AI and Breast Cancer Diagnosis

Blog Author: Archita Khaire

Breast cancer is the most common cancer worldwide. Last year 2.3 million women were diagnosed with breast cancer, and 685,000 deaths were reported. The National Cancer Institute states that more than $500 million is spent yearly on breast cancer research. Even with the significant investment, approximately 43,000 women in the U.S. are expected to die in 2021 from breast cancer.
Based on the human epidermal growth factor receptor 2 (Her2), progesterone receptor (PR), and estrogen receptor (ER), breast cancer is categorized into four molecular subtypes: 

Image Credit: Journal of Cancer

• Luminal A (ER+, PR±, and HER2-): Most prevalent subtype; low recurrence rate
• Luminal B (ER+, PR±, and HER2 ± ): Slightly worse prognosis than Luminal A; more frequent recurrence than A
• HER2-enriched (ER–, PR–, and HER2+): Faster cancer cell growth than luminal subtypes; worse prognosis
• Triple Negative (ER–, PR–, and HER2–): Most aggressive subtype; primarily seen in younger women

From early detection to survivorship, Artificial intelligence is helping doctors and researchers improve the survival rate for patients suffering from #breastcancer.
​
Diagnosis
There are many established risk factors for breast cancer, including gene mutations, estrogen levels, and breast density. A primary example: studies have shown that women with the BRCA1/2 gene mutation have a high risk of breast cancer.  

Image Credit: BMC Medical Genomics volume 14, Article number: 122 (2021)

Several deep learning algorithms are available to detect these mutations to predict break cancer risk. BRCA exchange provides data on thousands of BRCA variants to develop an understanding of a patient's cancer risk. 

Image credit: Nature Communications volume 11, Article number: 5727 (2020)

Researchers use deep learning algorithms to classify estrogen receptors from Breast Cancer Metabolomics Data. Immunohistochemistry (IHC) is often used to determine if cancer cells have estrogen and progesterone receptors. An excellent open-source library (histolab) available on GitHub can be used to get started with quantitative image analysis of stained slides.

Image credit: MIT News

Mammography is considered the first line of defense in breast cancer detection. MIT has developed an image-based deep learning algorithm that can detect cancer in #mammographs up to 5 years in advance. Another important feature of the algorithm is that it is equally effective for various races; this is not the case with many other algorithms that mostly rely on mammographs of white women to train their ML models.
Recently, a thermal imaging-based tool was approved by the European Union, and it uses cloud-based AI analytics to detect breast lesions.  

 Image credit: IEEE Transactions on Medical Imaging ( Volume: 39, Issue: 4, April 2020)

​Breast ultrasound is often used as a follow-up test if a #mammography detects abnormalities. Deep neural networks are available to classify ultrasound images as benign or malignant. 

Dense breasts are one of the most common risk factors for breast cancer development. Approximately 40% of women have dense breasts, indicating a higher risk for breast cancer. However, mammography may miss over 1/3 of cancers in patients with dense breasts. 3D ultrasound and 3D mammography (also known as tomosynthesis) are better alternatives to effectively identify tumors in dense breasts. Duke and IMAIOS have developed a free online app and ML algorithm to train the reading of breast #tomosynthesis images.

A Breast MRI is more invasive and only recommended for women at a higher than average risk of breast cancer. A Breast MRI has a higher rate of false-positive cases when compared to mammographies. Although AI is augmenting breast MRIs to improve accuracy, further research is needed to enhance their efficiencies. Recently, research presented at the 2019 Breast Imaging Symposium demonstrated the successful use of advanced algorithms on breast MRIs, which are traditionally used in cases of patients with particularly dense breast tissue.

Image Credit: Volume 34, Issue 5, 12 November 2018, Pages 840-851.e4

If mammographs and MRI fail to confirm cancer, a #biopsy is recommended to diagnose breast cancer. Combined with machine learning, Laser Raman Spectroscopy (LRS) offers real-time detection of cancer cells during a needle biopsy.​

Summary
With rapid advancements in Artificial Intelligence and robotic technology, the preliminary screening of breast cancer without physicians and the diagnosis of breast cancer without radiation are possible in the future!

References

• "Deep Learning Accurately Predicts Estrogen Receptor Status in Breast Cancer Metabolomics Data" J. Proteome Res. 2018, 17, 1, 337–347 Publication Date: November 7, 2017, https://doi.org/10.1021/acs.jproteome.7b00595
• "Deep learning-enabled breast cancer hormonal receptor status determination from base-level H&E stains" Naik, N., Madani, A., Esteva, A. et al. Deep learning-enabled breast cancer hormonal receptor status determination from base-level H&E stains. Nat Commun 11, 5727 (2020). https://doi.org/10.1038/s41467-020-19334-3
• Dai X, Cheng H, Bai Z, Li J. Breast Cancer Cell Line Classification and Its Relevance with Breast Tumor Subtyping. J Cancer 2017; 8(16):3131-3141. doi:10.7150/jca.18457. Available from https://www.jcancer.org/v08p3131.htm
• Varghese F, Bukhari AB, Malhotra R, De A (2014) IHC Profiler: An Open Source Plugin for the Quantitative Evaluation and Automated Scoring of Immunohistochemistry Images of Human Tissue Samples. PLoS ONE 9(5): e96801. https://doi.org/10.1371/journal.pone.0096801
• Yedjou, Clement G et al. “Health and Racial Disparity in Breast Cancer.” Advances in experimental medicine and biology vol. 1152 (2019): 31-49. doi:10.1007/978-3-030-20301-6_3
• N. Wu et al., "Deep Neural Networks Improve Radiologists’ Performance in Breast Cancer Screening," in IEEE Transactions on Medical Imaging, vol. 39, no. 4, pp. 1184-1194, April 2020, doi: 10.1109/TMI.2019.2945514.
• Mirsadeghi, L., Haji Hosseini, R., Banaei-Moghaddam, A.M. et al. EARN: an ensemble machine learning algorithm to predict driver genes in metastatic breast cancer. BMC Med Genomics 14, 122 (2021). https://doi.org/10.1186/s12920-021-00974-3
• Huang, Theodore. 2019. Statistical and Machine Learning Approaches for Family History Data. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.
• Pisano et al. NEJM 2005; 353: 1773
• Engmann NJ, et al, JAMA Oncol. 2017;3(9):1228-1236
• Mandelson et al. J Natl Cancer Inst 2000; 92:1081–1087
• AI Algorithm Detects Breast Cancer in MR Images. Imaging Technology News. https://www.itnonline.com/article/ai-algorithm-detects-breast-cancer-mr-images. Accessed Oct. 4, 2019.
• Kothari, Ragini et al. “Review of Laser Raman Spectroscopy for Surgical Breast Cancer Detection: Stochastic Backpropagation Neural Networks.” Sensors (Basel, Switzerland) vol. 20,21 6260. 2 Nov. 2020, doi:10.3390/s20216260
• Krishna Nand Keshavamurthy, Dmitry V. Dylov, Siavash Yazdanfar, Dharam Patel, Tarik Silk, Mikhail Silk, Frederick Jacques, Elena N Petre, Mithat Gonen, Natasha Rekhtman, Victor Ostroverkhov, Howard I. Scher, Stephen B. Solomon, Jeremy C. Durack
• https://www.who.int/news-room/fact-sheets/detail/breast-cancer#:~:text=Scope%20of%20the%20problem,the%20world's%20most%20prevalent%20cancer.
• https://www.cancer.gov/about-nci/budget/fact-book/data/research-funding
  https://www.komen.org/breast-cancer/facts-statistics/research-studies/topics/https://www.cdc.gov/cancer/breast/basic_info/dense-breasts.htm
• https://www.yalemedicine.org/conditions/3d-mammography
• https://www.imaios.com/en/Digital-Breast-Tomosynthesis/Breast-imaging-learning-tool
• https://github.com/histolab/histolab
• https://www.niramai.com/about/thermalytix/
• https://www.radiologyinfo.org/en/info/mammo
• https://news.mit.edu/2019/using-ai-predict-breast-cancer-and-personalize-care-0507
• https://www.gehealthcare.com/article/new-artificial-intelligence-breast-imaging-technology-demonstrates-profound-potential/?_ga=2.147395895.1035190492.1621813285-997565614.1621813285
• https://www.cancer.org/cancer/breast-cancer/understanding-a-breast-cancer-diagnosis/breast-cancer-hormone-receptor-status.html
• https://www.proteinatlas.org/learn/method/immunohistochemistry