Back to the list
Congress: ECR25
Poster Number: C-27014
Type: Poster: EPOS Radiologist (scientific)
Authorblock: I. Stamouli1, I. Gatos1, P. Katsakiori1, S. Tsantis1, T. Kalathas2, V. Chatzipavlidou2, N. Papathanasiou1, G. Kagadis1; 1Rion/GR, 2Thessaloniki/GR
Disclosures:
Ioanna Stamouli: Nothing to disclose
Ilias Gatos: Nothing to disclose
Paraskevi Katsakiori: Nothing to disclose
Stavros Tsantis: Nothing to disclose
Theodoros Kalathas: Nothing to disclose
Vasiliki Chatzipavlidou: Nothing to disclose
Nikos Papathanasiou: Nothing to disclose
George Kagadis: Nothing to disclose
Keywords: Artificial Intelligence, Molecular imaging, Nuclear medicine, PET-CT, CAD, Molecular imaging, Segmentation, Cancer, Molecular, genomics and proteomics, Multidisciplinary cancer care
Purpose Prostate Cancer (PCa) represents the most prevalent malignancy in men and ranks as the second most common cause of death [1]. A significant percentage of patients, ranging from 20% to 50%, will experience disease recurrence, which is characterized by elevated levels of Prostate Specific Antigen (PSA). This state is clinically defined as biochemical recurrence (BCR) (PSA > 0.2 ng/ml) [2]. During BCR, metastases are likely to occur in different anatomical regions. Although rising PSA levels indicate BCR, they do not provide...
Read more Methods and materials The study utilized a clinical dataset of 71 male patients who underwent PET/CT imaging. For each patient, the Maximum Intensity Projection (MIP) was created on the PET coronal plane series. Manual cropping of regions of interest (ROIs) was performed on suspicious findings, based on the diagnoses of experienced nuclear medicine physicians, generating 157 ROIs in total, of which 76 were labeled as FP and 81 as TP. Each ROI image was resized to 224x224.Then, the images were divided into two...
Read more Results For the original dataset, the model achieved mean sensitivity of 86.73% ± 5.54% and mean specificity of 78.36% ± 12.3%. The overall mean accuracy was 82.23% ± 7.11%, showcasing the model’s balanced performance across both classes.  The mean AUROCC was 89.77% ± 4.24%. When trained on the augmented dataset, the model maintained its sensitivity at 86.73% ± 5.54%, demonstrating that data augmentation did not compromise its ability to detect true positives. Notably, the mean specificity improved to 80.66% ± 8.55%, indicating...
Read more Conclusion The DenseNet201 model demonstrates robust performance in the classification of FP and TP findings in 18F-PSMA PET/CT images of biochemical recurrent patients with PCa. The introduction of data augmentation techniques led to improvements in specificity, accuracy, and overall model performance, as reflected in the higher AUROCC values.
Read more References [1]       Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin 2022;72:7–33. https://doi.org/10.3322/caac.21708.[2]       Cornford P, van den Bergh RCN, Briers E, Van den Broeck T, Cumberbatch MG, De Santis M, et al. EAU-EANM-ESTRO-ESUR-SIOG Guidelines on Prostate Cancer. Part II—2020 Update: Treatment of Relapsing and Metastatic Prostate Cancer. Eur Urol 2021;79:263–82. https://doi.org/10.1016/j.eururo.2020.09.046.[3]       Hofman MS, Lawrentschuk N, Francis RJ, Tang C, Vela I, Thomas P, et al. Prostate-specific membrane antigen PET-CT in patients with high-risk prostate cancer...
Read more
GALLERY