Hello BRIGHT Run Family,
I hope you are enjoying the gradual transition to Spring from Winter.
I am enjoying the weather and feeling happy after submitting some of my finished research projects for peer-review.
Each project that I finish takes a part of me, but not forever. It is returned after being reformatted with some invaluable knowledge and experience which I use in subsequent projects. The knowledge and experience are derived from reading, critical thinking, experimenting, writing, and of course, collaborating. I enjoy this journey but just like any other long journey that needs patience, it is a bag of mixed emotions.
Now, I am going to tell you about a very recent publication that I co-authored (A Competition, Benchmark, Code, and Data for Using Artificial Intelligence to Detect Lesions in Digital Breast Tomosynthesis | Breast Cancer | JAMA Network Open | JAMA Network). The publication involves research work that initiated while I was training at Duke University. Often it takes a few years from conception/inception to publication and the same happened here as well.
The work is related to using AI for digital breast tomosynthesis (DBT). DBT is often called 3D mammography as, instead of producing one image/slice (from one breast and view) in mammography (which is 2D), it produces a series of images.
To elaborate, DBT acquires multiple projections across an arc and reconstructs images of multiple ‘sections’ and produces a stack of slices. This helps reduce the effect of overlapping tissue in mammography and improves localization of a lesion. However, from the perspective of a radiologist/breast imager who interprets the DBT, the time needed for interpretation increases. Therefore, AI’s assistance in interpretation can help reduce the time taken. Also, it could be useful when there is a shortage of trained personnel for interpretation.
The goal of our work was to provide a foundation for future researchers interested in performing AI-based (computer-assisted) diagnosis methods for DBT. A grand-challenge was conducted to develop and compare AI-techniques that can detect lesions from DBT.
Teams around the world took part in that and came up with innovative ideas to improve the performance of AI-based lesion detection algorithms. By publicly sharing the data, related computer programs (codes) used in this grand challenge, and a performance benchmark, this work tries to help researchers understand this area of research quickly and apply the recent advancements in AI to build better detection algorithms.
I am happy to be part of the team that worked on this project. I am also thankful to the patients who contributed to this invaluable dataset and are helping to improve research for posterity. The knowledge and experience from this work are helping me with my ongoing and future projects.
Stay well and have a nice Spring!
Best,
Ashirbani
Dr. Ashirbani Saha is the first holder of the BRIGHT Run Breast Cancer Learning Health System Chair, a permanent research position established by the BRIGHT Run in partnership with McMaster University.
