Potential of GPT-4 for Detecting Errors in Radiology Reports: Implications for Reporting Accuracy.

Background Errors in radiology reports may occur because of resident-to-attending discrepancies, speech recognition inaccuracies, and large workload. Large language models, such as GPT-4 (ChatGPT; OpenAI), may assist in generating reports. Purpose To assess effectiveness of GPT-4 in identifying common errors in radiology reports, focusing on performance, time, and cost-efficiency. Materials and Methods In this retrospective study, 200 radiology reports (radiography and cross-sectional imaging [CT and MRI]) were compiled between June 2023 and December 2023 at one institution. There were 150 errors from five common error categories (omission, insertion, spelling, side confusion, and other) intentionally inserted into 100 of the reports and used as the reference standard. Six radiologists (two senior radiologists, two attending physicians, and two residents) and GPT-4 were tasked with detecting these errors. Overall error detection performance, error detection in the five error categories, and reading time were assessed using Wald χ2 tests and paired-sample t tests. Results GPT-4 (detection rate, 82.7%;124 of 150; 95% CI: 75.8, 87.9) matched the average detection performance of radiologists independent of their experience (senior radiologists, 89.3% [134 of 150; 95% CI: 83.4, 93.3]; attending physicians, 80.0% [120 of 150; 95% CI: 72.9, 85.6]; residents, 80.0% [120 of 150; 95% CI: 72.9, 85.6]; P value range, .522-.99). One senior radiologist outperformed GPT-4 (detection rate, 94.7%; 142 of 150; 95% CI: 89.8, 97.3; P = .006). GPT-4 required less processing time per radiology report than the fastest human reader in the study (mean reading time, 3.5 seconds ± 0.5 [SD] vs 25.1 seconds ± 20.1, respectively; P < .001; Cohen d = -1.08). The use of GPT-4 resulted in lower mean correction cost per report than the most cost-efficient radiologist ($0.03 ± 0.01 vs $0.42 ± 0.41; P < .001; Cohen d = -1.12). Conclusion The radiology report error detection rate of GPT-4 was comparable with that of radiologists, potentially reducing work hours and cost. © RSNA, 2024 See also the editorial by Forman in this issue.

Evaluation of prestyloid recess morphology and ulnar-sided contrast leakage in CT arthrography of the wrist.

BACKGROUND: In wrist arthrograms, aberrant contrast material is frequently seen extending into the soft tissue adjacent to the ulnar styloid process. Since the prestyloid recess can mimic contrast leakage in CT arthrography, this study aims to provide a detailed analysis of its morphologic variability, while investigating whether actual ulnar-sided leakage is associated with injuries of the triangular fibrocartilage complex (TFCC). METHODS: Eighty-six patients with positive wrist trauma history underwent multi-compartment CT arthrography (40 women, median age 44.5 years). Studies were reviewed by two board-certified radiologists, who documented the morphology of the prestyloid recess regarding size, opening type, shape and position, as well as the presence or absence of ulnar-sided contrast leakage. Correlations between leakage and the presence of TFCC injuries were assessed using the mean square contingency coefficient (rɸ). RESULTS: The most common configuration of the prestyloid recess included a narrow opening (73.26%; width 2.26 ± 1.43 mm), saccular shape (66.28%), and palmar position compared to the styloid process (55.81%). Its mean length and anterior-posterior diameter were 6.89 ± 2.36 and 5.05 ± 1.97 mm, respectively. Ulnar-sided contrast leakage was reported in 29 patients (33.72%) with a mean extent of 12.30 ± 5.31 mm. Leakage occurred more often in patients with ulnar-sided TFCC injuries (rɸ = 0.480; p < 0.001), whereas no association was found for lesions of the central articular disc (rɸ = 0.172; p = 0.111). CONCLUSIONS: Since ulnar-sided contrast leakage is more common in patients with peripheral TFCC injuries, distinction between an atypical configuration of the prestyloid recess and actual leakage is important in CT arthrography of the wrist.

Dose reduction potential in cone-beam CT imaging of upper extremity joints with a twin robotic x-ray system.

Cone-beam computed tomography is a powerful tool for 3D imaging of the appendicular skeleton, facilitating detailed visualization of bone microarchitecture. This study evaluated various combinations of acquisition and reconstruction parameters for the cone-beam CT mode of a twin robotic x-ray system in cadaveric wrist and elbow scans, aiming to define the best possible trade-off between image quality and radiation dose. Images were acquired with different combinations of tube voltage and tube current-time product, resulting in five scan protocols with varying volume CT dose indices: full-dose (FD; 17.4 mGy), low-dose (LD; 4.5 mGy), ultra-low-dose (ULD; 1.15 mGy), modulated low-dose (mLD; 0.6 mGy) and modulated ultra-low-dose (mULD; 0.29 mGy). Each set of projection data was reconstructed with three convolution kernels (very sharp [Ur77], sharp [Br69], intermediate [Br62]). Five radiologists subjectively assessed the image quality of cortical bone, cancellous bone and soft tissue using seven-point scales. Irrespective of the reconstruction kernel, overall image quality of every FD, LD and ULD scan was deemed suitable for diagnostic use in contrast to mLD (very sharp/sharp/intermediate: 60/55/70%) and mULD (0/3/5%). Superior depiction of cortical and cancellous bone was achieved in FDUr77 and LDUr77 examinations (p < 0.001) with LDUr77 scans also providing favorable bone visualization compared to FDBr69 and FDBr62 (p < 0.001). Fleiss' kappa was 0.618 (0.594-0.641; p < 0.001), indicating substantial interrater reliability. In this study, we demonstrate that considerable dose reduction can be realized while maintaining diagnostic image quality in upper extremity joint scans with the cone-beam CT mode of a twin robotic x-ray system. Application of sharper convolution kernels for image reconstruction facilitates superior display of bone microarchitecture.

Osteoarthritis of the Wrist: Pathology, Radiology, and Treatment.

Osteoarthritis (OA) is a degenerative disease that can manifest in any synovial joint under certain conditions. It leads to destruction of articular cartilage and adjacent bone, as well as formation of osteophytes at the edges of afflicted joint surfaces. Regarding the wrist, typical degenerative arthritis affects particular joints at a specific patient age, due to asymmetric load distribution and repetitive microtrauma. However, in the presence of instability or systemic diseases, early-onset degeneration can also impair the range of motion and grip strength in younger patients. Although advanced stages of OA display characteristic signs in radiography, the detection of early manifestations frequently requires computed tomography or magnetic resonance imaging (in some cases with additional arthrography). If a wrist becomes unstable, timely diagnosis and precise treatment are essential to prevent rapid disease progression. Therefore, close collaboration between radiologists and hand surgeons is obligatory to preserve the carpal function of patients.

Contrast-enhanced MRI of the wrist: Intravenous application of gadolinium improves diagnosis in ulnar-sided injuries of the TFCC.

PURPOSE: Although lesions of the triangular fibrocartilage complex (TFCC) frequently induce ulnar-sided wrist pain and potentially distal radioulnar joint instability, diagnosis can pose a challenge due to the intricate anatomy. This study aims to evaluate the benefits of contrast-enhanced sequences for the detection of TFCC injuries in magnetic resonance imaging of the wrist. METHOD: 94 patients underwent wrist MRI with intravenous application of gadolinium-based contrast agents. For each patient, two datasets were analysed independently by two board-certified radiologists: One set comprised only plain T1- and fat-saturated proton-density-weighted sequences, while the second dataset included contrast-enhanced T1-weighted images with fat suppression. Arthroscopy or clinical reports served as reference standard with the former being used whenever available. Diagnostic confidence and TFCC component assessability were subjectively evaluated. Contrast-to-noise ratios (CNR) were calculated serve as an objective indicator of image contrast. RESULTS: Lesions of the articular disc, the foveal and styloid ulnar attachment were present in 24 (25.5%), 61 (64.9%) and 53 (56.4%) patients. Access to contrast-enhanced T1 images improved the diagnostic accuracy for injuries of the styloid (R1/R2, 0.68/0.73 vs. 0.86/0.88) and foveal attachment (0.68/0.72 vs. 0.90/0.89) substantially compared to plain MRI (all p < 0.001), while no benefits could be identified for lesions of the central disc (0.89/0.90 vs. 0.87/0.90). Readers' diagnostic confidence and CNR for ulnar-sided lesions improved with contrast-enhanced T1 sequences available (p < 0.001). CONCLUSIONS: With superior CNR in lesions of the TFCC's foveal and styloid attachment, contrast-enhanced, fat-saturated T1-weighted sequences facilitate higher diagnostic accuracy and confidence than fat-saturated PD- and plain T1-weighted MRI.

Evaluation of Ultra-High-Resolution Cone-Beam CT Prototype of Twin Robotic Radiography System for Cadaveric Wrist Imaging.

RATIONALE AND OBJECTIVES: Cone-beam CT (CBCT) applications possess potential for dose reduction in musculoskeletal imaging. This study evaluates the ultra-high-resolution CBCT prototype of a twin robotic X-ray system in wrist examinations compared to high-resolution multidetector CT (MDCT). MATERIALS AND METHODS: Sixteen wrists of body donors were examined with the CBCT scan mode and a 384 slice MDCT system. Radiation-equivalent low-dose (CTDIvol(16cm)  = 3.3 mGy) and full-dose protocols (CTDIvol(16cm)  = 13.8 mGy) were used for both systems. Two observers assessed image quality on a seven-point Likert scale. In addition, software-assisted quantification of signal intensity fractions in cancellous bone was performed. Fewer pixels with intermediate signal intensity were considered to indicate superior depiction of bone microarchitecture. RESULTS: Subjective image quality in CBCT was superior to dose equivalent MDCT with p ≤ 0.03 for full-dose and p < 0.001 for low-dose scans, respectively. Median Likert values were 7/7 (reader 1 / reader 2) in full-dose CBCT, 6/6 in full-dose MDCT, 5/6 in low-dose CBCT and 3/3 in low-dose MDCT. Intraclass correlation coefficient was 0.936 (95% confidence interval, 0.897-0.961; p < 0.001), indicating excellent reliability. Objective analysis displayed smaller fractions of "indecisive" pixels with intermediate signal intensity for full-dose CBCT (0.57 [interquartile range 0.13]) compared to full-dose MDCT (0.68 [0.21]), low-dose CBCT (0.72 [0.19]), and low-dose MDCT (0.80 [0.15]) studies. No significant difference was observed between low-dose CBCT and full-dose MDCT. CONCLUSION: The new CBCT prototype provides superior image quality for trabecula and bone marrow in cadaveric wrist studies and enables dose reduction up to 75% compared to high-resolution MDCT.

Imaging of Carpal Instabilities.

BACKGROUND: The term "carpal instability" describes different debilitating wrist conditions, in which the carpus is unable to maintain its physiological range of motion and load transfer. Depending on the cause and location of the dysfunction, four groups can be defined: dissociative, non-dissociative, complex, and adaptive carpal instability. As the most common form by far, dissociative carpal instability can further be categorized as dorsal or palmar intercalated segment instability, contingent on the afflicted interosseous ligament. METHOD: This review article outlines the different entities of carpal instability, their pathophysiology, and their clinical presentation. It further discusses the diagnostic significance of different imaging methods as well as the established treatment options for each form of instability in context with the current literature. RESULTS AND CONCLUSION: Early detection and treatment of carpal instability are essential for preventing carpal osteoarthritis. Traumatic lesions of the scapholunate interosseous ligament are the most frequent cause of instability. They can occur in an isolated fashion or in context with other carpal injuries. While stress imaging and fluoroscopy facilitate the differentiation between dynamic and static forms of carpal instability, only MRI and CT/MR arthrography can directly reveal the extent of ligament discontinuity. KEY POINTS: · Carpal instability can manifest only in motion (dynamic) or at rest (static).. · Dissociative forms must be distinguished from non-dissociative, adaptive, or complex entities.. · Most instabilities are related to traumatic injuries or CPPD arthropathy.. · Fluoroscopy, stress imaging, and MR and CT arthrography are helpful for diagnosis.. CITATION FORMAT: · Grunz JP, Gietzen CH, Grunz K et al. Imaging of Carpal Instabilities. Fortschr Röntgenstr 2021; 193: 139 - 150.

Twin robotic x-ray system in small bone and joint trauma: impact of cone-beam computed tomography on treatment decisions.

OBJECTIVES: Trauma evaluation of extremities can be challenging in conventional radiography. A multi-use x-ray system with cone-beam CT (CBCT) option facilitates ancillary 3-D imaging without repositioning. We assessed the clinical value of CBCT scans by analyzing the influence of additional findings on therapy. METHODS: Ninety-two patients underwent radiography and subsequent CBCT imaging with the twin robotic scanner (76 wrist/hand/finger and 16 ankle/foot/toe trauma scans). Reports by on-call radiologists before and after CBCT were compared regarding fracture detection, joint affliction, comminuted injuries, and diagnostic confidence. An orthopedic surgeon recommended therapy based on reported findings. Surgical reports (N = 52) and clinical follow-up (N = 85) were used as reference standard. RESULTS: CBCT detected more fractures (83/64 of 85), joint involvements (69/53 of 71), and multi-fragment situations (68/50 of 70) than radiography (all p < 0.001). Six fractures suspected in radiographs were ruled out by CBCT. Treatment changes based on additional information from CBCT were recommended in 29 patients (31.5%). While agreement between advised therapy before CBCT and actual treatment was moderate (κ = 0.41 [95% confidence interval 0.35-0.47]; p < 0.001), agreement after CBCT was almost perfect (κ = 0.88 [0.83-0.93]; p < 0.001). Diagnostic confidence increased considerably for CBCT studies (p < 0.001). Median effective dose for CBCT was 4.3 µSv [3.3-5.3 µSv] compared to 0.2 µSv [0.1-0.2 µSv] for radiography. CONCLUSIONS: CBCT provides advantages for the evaluation of acute small bone and joint trauma by detecting and excluding extremity fractures and fracture-related findings more reliably than radiographs. Additional findings induced therapy change in one third of patients, suggesting substantial clinical impact. KEY POINTS: • With cone-beam CT, extremity fractures and fracture-related findings can be detected and ruled out more reliably than with conventional radiography. • Additional diagnostic information provided by cone-beam CT scans has substantial impact on therapy in small bone and joint trauma. • For distal extremity injury assessment, one-stop-shop imaging without repositioning is feasible with the twin robotic x-ray system.

3D cone-beam CT with a twin robotic x-ray system in elbow imaging: comparison of image quality to high-resolution multidetector CT.

BACKGROUND: Elbow imaging is challenging with conventional multidetector computed tomography (MDCT), while cone-beam CT (CBCT) provides superior options. We compared intra-individually CBCT versus MDCT image quality in cadaveric elbows. METHODS: A twin robotic x-ray system with new CBCT mode and a high-resolution clinical MDCT were compared in 16 cadaveric elbows. Both systems were operated with a dedicated low-dose (LD) protocol (equivalent volume CT dose index [CTDIvol(16 cm)] = 3.3 mGy) and a regular clinical scan dose (RD) protocol (CTDIvol(16 cm) = 13.8 mGy). Image quality was evaluated by two radiologists (R1 and R2) on a seven-point Likert scale, and estimation of signal intensity in cancellous bone was conducted. Wilcoxon signed-rank tests and intraclass correlation coefficient (ICC) statistics were used. RESULTS: The CBCT prototype provided superior subjective image quality compared to MDCT scans (for RD, p ≤ 0.004; for LD, p ≤ 0.001). Image quality was rated very good or excellent in 100% of the cases by both readers for RD CBCT, 100% (R1) and 93.8% (R2) for LD CBCT, 62.6% and 43.8% for RD MDCT, and 0.0% and 0.0% for LD MDCT. Single-measure ICC was 0.95 (95% confidence interval 0.91-0.97; p < 0.001). Software-based assessment supported subjective findings with less "undecided" pixels in CBCT than dose-equivalent MDCT (p < 0.001). No significant difference was found between LD CBCT and RD MDCT. CONCLUSIONS: In cadaveric elbow studies, the tested cone-beam CT prototype delivered superior image quality compared to high-end multidetector CT and showed a potential for considerable dose reduction.

The importance of radial multiplanar reconstructions for assessment of triangular fibrocartilage complex injury in CT arthrography of the wrist.

BACKGROUND: Triangular fibrocartilage complex (TFCC) lesions commonly cause ulnar-sided wrist pain and instability of the distal radioulnar joint. Due to its triangular shape, discontinuity of the TFCC is oftentimes difficult to visualize in radiological standard planes. Radial multiplanar reconstructions (MPR) may have the potential to simplify diagnosis in CT wrist arthrography. The objective of this study was to assess diagnostic advantages provided by radial MPR over standard planes for TFCC lesions in CT arthrography. METHODS: One hundred six patients (49 women, 57 men; mean age 44.2 ± 15.8 years) underwent CT imaging after wrist arthrography. Two radiologists (R1, R2) retrospectively analyzed three randomized datasets for each CT arthrography. One set contained axial, coronal and sagittal planes (MPRStandard), while the other two included an additional radial reconstruction with the rotating center either atop the ulnar styloid (MPRStyloid) or in the ulnar fovea (MPRFovea). Readers evaluated TFCC differentiability and condition. Suspected lesions were categorized using Palmer's and Atzei's classification and diagnostic confidence was stated on a five-point Likert scale. RESULTS: Compared to standard planes, differentiability of the superficial and deep TFCC layer was superior in radial reconstructions (R1/R2; MPRFovea: p < 0.001; MPRStyloid: p ≤ 0.007). Palmer and Atzei lesions were present in 86.8% (92/106) and 52.8% (56/106) of patients, respectively. Specificity, sensitivity and accuracy for central Palmer lesions did not differ in radial and standard MPR. For peripheral Atzei lesions, sensitivity (MPRStandard 78.6%/80.4%, MPRStyloid 94.6%/94.6%, MPRFovea 91.1%/89.3%) and accuracy (MPRStandard 86.8%/86.8%, MPRStyloid 96.2%/96.2%, MPRFovea 94.3%/93.4%) improved with additional styloid-centered (p = 0.004/0.008) and fovea-centered (p = 0.039/0.125) reconstructions. No substantial difference was observed between both radial MPR (p = 0.688/0.250). Interrater agreement was almost perfect for each dataset (κStandard = 0.876, κStyloid = 0.894, κFovea = 0.949). Diagnostic confidence increased with addition of either radial MPR (p < 0.001). CONCLUSIONS: Ancillary radial planes improve accuracy and diagnostic confidence for detection of peripheral TFCC lesions in CT arthrography of the wrist.

3D cone-beam CT of the ankle using a novel twin robotic X-ray system: Assessment of image quality and radiation dose.

PURPOSE: To evaluate image quality (IQ) and radiation dose in cone-beam computed tomography (CBCT) of the ankle using a novel twin robotic X-ray system. METHOD: We examined 16 cadaveric ankles with standard-dose (FD) and low-dose (LD) protocols using the new system's CBCT mode. For comparison, we performed multi-slice CT imaging (MSCT) with a clinical protocol. Three radiologists assessed IQ, noise and artifacts in bone and soft tissue on a five-point Likert scale (1= poor IQ; strong noise or artifacts; 5= excellent IQ; minimal noise or artifacts). Volume CT dose indices (CTDIvol) were calculated for radiation dose comparison between CBCT and MSCT. RESULTS: Overall IQ was described as very good or excellent by reader 1/2/3 in 62.5/87.5/56.3% of LD, 87.5/87.5/81.3% of FD and 100/87.5/87.5% of MSCT studies. Readers agreed that IQ was better in MSCT than LD (R1/R2/R3; p ≤ 0.008), two also found advantages of MSCT over FD (R1/R3; p ≤ 0.034). Soft tissue noise and artifacts were stronger in FD (all p ≤ 0.002) and LD (all p ≤ 0.001). In bone, artifacts and noise were also more severe in LD (all p < 0.001) and FD (all p ≤ 0.003). CTDIvol for clinical MSCT scans without dose modulation (15.0 ± 0.0 mGy) were higher than for FD (5.3 ± 1.0 mGy) and LD studies (2.9 ± 0.6 mGy; both p < 0.001). CONCLUSIONS: Despite MSCT providing better overall IQ than the twin robotic X-ray system's CBCT mode, both cone-beam protocols offer very good IQ in most studies and are suitable for clinical ankle imaging. Standard-dose and especially low-dose CBCT studies deliver up to five times less radiation dose than MSCT imaging.