Exploring the Potential use of Convolutional Neural Networks for Clinically Significant Prostate Cancer Detection through the lens of age
Bachelor thesis
Permanent lenke
https://hdl.handle.net/11250/3075640Utgivelsesdato
2023Metadata
Vis full innførselSamlinger
- Studentoppgaver (TN-IDE) [823]
Beskrivelse
Full text not available
Sammendrag
Prostate Cancer (PCa) represents a significant public health challenge in developed nations, with over 5,000 men being diagnosed with the disease annually in Norway alone, rendering it the most frequently occurring cancer among men in the country. Testing for PCa has proven to be a challenging process requiring various test in order to confirm the presence of PCa. These include the Prostate Specific Antigen (PSA) test, a simple blood test that measures PSA levels in the blood, Digital Rectal Examination (DRE), which involves a medical practitioner inserting a finger into the patient’s rectum to assess the prostate’s size and texture. In addition, biopsies are also taken, which involves obtaining a tissue sample for further study. Despite their utility, these testing methods are susceptible to erroneous conclusions, with PSA levels potentially being influenced by factors beyond prostate size, biopsy samples possibly being cancer-free, and the possibility of human error during a DRE.In light of these challenges, this thesis explores the potential of Convolutional Neural Networks (CNNs) as a screening tool for PCa and severity assessment. Specifically, we examine the role of age as the primary indicator/determiner of PCa and its severity. Our methodology involves evaluating the accuracy of the predicted age through the use of Mean Absolute Error (MAE), comparing it to the patient’s actual age. By using this approach, we aim to determine the viability of CNNs in improving PCa detection and severity assessment while overcoming the limitations of current screening methods such as PSA.The results of this thesis show that DL for age prediciting can be a valuable technique in revealing PCa, with significant differences between the MAE of patients with cancer (159) and patients without it (91) in an independent test cohort. Prostate Cancer (PCa) represents a significant public health challenge in developed nations, with over 5,000 men being diagnosed with the disease annually in Norway alone, rendering it the most frequently occurring cancer among men in the country. Testing for PCa has proven to be a challenging process requiring various test in order to confirm the presence of PCa. These include the Prostate Specific Antigen (PSA) test, a simple blood test that measures PSA levels in the blood, Digital Rectal Examination (DRE), which involves a medical practitioner inserting a finger into the patient’s rectum to assess the prostate’s size and texture. In addition, biopsies are also taken, which involves obtaining a tissue sample for further study. Despite their utility, these testing methods are susceptible to erroneous conclusions, with PSA levels potentially being influenced by factors beyond prostate size, biopsy samples possibly being cancer-free, and the possibility of human error during a DRE.In light of these challenges, this thesis explores the potential of Convolutional Neural Networks (CNNs) as a screening tool for PCa and severity assessment. Specifically, we examine the role of age as the primary indicator/determiner of PCa and its severity. Our methodology involves evaluating the accuracy of the predicted age through the use of Mean Absolute Error (MAE), comparing it to the patient’s actual age. By using this approach, we aim to determine the viability of CNNs in improving PCa detection and severity assessment while overcoming the limitations of current screening methods such as PSA.The results of this thesis show that DL for age prediciting can be a valuable technique in revealing PCa, with significant differences between the MAE of patients with cancer (159) and patients without it (91) in an independent test cohort.