Development of Infrared Based Tests for the Diagnosis of Prostate Cancer

  • Julie Brooks

Student thesis: Phd


Prostate cancer is the most frequently diagnosed cancer in males within the western world. Current practices to identify prostate cancer are limited to screening for prostate specific antigen in blood samples and a digital rectal examination, both of which lack sensitivity. If a post-operative examination identifies malignancies, the prognosis is generally uncertain. Although the majority of prostate tumours are indolent in nature, for some patients the cancer will rapidly metastasise and life expectancy can be as short as 1-2 months. At present, identifying high-risk patients is estimated based upon the outcome of collated pathological results, such as PSA score and Gleason score. The aim of this project was to investigate if spectral biomarkers relating to prostate cancer can be identified using infrared based platforms. The project is divided into two clinical research areas; blood analysis and tissue analysis. The first strand of the project investigates the use attenuated total reflection-Fourier transform infrared spectroscopy as a ‘liquid biopsy’ application for the diagnosis of prostate cancer in blood serum. To begin with, a number of preliminary investigations were carried out in the pre-analytical stage. This included a drying study, a small pilot study, and investigations into the spectral variance within sample replicates. Once all pre-analytical investigations were resolved, a clinical diagnostic study was initiated to determine if patients with prostate cancer can be discriminated from those with BPH using 1 μL of blood serum. The study examined the spectral profiles of 58 patients (26 with BPH and 32 with prostate cancer) in triplicate measurements, and found that patients can be discriminated on benign and malignant prostate conditions. A radial basis function support vector machine (RBF-SVM) model achieved mean sensitivity and specificity values of 94%, and 77%, respectively. The second strand of the project investigates if spectral markers of metastatic prostate cancer can be identified in post-operative prostate needle biopsies using Fourier transform infrared spectroscopic chemical imaging and a Random Forest classification model. The study progresses in three stages and investigates differences in classification accuracy when using different intra-core spectral profiles for training a classification model, and the effects on classification accuracy using different patients in the training and testing groups. In the first two stages 36 patients (17 metastatic and 19 non-metastatic patients) are investigated and in the third stage at total of 60 are investigated (30 metastatic and 30 non-metastatic patients). Annotations were performed on infrared images to extract only tumour epithelium for analysis. By dividing the patients in to training and testing groups, a robust Random Forest model was constructed using tumorous epithelium from metastatic and non-metastatic cores. The Random Forest model achieved sensitivities and specificities ranging from 57 – 93% and 41 – 99%, respectively. In addition to the metastatic study, a further study was carried out to investigate if surrogate spectral marker for EphA2 expression in TURP cores can be identified using chemical imaging. High EphA2 in prostate epithelium has been linked to poorer prognosis and can be detected in the early stages of the disease. Again, the study uses the application of annotating infrared images to extract key spectral profile for interrogation. In the first instant, tumour and normal associated tissue cores were assessed to examine if same tissue types can be discriminated according to EphA2 expression. The study then goes on to use EphA2 stained images as references in annotating specific homogenous regions of EphA2 on infrared images with the notion of building a classifier to identify low and high EphA2 expression in independent tissue cores. Image registration is applied as a way to improve annotations on a pixel level, however,
Date of Award31 Dec 2017
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorNicholas Lockyer (Supervisor) & Peter Gardner (Supervisor)


  • metastatic prostate cancer
  • PSA test
  • biopsy
  • blood serum
  • chemical imaging
  • FTIR spectroscopy
  • prosstate cancer

Cite this