ChatGPT, which can automatically generate written responses to queries using internet sources, soon went viral after its release at the end of 2022. The performance of ChatGPT on medical exams shows results near the passing threshold, making it comparable to third-year medical students. It can also write academic abstracts or reviews at an acceptable level. However, it is not clear how ChatGPT deals with harmful content, misinformation or plagiarism; therefore, authors using ChatGPT professionally for academic writing should be cautious. ChatGPT also has the potential to facilitate the interaction between healthcare providers and patients in various ways. However, sophisticated tasks such as understanding the human anatomy are still a limitation of ChatGPT. ChatGPT can simplify radiological reports, but the possibility of incorrect statements and missing medical information remain. Although ChatGPT has the potential to change medical practice, education and research, further improvements of this application are needed for regular use in medicine.

Many artificial intelligence (AI) algorithms have been developed for medical practice, but few have led to clinically used products. The recent hype of ChatGPT shows us that simple, user-friendly interfaces are one major factor in the applications’ popularity. The majority of AI-based applications in clinical practice are still far from simple-to-use applications with user-friendly interfaces. Therefore, simplifying operations is one key to AI-based medical applications’ success.

Background: Soft-tissue sarcomas (STS) are rare tumors of the soft tissue. Recent diagnostic studies on STS mainly dealt with only few cases of STS and did not investigate the post-therapeutic performance of MRI in a routine clinical setting. Therefore, we assessed the long-term diagnostic accuracy of MRI for detecting recurrent STS at a multidisciplinary sarcoma center.Methods: In all, 1055 postoperative follow-up MRIs of 204 patients were included in the study. MRI follow-up scans were systematically reviewed for diagnostic values (true-positive/−negative and false-positive/−negative results) in detecting recurrences. Pathological reports and follow-up MRIs were set as baseline references.Results: The median age of the patients was 55.3 ± 18.2 years. Of the patients, 34.8% presented with recurrences. Here, 65 follow-up scans were true positive, 23 false positive, 6 false negative, and 961 true negative. The overall sensitivity and specificity of MRI for detecting recurrences were 92 and 98%, respectively, with an accuracy of 97%. For intramuscular lesions and after surgery alone the sensitivity was higher (95 and 97%, respectively) than for subcutaneous lesions and surgery with additional radiation therapy (83 and 86%, respectively), at similarly high specificities (96–98%). The 6 false-negative results were found in streaky (n = 2) and small ovoid/nodular (n = 4) recurring lesions. The false-positive lesions imitated streaky (n = 14), ovoid/nodular (n = 8), and polycyclic/multilobulated recurring tumors (n = 1). All false-positive results were found in patients in whom the primary tumors were polycyclic/multilobulated in appearance.Conclusion: MRI shows a high diagnostic accuracy for detecting recurrent STS, with a high sensitivity and specificity. The diagnostic accuracy decreases in subcutaneous lesions and after surgery with radiation therapy, compared to intramuscular lesions and surgery alone. Radiologists should pay particular attention to streaky and small ovoid/nodular recurring lesions and patients with polycyclic/multilobulated primary tumors.

We developed and tested a neural network for automated detection and stability analysis of vertebral body fractures on computed tomography (CT). 257 patients who underwent CT were included in this Institutional Review Board (IRB) approved study. 463 fractured and 1883 non-fractured vertebral bodies were included, with 190 fractures unstable. Two readers identified vertebral body fractures and assessed their stability. A combination of a Hierarchical Convolutional Neural Network (hNet) and a fracture Classification Network (fNet) was used to build a neural network for the automated detection and stability analysis of vertebral body fractures on CT. Two final test settings were chosen: one with vertebral body levels C1/2 included and one where they were excluded. The mean age of the patients was 68 ± 14 years. 140 patients were female. The network showed a slightly higher diagnostic performance when excluding C1/2. Accordingly, the network was able to distinguish fractured and non-fractured vertebral bodies with a sensitivity of 75.8 % and a specificity of 80.3 %. Additionally, the network determined the stability of the vertebral bodies with a sensitivity of 88.4 % and a specificity of 80.3 %. The AUC was 87 % and 91 % for fracture detection and stability analysis, respectively. The sensitivity of our network in indicating the presence of at least one fracture / one unstable fracture within the whole spine achieved values of 78.7 % and 97.2 %, respectively, when excluding C1/2. The developed neural network can automatically detect vertebral body fractures and evaluate their stability concurrently with a high diagnostic performance.

Systematic reviews on the grading of STS using MRI are lacking. This review analyses the role of different MRI features in inferring the histological grade of STS. A systematic review was conducted and is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) checklist. The electronic databases of PubMed/MEDLINE were systematically searched for literature addressing the correlation of MRI findings in soft tissue sarcoma with tumor grade. As keywords "MRI", "magnetic resonance imaging", "sarcoma", "grade", "grading", and "FNCLCC" have been selected. 14 studies have been included in this systematic review. Tumor size (p = 0.015 (51 patients) to p = 0.81 (36 patients)), tumor margin (p < 0.001 (95 patients) to 0.93 (36 patients)), necrosis (p = 0.004 (50 patients) to p = 0.65 (95 patients)), peritumoral edema (p = 0.002 (130 patients) to p = 0.337 (40 patients)), contrast enhancement (p < 0.01 (50 patients) to 0.019 (51 patients)) and polycyclic/multilobulated tumor configuration (p = 0.008 (71 patients)) were significantly associated with STS malignancy grade in most of the included studies. Heterogeneity in T2w images (p = 0.003 (130 patients) to 0.202 (40 patients)), signal intensity in T1w images/ hemorrhage (p = 0.02 (130 patients) to 0.5 (31 patients)), peritumoral contrast enhancement (p < 0.001 (95 patients) to 0.253 (51 patients)) and tumoral diffusion restriction (p = 0.01 (51 patients) to 0.53 (52 patients)) were regarded as significantly associated with FNCLCC grade in some of the studies which investigated these features. Most other MRI features were not significant. Several MRI features, such as tumor size, necrosis, peritumoral edema, peritumoral contrast enhancement, intratumoral contrast enhancement, and polycyclic/multilobulated tumor configuration may indicate the malignancy grade of STS. However, further studies are needed to gain consensus.