IBM New AI Predict Breast Cancer A Year Before It Appears

Healthcare is the major sector that is getting revolutionize with the power of AI, and this week IBM shows the world an another example of it. IBM’s researchers have come up with the solution that achieves an impressive level of accurate detection of breast cancer.

Breast Cancer is the global leading cause of cancer-related deaths in women, alone in the US the average risk of a woman developing breast cancer in her life is about 12%, i.e 1 in 8 women.



IBM’s researchers published research in Radiology, that can seed-up the early detection of Breast Cancer. The team builds an AI capable of predicting the development of malignant breast cancer in patients within a year.

The AI model uses both machine learning and deep learning to learn and make decisions from both imaging data and from comprehensive patient’s health history.




The model analyses mammograms and patients clinical information in electronic health records (EHR) to accurately diagnose breast cancer even in the cases where the disease is missed by the radiologists.

This research is based on a study conducted by scientists at IBM’s Zurich office and the University of Zurich, which architected a system that can detect and classify tumor and immune cells as well as their relationships.

To train the AI, the researchers collected a dataset of 52,936 images from 13,234 women who underwent at least one mammogram between 2013 and 2017, and who had health records for at least one year prior to the mammogram.



The AI mode was get trained over more than 9,000 matching sets of mammograms and health records.

It was trained to analyze and predict biopsy malignancy and to differentiate a normal and abnormal examination and to mapped connections among clinical risk factors to anticipate biopsy malignancy and differentiate normal from abnormal screening examinations.

During the training process, the team majorly used the two standard views in mammography that are craniocaudal (CC) and mediolateral oblique (MLO) which are often compared in assessing lesions.

The researchers started, by identifying a subset of the clinical features with the highest contribution to positive biopsy prediction.

These were fused into a deep neural network (DNN) that was trained on each mammogram for each prediction task. With the use of DNN., the team extracted the probability of each prediction task, as well as imaging features, for each view.

For the probability of either cancer-positive biopsy or normal/abnormal, differentiation was estimated using a gradient boosting machines model.

Lastly, the imaging features they concatenated the imaging features as well as the entire set of clinical features into a single representation of patients’ breasts.




If we talk about the results, then the AI algorithm was validated with data from 1055 women and was tested in 2548 women (mean age, 55 years ± 10 [standard deviation]).

In the test set, the algorithm identified 48% of false-negative findings on mammograms.

It was found the model obtained an area under the receiver operating curve (AUROC) of 0.78, improving breast cancer risk prediction in comparison to common risk models like the Gail model.

Moreover, the model generates a specificity of 77.3 percent at a sensitivity. The AI was also successful to elevated possible risk which was not previously used by other models, such as white blood cell profiles and thyroid function tests.

This factors can significantly reduce the chance of a bad diagnosis by establishing connections between traits you wouldn’t spot in imagery alone, such as iron deficiencies and thyroid function.

It was found the model obtained an area under the receiver operating curve (AUROC) of 0.91, and specificity of 77.3 percent at a sensitivity of percent. The addition of clinical data to the mammograms has significantly increased the model’s AUROC and sensitivity.

In my opinion, the technology of this kind is very much needed. This technology can be potential to reduce the number of women being sent for unnecessary tests or experiencing the trauma of being needlessly assigned as high risk by traditional models.

More in AI

Adobe’s New AI Can Automatically Detect Deepfakes and Photoshopped Images

Google New Project Combines Satellites With AI To Fight Against Pollution

Olympic Of Artificial Intelligence Is All Set To Start In June 2019



Leave a Reply

Your email address will not be published. Required fields are marked *