Using Low Energy Mono-Energetic CT Reconstruction to Improve Detection of High-Risk Prostate Cancer
Using Low Energy Mono-Energetic CT Reconstruction to Improve Detection of High-Risk Prostate Cancer
Dominic Kite, Paul Ward, Mathuri Sakthithasan, Richard Keogan, Elena Lambert, Dow-Mu Koh, Paul Burn
To evaluate the diagnostic effectiveness of low energy mono-energetic CT reconstruction in detection of high-risk prostate cancer.
Single-centre retrospective study, scans performed at Taunton Diagnostic Unit, Alliance Medical Ltd. Participants: Consecutive patients with histologically-confirmed, newly diagnosed, high-risk prostate cancer (Cambridge Prognostic Group 4/5) who had prostate MRI and CT staging (in early 2023). A convenience series of patients without known prostate cancer were evaluated as controls (from same time period). Data Collection: Images were acquired on contrast-enhanced CT acquired using Philips IQon detector-based spectral CT with conventional and 50keV mono-energetic reconstructions (Low MonoE). Part One – for the cancer group, the attenuation (HU) of each index lesion and background prostate (peripheral zone) was measured for each reconstruction using best-fit ovoid regions of interest. MRI provided reference for tumour site. Part Two – for all patients, two readers, blinded to clinicopathological details, independently scored each reconstruction for likelihood of cancer (Likert scale 1-5, 5=definite cancer, Likert 4 and 5 designated test positive). Data Analysis: The mean attenuation values and difference in values between lesion and background were compared using paired t-test. Inter-reader agreement for Likert scoring was assessed using Cohen’s kappa test. The diagnostic sensitivity of cancer detection using Low MonoE and conventional reconstruction were compared by McNemar’s test (subset analysis containing cancer cases only).
Figure 1. Prostate cancer (confirmed with MRI and histology) (arrows) better seen on Low MonoE (right) than conventional reconstruction (left).
Figure 2. Left - violin plot of attenuation of prostate cancer on conventional and Low MonoE reconstruction.
Figure 3. Right - violin plot of difference in attenuation between prostate cancer and background prostate on conventional and Low MonoE reconstruction.
RESULTS Part One 34 cancer patients (mean age 72 years) and 9 controls (mean age 67 years) were included. For cancer patients, ISUP grade group (GG): GG2 8 patients, GG3 12, GG4 3, and GG5 11. Prostate cancer attenuation (mean HU±SD) was higher on Low MonoE(108±30), compared with conventional reconstruction (67±18) [p<0.0001]. See Figure 1 and 2. Lesion to background attenuation difference (mean HU±SD) was greater on Low MonoE (40±30) than conventional reconstruction (21±19) [p<0.0001]. See Figure 3. RESULTS Part Two Pooled sensitivity and accuracy (95% confidence interval) for detecting cancer were 65% (52-76%) and 64% (53-74%) respectively using Low MonoE reconstruction versus 12% (5-22%) and 30% (21-41%) using conventional reconstruction [McNemar’s p<0.001]. There was a higher rate of false positives on Low MonoE (39%) compared to conventional reconstruction (0%). Inter-reader agreement (cancer cases): conventional, moderate (kappa 0.44); mono-energetic, fair (kappa 0.25).
Table 1. Cross-tabulations of Likert scoring on conventional (white) and Low MonoE reconstruction (grey)
DISCUSSION AND LIMITATIONS Our control group is small. In addition, assumption of benignity in this group is based on clinical background and not (inevitably), with biopsy confirmation. Although the sensitivity and accuracy for cancer detection was improved using Low MonoE reconstruction, this was associated with higher false positives and lower inter-reader agreement. Future studies can explore different values of MonoE reconstruction to determine whether this can improve diagnostic performance. CONCLUSIONS Low energy mono-energetic reconstructions significantly increase lesion attenuation and sensitivity for detection of high-risk prostate cancer. Use of Low MonoE in clinical practice may aid cancer detection, pending larger studies investigating specificity.
