s instinct based on the ‘gist’ of the image will suggest that there is still something odd about the image. This gist phenomenon has never been explored even though it could yield major benefits when describing a woman’s risk profile. Therefore, the objectives of the proposed work are to: • study previous mammograms with “visible signs of cancer”, which were reported as normal, to identify factors that contribute to missed cancers so that technological and educational interventions can be tailored to ensure early diagnosis of these cancers; • establish whether previous mammograms with “no visible sign of cancer” contain information that can predict a woman’s risk of developing cancer in the future so that these can be characterized and used to tailor screening strategies. This work will result in tailored radiologist training and computer-aided diagnosis algorithms to better recognise “missed” cancers, identify women at a higher risk of a future malignancy and facilitate early detection and treatment of the disease.