Background Information

Prostate cancer is a common cancer in men in the United States and worldwide. In spite of its high incidence and prevalence, it grows very slow. Because of the harmlessness of low-risk prostate cancer and its slow progression, early-detection tests of prostate cancer have been mired in controversy for the potential over-diagnosis, over-treatment and treatment-related side effects. Efficient detection and risk assessments of prostate cancer are in need.

Prostate Cancer

Prostate cancer (PCa) happens when malignant cells in the tissues of the prostate grow abnormally and out of control. It is an common malignancy worldwide. Statistics shows globally it is the most frequent malignancy in men, the second most common type of cancer and the fifth leading cause of cancer-related death in men. 

In spite of its high incidence and prevalence, the progression of prostate cancer in most men is relatively slow. Many prostate cancers never even become large or troublesome enough to be diagnosed clinically. 

Men with prostate cancer may be characterized as low-, intermediate-, or high-risk for having/developing metastatic disease or dying of prostate cancer. Low-risk prostate cancer is not likely to grow or spread for many years and high-risk prostate cancer may grow or spread within a few years. 

Controversy of Prostate Cancer Screening

Because of the harmlessness of low-risk prostate cancer and the slow progression of prostate cancer, there have been controversies around the early-detection of prostate problems and needs of accurate detection and efficient tests for risk-stratification. Prostate cancer screening is widely used in early-detection and risk-stratification of prostate cancer. However, it is controversial due to uncertainty of whether the benefits of screening ultimately outweigh the risks of over-diagnosis and over-treatment.

These downsides lead to the current confusion and disagreement among public health societies regarding which patients should be offered screening for prostate cancer and the need of efficient detection and risk stratification methods.

Read more about PCa

Histological Grading and Gleason Score

The wide-ranging biologic malignancy of prostate cancer is strongly correlated with its extensive and diverse morphologic appearances. The grade of prostate cancer is a commonly used variable on which the risk-stratification is based. Histological grading is a valuable research tool that is used more extensively and systematically in patient care.

The Gleason grading system is used to help evaluate the prognosis of men with prostate cancer using samples from a prostate biopsy. A Gleason score is given to cancer based upon its microscopic appearance, which is based entirely on the histological pattern of arrangement of carcinoma cells in HE-stained sections. It is calculated based on the dominant histological grades, from grade 1 (well differentiated) to grade 5 (very poorly differentiated). The classical score is derived by adding the two most prevalent pattern grades, yielding a score ranging from 2 to 10. The score is often provided by its separate components (e.g., Gleason score 3 + 3 = 6; or 4 + 5 = 9). Gleason scores provide useful cancer staging information in determining prognosis.

Cancers with a higher Gleason score are more aggressive and have a worse prognosis. Increasing Gleason grade is directly related to a number of histopathologic end points, including tumor size, margin status, and pathologic stage. Higher number indicates greater risks and higher mortality.

In this study, the Gleason scores were collected through pretreatment prediction of pathologic stage based upon needle biopsy Gleason grade. The needle biopsy results were interpreted by pathologists and only histological grades greater than 3 (which are considered to be cancerous) were recorded

Because of the advantage of noninvasive in-vivo assessment, MRI is more widely used in the collection and  evaluation of cancerous data. In this project, diffusion [MRI] histology (D-Histo), an advanced MRI technique developed to better differentiate between various pathologies in tumor, is used to for the classification of prostate cancer. 

Magnetic resonance imaging (MRI) machine.

Magnetic Resonance Imaging

Magnetic resonance imaging (MRI), with superior soft-tissue resolution, has a fundamental role in the diagnosis and management of many types of tumors. It uses the electromagnetic properties of hydrogen molecules to collect information about organs and other tissues and converts this to an image. Cancerous tissue has a different set of magnetic properties than surrounding normal tissue. MRI is able to capture these differences. Especially, diffusion-weighted MRI has the advantage of being able to perform noninvasive in-vivo assessment of the tumors, which is of great practical use in guiding the treatment of patient.

Read more about MRI

Diffusion MRI Histology

Diffusion [MRI] histology (D-Histo) is an advanced MRI technique developed in Dr. Song’s research group in BMI lab to better differentiate between various pathologies in tumor. It examines tumor micro-environment by modeling diffusion-weighted MRI signals as a linear combination of an anisotropic diffusion tensor and a spectrum of isotropic diffusion tensors. This technique generates multiple D-Histo maps, each measure some properties of the sample tissue like fiber and water ratio and present values of these properties as imaging voxels (with each voxel consisted of multiple pixels in a squared region).

Problem Statement

The project attempts to address two problems, both using data of imaging voxels generated by D-Histo techniques.

The first problem is classifying cancerous prostate tissues and benign prostate tissues, which is the task of prostate cancer detection.

The second problem is differentiating low-risk and high-risk prostate cancer tissues, which is the task of risk-stratification. The staging criteria used follow the Gleason scoring system. For simplicity, those with Gleason scores greater or equal to 4+4 are treated as high-risk cancer tissues and those with scores below that are treated as low-risk cancer tissues. 


To design and test a framework that can automatically perform the tasks of using D-Histo data obtained to perform these two classification tasks on prostate tissues.