Artificial Intelligence: What’s Good for Breast Imaging is Good for Prostate Imaging
What do breast cancer (BCa) and prostate cancer (PCa) have in common? As it turns out, they share several characteristics:
- They are both the second leading cause of cancer death in the U.S.
- They are both related to sexuality and reproduction
- They are both subject to hormone-driven cancers that respond to drugs that affect hormone levels
- Both often have a “latency period” that can last for years from the onset of a primary tumor until cancer begins to spread more widely in the body
- They both have a number of cell lines, some of which are more aggressive than others
The same gene mutations, BRCA1 and BRCA2, can be carried by men and women and contribute to more dangerous disease
- Both cancers can be screened for by imaging, and if detected early before there are symptoms, treatment is more likely to be successful.
The last point can be elegantly summed up in a statement that is true for both BCa and PCa: Early detection saves lives.
Detecting BCa early
Since 1976, when the American Cancer Society first recommended that women should get regular mammograms to screen for BCa, imaging became the companion standard of detection along with breast self-exam. Mammograms use very small doses of x-rays to scan for differences in tissue (breasts are compressed to flatten them to make x-ray penetration more effective). More recently, breast MRI is increasingly used for screening women with BCa risk factors. Although MRI is more costly and time-consuming, no compression is needed and there is no radiation. In addition, there are abbreviated imaging sequences that reduce the time in the scanner. In either case, image-based screening is worth it. If BCa is detected early when the tumor is still contained in the breast, the 5-year survival rate is 99%.
Detecting PCa early
Historically, screening for PCa was done using the PSA blood test, which came into wide use in the early 1990s. However, it is not specific for cancer, so when a suspiciously high result alerted a doctor, the patient was usually rushed to a needle biopsy (over detection). This often led to whole gland treatment for insignificant (indolent or nonaggressive) disease, leaving countless men with urinary and sexual side effects. More recently, doctors are responding to abnormal PSA results with a) a repeat blood test within a few months and b) multiparametric MRI (mpMRI) of the prostate. The use of imaging changed the landscape of PCa detection and diagnosis, since mpMRI is highly accurate for detecting significant disease. There are very low rates of false positives, and an added benefit is such precise depiction and characterization of tumors and in-bore biopsy requires only a minimum number of needles, and accurate diagnosis means that treatment can be planned to match the disease.
Artificial Intelligence improves detection in both cancers
Here’s something else that BCa and PCa have in common: a new discipline called radiomics is improving the speed and accuracy of cancer detection in both breast and prostate.
It is said that a picture is worth a thousand words. In the case of medical imaging, a picture is worth thousands of bits of data. In fact, mammograms and MRI scans ARE data! Traditionally, when a radiologist interpreted such images, he/she was using eyes and brain to analyze and translate the “data” (light and dark shapes) into verbal information about the patient. The more scans that went into the training and experience of the interpreter, the better able to identify (detect) cancer the radiologist became.
This is similar to “training” computer programs to recognize suspicious tumors based on clues too small for the human eye to perceive. The images used to train computers involve human input in terms of segmentation and user-defined characteristics,[i] but when it comes to reading the images, computers and humans differ. When a human views a scan, his/her evaluation is based on its qualitative characteristics. Computers, on the other hand, rely on quantitative values attached to each bit of data. From this, basic binary features are extracted: yes, this is cancer or no, this is not cancer. These values can then be assessed by Deep Learning algorithms, often “trained” on far more cases than a human radiologist might encounter in a professional lifetime. The algorithm can “compare” a patient’s breast or prostate scan against a large number of other scans more efficiently than a single radiologist.
Does this make the computer can replace the doctor? No. The computer can efficiently bring a red-flag detection to the doctor’s attention. This is a definite help for the doctor, but the final call still rests with the physician. Furthermore, the computer cannot know the patient, or all the factors that go into the patient’s family history, lifestyle, personality, emotional and psychological traits, possible exposure to pollution or toxins, etc. The computer can’t have a dialogue with the patient, or see the fear on a patient’s face, or work together with the patient to formulate next steps.
Still, by optimizing the sensitivity of screening mammography and MRI, the accuracy of early detection can only mean better diagnostic and treatment outcomes, and more women’s and men’s lives saved.
NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.
[i] Gillies RJ, Schabath MB. Radiomics Improves Cancer Screening and Early Detection. Cancer Epidemiol Biomarkers
Prev. 2020 Dec; 29(12): 2556-2567.
- Artificial Intelligence, Breast imaging, Prostate imaging