New developments in computed tomography : low dose techniques, image reconstruction, and spectral imaging

CT is an indispensable modality in any modern radiology department. It has revolutionized medical imaging, and the global frequency of CT scans is steadily increasing. As novel developments in CT hardware and software emerge, it is important to evaluate the clinical implications of these advancements. CT protocols should be optimized to maximize image quality for specific diagnostic tasks while reducing radiation exposure for both patients and radiology personnel in accordance with the ALARA principle. This dissertation has assessed how novel CT technologies related to radiation dose reduction, image reconstruction, and spectral imaging can be implemented in four specific diagnostic tasks: 1) CT-guided musculoskeletal (MSK) biopsies, 2) evaluation of distal extremities in an emergency setting, 3) CT head examinations of trauma patients, and 4) CT abdomen examinations of patients with PDAC.

The overarching aim of this thesis was to assess how novel CT technology can be implemented in various CT protocols to reduce radiation dose and to enhance image quality.

In Study I, we assessed a low-dose CT protocol for MSK biopsies that utilized several radiation dose reduction methods, including single-shot axial wide scanning, tube current modulation, and an IR algorithm. 57 patients underwent biopsies with the low-dose CT protocol (Group A), and 58 patients underwent biopsies with a standard CT protocol (Group B). A spreadsheet was created for Protocol A to calculate the start and end positions of the control scan, replicating the functionality built into the software of Protocol B. The mean total dose length product (DLP) was significantly lower in Group A compared to Group B (41.5+2.9 vs. 257.4+22.0 mGy*cm; p < 0.001), corresponding to an 83.9% dose reduction. No significant differences were observed in biopsy success rate, quantitative and qualitative image quality parameters, or workflow parameters.

In Study II, an ultra-low-dose CT (ULD-CT) protocol for distal extremities was evaluated in comparison to digital radiography (DR) in terms of diagnostic accuracy, radiation dose, and workflow efficiency. 207 extremities in 203 consecutive patients from the emergency department were included. Patients underwent DR followed by an ULD-CT protocol that incorporated several radiation dose reduction techniques. DR detected fractures in 73/207 (35.3%) extremities, while ULD-CT detected fractures in 109/207 (52.7%) extremities, resulting in an odds ratio for fracture detection between ULD-CT and DR of 2.0 (95% CI, 1.4-3.0). ULD-CT also yielded additional relevant diagnostic information, such as articular involvement and loose intraarticular bodies. The mean total time, including both the examination time and the time to produce a preliminary report, was shorter for ULD-CT than DR (7.6+2.5 vs. 9.8+4.7 min, p = 0.002). The diagnostic information from ULD-CT led to changes in treatment recommendation for 34/207 (16.4%) extremities, with treatment upgrades in 27/207 (13.0%) and downgrades in 7/207 (3.4%) extremities.

In Study III, the image quality between an IR algorithm (ASiR-V) and a DLR algorithm for head CT scans of 94 consecutive trauma patients was compared. Quantitative image quality parameters included image noise, signal-to-noise ratio (SNR) in gray and in white brain matter, and contrast-to-noise ratio (CNR) between gray and white brain matter. Qualitative image quality parameters comprised: image noise, artifacts, brain structures, and the conspicuity of intracranial hemorrhage (ICH). All image quality parameters were compared across four image reconstructions: ASIR-V and three strength levels of DLR, high (DLIR-H), medium (DLIR-M), and low (DLIR-L), for the same CT scan. DLIR-H and DLIR-M demonstrated significantly higher SNR and CNR, and lower image noise than ASiR-V and DLIR-L (p < 0.001 for all pairwise comparisons). DLIR-H had higher SNR and CNR and lower image noise than DLIR-M (p < 0.001-0.016). DLIR- H achieved up to 53.3% higher CNR and up to 82.9% higher SNR compared to ASİR-V. No significant difference in CNR was observed between DLIR-L and ASiR-V. Reader scores were higher for DLIR-H and DLIR-M compared to ASiR-V and DLIR-L, with scores generally increasing as the strength of the DLR increased. No CT images reconstructed with DLIR-H received a non-diagnostic grade for ICH. The scores of the less experienced and second most experienced readers showed greater improvement compared to those of the most experienced reader.

In Study IV, the image quality of MD-iodine images was compared to the image quality of virtual monoenergetic images (VMIs) at different keV levels, for assessing PDAC in the portal venous phase (PVP). From spectral data of 50 PCD-CT examinations (46 patients) with PDAC, MD-iodine images and VMIs at 55 keV and 70 keV were generated. Quantitative and qualitative image quality was compared across these three image reconstructions. Quantitative image quality was evaluated by calculating the CNR between PDAC and normal pancreatic parenchyma. Qualitative image quality parameters included PDAC conspicuity, image noise, sharpness of pancreatic and surrounding structures ("structures"), and overall image quality. MD-iodine had significantly higher CNR compared to 70 keV (p = 0.003), but lower compared to 55 keV (p = 0.049). MD-iodine received significantly lower reader scores for all parameters compared to both VMI reconstructions (p < 0.001 for pairwise comparisons). VMI at 55 keV was superior to both MD-iodine and VMI at 70 keV in terms of both quantitative and qualitative image quality.

List of scientific papers

I. First experiences of a low-dose protocol for CT-guided musculoskeletal biopsies combining different radiation dose reduction techniques.

Alagic Z, Alagic H, Bujila R, Srivastava S, Jasim S, Lindqvist M, Wick MC. Acta Radiol. 2020 Jan;61(1):28-36. https://doi.org/10.1177/0284185119847676

II. Ultra-low-dose CT for extremities in an acute setting: initial experience with 203 subjects.

Alagic Z, Bujila R, Enocson A, Srivastava S, Koskinen SK. Skeletal Radiol. 2020 Apr;49(4):531-539. https://doi.org/10.1007/s00256-019-03309-7

III. Deep learning versus iterative image reconstruction algorithm for head CT in trauma.

Alagic Z, Diaz Cardenas J, Halldorsson K, Grozman V, Wallgren S, Suzuki C, Helmenkamp J, Koskinen SK. Emerg Radiol. 2022 Apr;29(2):339-352. https://doi.org/10.1007/s10140-021-02012-2

IV. Photon-counting detector computed tomography: iodine density versus virtual monoenergetic imaging of pancreatic ductal adenocarcinoma.

Alagic Z, Valls Duran C, Suzuki C, Halldorsson K, Svensson-Marcial A, Saeter R, Koskinen SK. Abdom Radiol. 2024 Sep; Online ahead of print. https://doi.org/10.1007/s00261-024-04605-0