Detection of hypovascular colorectal liver metastases with photon-counting detector CT versus energy-integrating detector CT: a blinded intra-patient comparison of routine clinical protocols using an MRI-anchored reference standard

Background Contrast-enhanced CT remains the staging workhorse in metastatic colorectal cancer but under-detects small hypovascular colorectal liver metastases (CRLM) relative to MRI. Photon-counting detector CT (PCCT) offers higher spatial resolution and lower electronic noise, yet clinical evidence on lesion detection is scarce. Methods In this single-centre, intra-patient reader study, 15 patients with colorectal cancer and hepatic metastases underwent both energy-integrating detector CT (EID-CT) and PCCT within the same month with each scanner’s routine protocol (PCCT: 140 kV, 0.4 mm sections; EID-CT: 100 kV, 1.0 mm sections), during documented oncological stability, plus contrast-enhanced liver MRI. A consensus panel defined an MRI-anchored reference map of 46 metastases. Three radiologists of differing experience independently reviewed both arms in randomised order, blinded to scanner type, clinical data and reference standard, without cue to lesion number or location. Detection, conspicuity and diagnostic confidence were scored per reference lesion. PCCT was read at conventional-equivalent monoenergies (65–70 keV). Results Per-lesion sensitivity against the MRI-anchored reference was 138/138 reads (100%; 95% CI 97.3–100) for PCCT versus 122/138 (88.4%; 95% CI 82.0–92.7) for EID-CT; all 16 discordant reads favoured PCCT (exact McNemar P < 0.001; patient-clustered bootstrap difference +11.5 points, 95% CI 1.6–24.1). Eight of 46 metastases (17.4%) were missed on EID-CT and none on PCCT. The complete metastatic burden was reproduced in 45/45 patient-level PCCT reads versus 37/45 (82.2%) on EID-CT. Composite conspicuity was 4.57 ± 0.48 versus 3.53 ± 0.88 (+1.04, patient-clustered 95% CI 0.72–1.33; P < 0.001) and diagnostic confidence 4.28 ± 0.53 versus 3.62 ± 0.87 (+0.66, 95% CI 0.42–0.89; P < 0.001). The confidence gain was smaller for the least experienced reader (mixed-effects model P = 0.004). False-positive findings were not recorded, so specificity could not be assessed. Conclusion In this preliminary comparison of two routine clinical protocols, PCCT showed higher per-lesion sensitivity than EID-CT, identifying every reference metastasis, and improved conspicuity and confidence, even without low-keV optimisation. Because section thickness, tube voltage, pitch and kernel differed between protocols, the advantage reflects the PCCT system as a whole rather than the detector alone. PCCT may narrow the CT–MRI detection gap in CRLM, a hypothesis requiring confirmation in larger free-response studies with matched reconstruction parameters.