Wildman-Tobriner Benjamin, Felice Nicholas, Kalisz Kevin R, Allen Brian C, Thomas Sarah P, Kruse Danielle E, Segars William Paul, Harrawood Brian, Bashir Mustafa R, Marin Daniele, Morrison Samantha, Erkanli Alaattin, Samei Ehsan, Abadi Ehsan
From the Department of Radiology, Duke University Hospital, 2301 Erwin Rd, Box 3808, Durham, NC 27701 (B.W.T., K.R.K., B.C.A., S.P.T., D.E.K., B.H., M.R.B., D.M., E.S., E.A.); Department of Biostatistics and Bioinformatics (N.F., S.M., A.E.) and Department of Medical Physics (W.P.S., E.S., E.A.), Duke University, Durham, NC.
Radiology. 2025 Jan;314(1):e241568. doi: 10.1148/radiol.241568.
Background Detection of hepatic metastases at CT is a daily task in radiology departments that influences medical and surgical treatment strategies for oncology patients. Purpose To compare simulated photon-counting CT (PCCT) with energy-integrating detector (EID) CT for the detection of small liver lesions. Materials and Methods In this reader study (July to December 2023), a virtual imaging framework was used with 50 anthropomorphic phantoms and 183 generated liver lesions (one to six lesions per phantom, 0.4-1.5 cm in diameter). Virtual CT platforms simulated PCCT and EID CT scanners. Phantoms were virtually scanned using routine (6 mGy) and low-dose (1.5 mGy) conditions and reconstructed with three kernels. A subset of 300 scans (150 PCCT vs EID CT pairs) were selected. Four radiologists independently reviewed all scans to mark liver lesions, assigned confidence scores for detection, and rated scan quality. Analysis was performed on a per-lesion basis to determine sensitivity for several variables and on a per-scan basis for scan quality. The McNemar test, two-sided paired tests, and mixed-effects logistic regression models were fitted; < .05 was considered indicative of statistically significant difference. Results Consensus reader sensitivity in detecting lesions was 82.1% (451 of 549) for PCCT versus 77.6% (426 of 549) for EID CT ( < .001), with a mean sensitivity gain of 4.3 percentage points ± 1.3 ( < .001 to = .02 per reader). Readers had better subjective confidence for lesions at PCCT (mean score, 61.5 ± 22 vs 56.1 ± 24 [on a 101-point scale]; < .001). Sensitivity was lower for lesions smaller than 1 cm, with more pronounced difference between PCCT and EID CT (74.0% [271 of 366] vs 67.2% [246 of 366]; < .001). At the lower dose level, PCCT showed higher sensitivity than EID CT (68.9% [168 of 244] vs 61.1% [149 of 244]; < .001) for subcentimeter lesions. In a multivariable model, PCCT was independently associated with increased odds of lesion detection (odds ratio, 1.55; < .001). Image quality was slightly higher for PCCT (mean score, 55.3 vs 50.6 [on a 101-point scale]; < .001). Conclusion Compared with EID CT, PCCT showed better sensitivity in the detection of small liver lesions. © RSNA, 2025 See also the editorial by Menu in this issue.
CT检测肝转移瘤是放射科的日常工作,会影响肿瘤患者的药物及手术治疗策略。目的:比较模拟光子计数CT(PCCT)与能量积分探测器(EID)CT对肝脏小病灶的检测能力。材料与方法:在本读者研究(2023年7月至12月)中,使用虚拟成像框架及50个仿真人体模型和183个生成的肝脏病灶(每个模型1至6个病灶,直径0.4 - 1.5厘米)。虚拟CT平台模拟PCCT和EID CT扫描仪。在常规(6 mGy)和低剂量(1.5 mGy)条件下对模型进行虚拟扫描,并用三种核函数进行重建。选取300次扫描的子集(150对PCCT与EID CT扫描)。四位放射科医生独立审阅所有扫描以标记肝脏病灶,为检测结果指定置信度评分,并对扫描质量进行评级。基于每个病灶进行分析以确定对多个变量的敏感性,并基于每次扫描分析扫描质量。采用McNemar检验、双侧配对检验和混合效应逻辑回归模型;P < 0.05被认为具有统计学显著差异。结果:读者对PCCT检测病灶的共识敏感性为82.1%(549个病灶中451个),而EID CT为77.6%(549个病灶中426个)(P < 0.001),平均敏感性增益为4.3个百分点±1.3(每位读者P < 0.001至P = 0.02)。读者对PCCT上的病灶主观置信度更高(平均评分,61.5 ± 22 vs 56.1 ± 24[101分制];P < 0.001)。对于小于1厘米的病灶,敏感性较低,PCCT和EID CT之间的差异更明显(74.0%[366个病灶中271个] vs 67.2%[366个病灶中246个];P < 0.001)。在较低剂量水平下,对于亚厘米病灶,PCCT的敏感性高于EID CT(68.9%[244个病灶中168个] vs 61.1%[244个病灶中149个];P < 0.001)。在多变量模型中,PCCT与病灶检测几率增加独立相关(优势比,1.55;P < 0.001)。PCCT的图像质量略高(平均评分,55.3 vs 50.6[101分制];P < 0.001)。结论:与EID CT相比,PCCT在检测肝脏小病灶方面表现出更好的敏感性。©RSNA,2025 另见本期Menu的社论。