ABSTRACT
Objectives: The aim of this study was to evaluate the diagnostic performance of nonspecialist readers with and without the use of an artificial intelligence (AI) support tool to detect traumatic fractures on radiographs of the appendicular skeleton. Methods: The design was a retrospective, fully crossed multi-reader, multi-case study on a balanced dataset of patients (≥2 years of age) with an AI tool as a diagnostic intervention. Fifteen readers assessed 340 radiographic exams, with and without the AI tool in 2 different sessions and the time spent was automatically recorded. Reference standard was established by 3 consultant radiologists. Sensitivity, specificity, and false positives per patient were calculated. Results: Patient-wise sensitivity increased from 72% to 80% (P < .05) and patient-wise specificity increased from 81% to 85% (P < .05) in exams aided by the AI tool compared to the unaided exams. The increase in sensitivity resulted in a relative reduction of missed fractures of 29%. The average rate of false positives per patient decreased from 0.16 to 0.14, corresponding to a relative reduction of 21%. There was no significant difference in average reading time spent per exam. The largest gain in fracture detection performance, with AI support, across all readers, was on nonobvious fractures with a significant increase in sensitivity of 11 percentage points (pp) (60%-71%). Conclusions: The diagnostic performance for detection of traumatic fractures on radiographs of the appendicular skeleton improved among nonspecialist readers tested AI fracture detection support tool showed an overall reader improvement in sensitivity and specificity when supported by an AI tool. Improvement was seen in both sensitivity and specificity without negatively affecting the interpretation time. Advances in knowledge: The division and analysis of obvious and nonobvious fractures are novel in AI reader comparison studies like this.
ABSTRACT
Introduction In this retrospective study, we aimed to evaluate benign internal mammary lymph nodes (IMLNs) in terms of frequency, number, size, long axis/short axis (L/S) ratio, intercostal location, presence of fatty hilum, and stability using breast magnetic resonance imaging (MRI) and discuss the findings by reviewing existing literature. Methods This single-center study consisted of 130 women between the ages of 24 and 76 years, who had at least two breast MRI examinations in our institution, with the latest exam performed between January 1, 2019 and November 1, 2019, were eligible for the study. MRIs of the study group were independently reviewed by two radiologists. Results IMLN was detected in 31.1% of the 427 MRIs, with a total number of 256 nodes. The most common indication for breast MRI was high-risk screening (66.2%). The median number of nodes per patient was 1 (range: 1-6). The median follow-up time was 19.5 months (range: 6-141 months). None of these benign nodes showed significant interval growth. Mean L/R ratio of the nodes was 1.9. One hundred and four nodes ( n = 104, 40.6%) had a L/S ratio less than 2 and 43.2% ( n = 45) of the nodes with a L/S ratio less than 2, had a long axis measuring less than or equal to 3mm. IMLN of patients with breast implants had the largest mean long axis. The fatty hilum was identified in 34.3% ( n = 68) of the 256 nodes. The size of the lymph nodes where fatty hilum was visualized was significantly larger than the ones where fatty hilum was not visualized ( p < 0.001). Fatty hilum could be visualized in only 2.7% of the nodes with a long axis smaller than 3 mm. Conclusion IMLN is a frequent finding on breast MRI. We have shown that benign IMLNs might be large sized in specific cases like patients with breast implants. When small sized (≤3mm), they are more likely to be rounded (L/S ratio <2). The fatty hilum that is a feature of benignity might not be visualized in nodes less than or equal to3mm.
ABSTRACT
OBJECTIVES: Breast implant-associated anaplastic large-cell lymphoma (BIA-ALCL) has been recognised in recent years, and there is extensive ongoing research. Although the exact mechanism and cause are still unclear, we now know that the disease is more associated with textured implants. To the best of our knowledge, no previous studies investigating the radiological differential of various implants have been conducted. In this essay, we aimed to demonstrate dicriminating in vitro and in vivo imaging features of variuos types of breast implant devices using mammography, ultrasound, and Magnetic Resonance Imaging (MRI). METHODS: Five different implant devices from various manufacturers with various surface textures, including smooth, micro-textured, regular macro-textured, lightweight macro-textured, and polyurethane- coated were used. In vitro mammography was performed with a digital mammogram (Amulet Innovality, Fuji, Japan), and in vitro and in vivo sonography were performed with Esaote MyLab9 using a 7.5 MHz linear probe. In vitro MRI was performed with a 1.5T magnet (Symphony TIM upgrade and Aera, Siemens Healthcare, Erlangen, Germany) with a 7-channel breast coil (Sense coil, Innova, Germany). MRI studies included fat sat T2 weighted sequences (T2WS), non-fat sat T2WS, and silicone only sequences. RESULTS: Each imaging technique had different contributions to dealing with this challenge. Mammography and MRI were limited to identifying the capsule's double bands. We could only differentiate the lightweight macro-textured implant on the mammogram as the borosilicate microspheres were represented by tiny, round lucencies within the gel. Ultrasound imaging with the proper technique was very helpful in identifying the surface. The inner capsule (implant shell) was identified as parallel double echogenic bands on the in vitro sonogram. Bands of the smooth implant were better delineated compared to the textured implants. The double echogenic bands of the polyurethane-coated implant were not even identified individually. The reverberation artifact caused by the smooth implant was the main discriminating in vivo sonographic feature of smooth implants. The hyperintense polyurethane-coated capsule was identified on fat-saturated T2WS and non-fat-saturated T2WS via in vitro MRI. The tiny hypointense microspheres of the lightweight implant were also identified on the silicone-only sequence of the in vitro MRI. CONCLUSION: In this study, we have shown that breast implant material and type may differ with the help of in vitro and in vivo imaging characteristics on different radiological modalities. These different imaging features could be used for recognising and labelling the implant type, especially macrotextured implants that are reported to be more associated with breast implant-associated anaplastic large-cell lymphoma (BIA-ALCL) compared to other types. We believe evaluating these imaging characteristics during daily practice will help radiologists become aware of the implant type and possible complications or diseases associated with that type.