Prostate cancer has sometimes been called “the emperor's cancer” because prominent figures such as John Kerry have been diagnosed with it, and “father's cancer” because it commonly affects men over 50.
Prostate cancer cannot be distinguished from other prostate diseases by its symptoms alone, it requires accurate testing, and the treatment process is different depending on the disease, so accurate diagnosis is of paramount importance.
AI may assist prostate-cancer diagnosis. This article examines a JLK Inspection patent concerning detection of prostate-cancer regions using a deep-learning model.
[Patent]
Click the image to view the published patent application.According to this patent, when the MRI taken is entered into the diagnostic region detection model, the parametric MRI with the diagnostic region is output. That is, the diagnostic region detection model performs a function of detecting which region contains the prostate (diagnostic region).
On the other hand, the cancer region detection model is inputted with a parametric MRI of the diagnostic region to output a probability map indicating the area in which the cancer is expected. Here, the probability map is expressed as what is the probability that each of the coordinate regions contains cancer. An area with a probability value that exceeds the threshold value in the probability map may be recognized as an area containing cancer.
The key is to adjust the detection threshold using clinical information, specifically the prostate-specific antigen (PSA) value.
A higher prostate specific antigen is associated with a higher risk of prostate cancer. In this case, the sensitivity of the cancer region detection model will increase.
On the other hand, a lower prostate specific antigen is associated with a lower risk of prostate cancer. The sensitivity of the cancer region detection model decreases.
The central feature of the invention is post-processing that improves accuracy by adjusting the threshold used to detect cancerous regions.
This article reflects the information available when it was published. Contact us to discuss your circumstances.