Jump to: navigation, search

Computer-aided detection of prostate cancer

Prostate Cancer

Prostate cancer is the second most common cancer in men, with 10000 new cases and 2500 deaths every year in the Netherlands (Cancer Society). These numbers increase every year due to the ageing of the general population.

Screening can help reduce mortality. The results of a large multi-center ERSPC trial Schroder et al NEJM 2009 showed that a screening program using a prostate specific antigen (PSA) blood test and transrectal ultrasound (TRUS) biopsy could reduce mortality up to 30%. However, this mortality reduction would lead to a massive amount of over-treatment in men. To save one life, 1410 healthy men would have to do a PSA test, 225 biopsied, and 48 men treated for an indolent cancer. Screening was therefore not recommended.

Prostate MRI

Prostate MRI has become very accurate in detection of prostate cancer. A recent study showed that compared to TRUS biopsy, MR biopsy finds 18% more clinically significant cancers while reducing the number of required biopsies by a third Pokorny et al Eur Urol 2014. However, reported diagnostic accuracy varies strongly. One recent Prostate MR evaluation study in biopsy naive men with template guided biopsy reference standard showed a sensitivity of 97% for clinically significant cancers Grey et al BJUI 2015. The negative predictive value (NPV) was 98%, which explains why a negative MR can avoid an unnecessary biopsy. However, the specificity was only 60%. Another similar study Abd-Alazeez et al 2014 reported a specificity of 21%. Yet the Pokorny study reports a specifity of 95%.

Improving prostate MR specificity is one of the current challenges. For screening purposes specificity it is currently not enough (it is in the order of 90-99% in breast cancer screening). Prostate MR is a complex multi-parametric image analysis problem. This is the domain where Computer Aided Diagnosis can help improve prostate cancer MR diagnosis. CAD's trainable pattern recognition and multi-variate statistical analysis can optimally use available information.

Improving prostate MR reporting variability is another challenge. A structured reporting system has been defined in PIRADS 2.0 2015. Computer Aided Diagnosis is well known for its ability to reduce reader variability when used as a second reader.

This long term project explores how CAD can help improve prostate cancer MR diagnosis.

Top-Left: Diffusion weighted image. Top-Right: Apparant diffusion coefficient-map. Bottom-Left: Dynamic contrast enhanced image. Bottom-Right: T2-weighted image

A general prostate MRI exam consists of multiple MR images to give both anatomical and functional information about the prostate tissue. The three typically acquired image types are T2-weighted, diffusion-weighted and dynamic contrast enhanced images. High-resolution T2-weighted images are used to assess the anatomical structure of the prostate tissue. The diffusion-weighted images provide an indication of the cellular structure of the prostate. Dynamic contrast enhanced images can be used to determine attributes like blood vessel surface area and permeability in prostate tissue.

One of the remaining issues with prostate MRI is that it requires substantial experience to detect and categorize prostate cancer. The availability of such experienced radiologists is scarce. In addition, even for an experienced radiologist it is difficult to extract all the information from the data. To give an indication, in a typical exam 3 T2-weighted images are acquired, in addition to 5 diffusion-weighted images and 35 dynamic-contrast enhanced images.

Development of a computer-aided detection system for prostate cancer

Computer-aided detection of prostate cancer can perhaps provide a solution. Computer algorithms allow us to combine the enormous amount of images into a much smaller amount of images with a high information content. For example, using pharmacokinetic analysis we can reduce the 35 dynamic contrast enhanced images to 2 pharmacokinetic parameter maps.

Left: Segmentation of the prostate zones. Middle: Cancer probability map with ground truth in white. Right: CAD result in red with probability value, ground truth in white.

These images can be used by the radiologist directly, but using machine learning and pattern recognition techniques computer-aided detection of suspicious regions is also possible. This creates a synthetic 'second reader', allowing every image set to be read by two readers, the radiologist and the computer. In the end this will help reduce the workload for the radiologists while increasing their detection performance.

Computer-aided detection of prostate cancer in clinical practice

To make our algorithms available to the clinic a prostate workstation was created. This workstation is supposed to streamline the reading and reporting of prostate MRI, while also providing advanced features like pharmacokinetic analysis and computer-aided detection in a user-friendly way. The first iteration of our workstation is in current clinical practice and is used on a daily basis to report prostate MR. Currently work is well underway for the next version.

Example image of the prostate workstation

The close interaction with the clinic provides us with valuable feedback and support to improve our algorithms. This will result in better tools, which allow the radiologists to work more efficiently and more accurately. In the end this will result in an improvement in patient care.



Key publications