Artificial Intelligence for Clinical Interpretation of Bedside Chest Radiographs.

Background Supine chest radiography for bedridden patients in intensive care units (ICUs) is one of the most frequently ordered imaging studies worldwide. Purpose To evaluate the diagnostic performance of a neural network-based model that is trained on structured semiquantitative radiologic reports of bedside chest radiographs. Materials and Methods For this retrospective single-center study, children and adults in the ICU of a university hospital who had been imaged using bedside chest radiography from January 2009 to December 2020 were reported by using a structured and itemized template. Ninety-eight radiologists rated the radiographs semiquantitatively for the severity of disease patterns. These data were used to train a neural network to identify cardiomegaly, pulmonary congestion, pleural effusion, pulmonary opacities, and atelectasis. A held-out internal test set (100 radiographs from 100 patients) that was assessed independently by an expert panel of six radiologists provided the ground truth. Individual assessments by each of these six radiologists, by two nonradiologist physicians in the ICU, and by the neural network were compared with the ground truth. Separately, the nonradiologist physicians assessed the images without and with preliminary readings provided by the neural network. The weighted Cohen κ coefficient was used to measure agreement between the readers and the ground truth. Results A total of 193 566 radiographs in 45 016 patients (mean age, 66 years ± 16 [SD]; 61% men) were included and divided into training (n = 122 294; 64%), validation (n = 31 243; 16%), and test (n = 40 029; 20%) sets. The neural network exhibited higher agreement with a majority vote of the expert panel (κ = 0.86) than each individual radiologist compared with the majority vote of the expert panel (κ = 0.81 to ≤0.84). When the neural network provided preliminary readings, the reports of the nonradiologist physicians improved considerably (aided vs unaided, κ = 0.87 vs 0.79, respectively; P < .001). Conclusion A neural network trained with structured semiquantitative bedside chest radiography reports allowed nonradiologist physicians improved interpretations compared with the consensus reading of expert radiologists. © RSNA, 2022 Supplemental material is available for this article. See also the editorial by Wielpütz in this issue.

Serum levels of soluble B and T lymphocyte attenuator predict overall survival in patients undergoing immune checkpoint inhibitor therapy for solid malignancies.

Therapy with immune checkpoint inhibitors (ICIs) can lead to durable tumor control in patients with various advanced stage malignancies. However, this is not the case for all patients, leading to an ongoing search for biomarkers predicting response and outcome to ICI. The B and T lymphocyte attenuator (BTLA) is an immune checkpoint expressed on immune cells that was shown to modulate therapeutic responses. Here, we evaluate circulating levels of its soluble form, soluble B and T lymphocyte attenuator (sBTLA), as a biomarker for the prediction of treatment response and outcome to ICI therapy. Serum levels of sBTLA were analyzed by multiplex immunoassay in n = 84 patients receiving ICI therapy for solid malignancies and 32 healthy controls. BTLA expression was evaluated on peripheral blood mononuclear cells in a subset of patients (n = 6) using multicolor flow cytometry. Baseline sBTLA serum levels were significantly higher in cancer patients compared to healthy controls. Importantly, circulating sBTLA levels were an independent prognostic factor for overall survival (OS). As such, patients with initial sBTLA levels above the calculated prognostic cutoff value (311.64 pg/mL) had a median OS of only 138 days compared to 526 for patients with sBTLA levels below this value (P = .001). Uni- and multivariate Cox regression analyses confirmed the prognostic role of sBTLA in the context of ICI therapy. Finally, we observed a significant correlation between sBTLA levels and the frequency of CD3 + CD8 + BTLA+ T cells in peripheral blood. Thus, our data suggest that circulating sBTLA could represent a noninvasive biomarker to predict outcome to ICI therapy, helping to select eligible therapy candidates.

Percutaneous transhepatic biliary drainage (PTBD) in patients with dilated vs. nondilated bile ducts: technical considerations and complications.

OBJECTIVES: The aim of this study was to compare success, technical complexity, and complication rates of percutaneous transhepatic biliary drainage (PTBD) in patients with dilated vs. nondilated bile ducts. METHODS: In a retrospective analysis, we evaluated all consecutive PTBD performed in our department over a period of 5 years. Technical success, technical data (side, fluoroscopy time, radiation dose, amount of contrast media, use of disposable equipment), procedure-related complications and peri-interventional mortality were compared for patients with dilated vs. non-dilated bile ducts. Independent t test and χ2 test were used to evaluate the statistical significance. RESULTS: A total of 253 procedures were performed on 187 patients, of whom 101/253 had dilated bile ducts and 152/253 not. In total, 243/253 procedures were successful. PTBD was significantly more often successful in patients with dilated vs. nondilated bile ducts (150/153 vs. 93/101; p 0.02). Overall complication rate (13%) did not differ significantly between patients with dilated vs. nondilated bile ducts. Procedures in patients with normal, nondilated bile ducts were associated with a significantly higher rate of post-interventional bleeding (5/101 vs. 0/152). Mean fluoroscopy time (42:36 ± 35:39 h vs. 30:28 ± 25:10 h; p 0.002) and amount of contrast media (66 ± 40 ml vs. 52 ± 24 ml; p 0.07) or use of disposables were significantly higher in patients with nondilated ducts. A significantly lower fluoroscopy time and amount of contrast medium were used in left hepatic PTBD. CONCLUSION: Despite the higher technical complexity, PTBD with nondilated bile ducts was associated with similar overall complication rates but higher bleeding complications compared with PTBD with dilated bile ducts. KEY POINTS: • PTBD was associated with similar overall complication rates in patients with dilated vs. nondilated bile ducts. • Although overall complication rates were low, PTBD in patients with nondilated bile ducts was associated with a higher incidence of post-interventional bleeding. • PTBD in patients with nondilated bile ducts is technically more complex.

Induction of Contralateral Hepatic Hypertrophy by Unilobar Yttrium-90 Transarterial Radioembolization versus Portal Vein Embolization: An Animal Study.

PURPOSE: To compare hepatic hypertrophy in the contralateral lobe achieved by unilobar transarterial radioembolization (TARE) versus portal vein embolization (PVE) in a swine model. METHODS: After an escalation study to determine the optimum dose to achieve hypertrophy after unilobar TARE in 4 animals, 16 pigs were treated by TARE (yttrium-90 resin microspheres) or PVE (lipiodol/n-butyl cyanoacrylate). Liver volume was calculated based on CT before treatment and during 6 months of follow-up. Independent t-test (P < .05) was used to compare hypertrophy. The relationship between hypertrophy after TARE and absorbed dose was calculated using the Pearson correlation. RESULTS: At 2 and 4 weeks after treatment, a significantly higher degree of future liver remnant hypertrophy was observed in the PVE group versus the TARE group, with a median volume gain of 31% (interquartile range [IQR]: 16%-66%) for PVE versus 23% (IQR: 6%-36%) for TARE after 2 weeks and 51% (IQR: 47%-69%) for PVE versus 29% (IQR: 20%-50%) for TARE after 4 weeks. After 3 and 6 months, hypertrophy converged without a statistically significant difference, with a volume gain of 103% (IQR: 86%-119%) for PVE versus 82% (IQR: 70%-96%) for TARE after 3 months and 115% (IQR: 70%-46%) for PVE versus 86% (IQR: 58%-111%) for TARE after 6 months. A strong correlation was observed between radiation dose (median 162 Gy, IQR: 139-175) and hypertrophy. CONCLUSIONS: PVE resulted in rapid hypertrophy within 1 month of the procedure, followed by a plateau, whereas TARE resulted in comparable hypertrophy by 3-6 months. TARE-induced hypertrophy correlated with radiation absorbed dose.

Insufficient future liver remnant and preoperative cholangitis predict perioperative outcome in perihilar cholangiocarcinoma.

BACKGROUND: Major liver resection has evolved as the mainstay of treatment for patients with perihilar cholangiocarcinoma (pCCA). Here we assessed the suitability of preoperative future liver remnant (FLR) measurement to predict perioperative complications, since surgical morbidity and mortality are high compared to other malignancies. METHODS: Between 2011 and 2016, 91 patients with pCCA underwent surgery in curative intent at our institution. The associations of surgical complications with FLR and clinico-pathological characteristics were assessed using logistic regression analyses. Different methods of FLR assessment, the calculated-FLR (cFLR; ratio of FLR to total liver volume), standardized FLR (sFLR; ratio of FLR to liver volume estimated by body surface area) and FLR to bodyweight ratio (FLR/BW) were tested for validity. RESULTS: Multivariable analysis identified preoperative cholangitis (Exp(B) = 0.236; p = 0.030) as the single significant predictor of postoperative mortality and cFLR (Exp(B) = 0.009, p = 0.004) as the single significant predictor of major postoperative morbidity (Clavien-Dindo ≥ 3b). Based on these findings we designed a futility criterion (cFLR<40% OR preoperative cholangitis) predicting in-house mortality. CONCLUSIONS: In patients with pCCA, the preoperative FLR<40% as well as preoperative cholangitis are two risk factors to independently predict perioperative morbidity and mortality. The cFLR should be the preferred method of liver volumetry.

Favorable COVID-19 course despite significant comorbidities in a ruxolitinib-treated patient with primary myelofibrosis.

COVID-19 carries a high risk of severe disease course, particularly in patients with comorbidities. Therapy of severe COVID-19 infection has relied on supportive intensive care measures. More specific approaches including drugs that limit the detrimental "cytokine storm", such as Janus-activated kinase (JAK) inhibitors, are being discussed. Here, we report a compelling case of a 55-yo patient with proven COVID-19 pneumonia, who was taking the JAK1/2 inhibitor ruxolitinib in-label for co-existing primary myelofibrosis for 15 months prior to coronavirus infection. The patient had significant comorbidities, including chronic kidney disease, arterial hypertension, and obesity, and our previous cohort suggested that he was thus at high risk for acute respiratory distress syndrome (ARDS) and death from COVID-19. Since abrupt discontinuation of ruxolitinib may cause fatal cytokine storm and ARDS, ruxolitinib treatment was continued and was well tolerated, and the patient´s condition remained stable, without the need for mechanical ventilation or vasopressors. The patient became negative for SARS-CoV-2 and was discharged home after 15 days. In conclusion, our report provides clinical evidence that ruxolitinib treatment is feasible and can be beneficial in patients with COVID-19 pneumonia, preventing cytokine storm and ARDS.

Salvage RFA in patients with intrahepatic recurrence after major hepatic surgery for colorectal cancer liver metastases: mid-term outcome.

OBJECTIVES: To evaluate the mid-term outcome of salvage radiofrequency ablation (RFA) treatment in patients who exhibited intrahepatic recurrence after major hepatectomy for colorectal liver metastases (CRCLM). METHODS: Observational study on 23 consecutive patients (mean age 59 ± 9 years; 14/9 male/female) who, after a single (11/23) or multiple rounds (12/23) of major hepatic surgery, developed recurrent CRCLM in the liver remnant. Patients with a maximum of three metastases measuring up to 3 cm, and without relevant extrahepatic disease, underwent CT-guided RFA. Using the Kaplan-Meier-method, median intrahepatic progression-free-survival (ihPFS) and overall survival (OS) times after salvage RFA were compared with the same patients' time between the respective last round of surgery and diagnosis of intrahepatic recurrence leading to RFA. RESULTS: Median follow-up was 26 months (range 12-103 months). Median ihPFS time after RFA was 8 months (range 1-81 months). Median ihPFS time after the respective last round of surgery and RFA in the same patients had been 5 months (range 1-23 months), thus yielding similar ihPFS times after surgery vs. after salvage RFA (p = 0.238; Mood's median test). After RFA, 15/23 (65%) of patients developed new hepatic metastases within the first year post-RFA. Median OS was 37 months, with a 1-year OS rate of 100%, 3-year OS rate of 57%, and 5-year OS rate of 24%. No major complications were observed. CONCLUSIONS: Patients who exhibit intrahepatic recurrence of CRCLM after major hepatectomy will experience intrahepatic recurrence after salvage RFA as well, and after similar time intervals. However, long-term ihPFS may still be achieved in some patients. KEY POINTS: • Patients who exhibit intrahepatic recurrence of colorectal liver metastases after major hepatectomy will experience intrahepatic recurrence after salvage RFA as well, and after similar time intervals. • About two-thirds of all patients develop new metastases elsewhere in the liver within 1 year after RFA. • However, long-term intrahepatic progression-free survival may still be achieved in some patients.

Predictors of Occlusion of Hepatic Blood Vessels after Irreversible Electroporation of Liver Tumors.

PURPOSE: To examine predictors of midterm occlusion in portal and hepatic veins within or adjacent to the ablation zone after irreversible electroporation (IRE) of liver tumors. MATERIALS AND METHODS: This retrospective cohort analysis included 39 patients who underwent CT-guided IRE of liver tumors. Vessels within or adjacent to the ablation zone were identified on CT images acquired immediately after the procedure, and the positional relationships with the ablation zone (within/adjacent), locations (proximal/distal), and diameters (< 4 mm or ≥ 4 mm) were evaluated. Using contrast-enhanced follow-up scans, each vessel was classified as patent, stenosed, or occluded. Associations between vessel occlusion and each variable were investigated. RESULTS: Overall, 33 portal veins and 64 hepatic veins were analyzed. Follow-up scans showed occlusion in 12/33 (36.7%) portal veins and 17/64 (26.6%) hepatic veins. Vessels within the ablation zone were occluded significantly more frequently than vessels adjacent to the ablation zone (portal: 55.6% [10/18] vs 13.3% [2/15], P = .04; hepatic: 45.4% [15/33] vs 6.4% [2/31], P = .011). Vessels with a diameter < 4 mm were also occluded significantly more frequently than vessels with a diameter ≥ 4 mm (portal: 72.7% [8/11] vs 18.1% [4/22], P = .011; hepatic: 54.8% [17/31] vs 0% [0/33], P < .001). The respective positive and negative predictive values for occlusion of vessels categorized as both within and < 4 mm were 88% (7/8) and 82% (20/25) for portal veins and 79% (15/19) and 96% (43/45) for hepatic veins. CONCLUSIONS: Midterm vessel occlusion after liver IRE could be predicted with relatively high accuracy by assessing ablation location and vessel diameter.

Single-needle electroporation and interstitial electrochemotherapy: in vivo safety and efficacy evaluation of a new system.

OBJECTIVES: We conducted an in vivo trial to investigate the safety and efficacy of a newly developed system for the application of a combined therapy consisting of irreversible electroporation (IRE) and electrochemotherapy (IRECT) in the liver. The system is conceived as a single-needle multitined applicator with expandable electrodes that allow interstitial injection of fluids, e.g., chemotherapy. METHODS: Experiments were conducted in ten domestic pigs. The applicator was placed in different liver lobes under CT guidance. In one lobe, the applicator was used for conventional IRE (1500 V, 120 pulses, pulse length 100 µs). In the other lobe, the same procedure was performed preceded by the injection of a doxorubicin mixture through the expandable electrodes (IRECT). Contrast-enhanced CT and MRI were performed on days 1, 3, and 7 after the procedure. Accordingly, three animals were sacrificed on days 1, 3, and 7 after the imaging and ablation volumes were evaluated histopathologically. Related t test was used to compare the groups. RESULTS: Technical success was achieved in 9/10 experiments. One animal deceased during the intervention because of ventricular fibrillation. Follow-up CT 1 and 3 days after intervention showed a significant (p < 0.05) difference in the ablation volumes of IRECT vs IRE, respectively, of 4.47 ± 1.78 ml vs 2.51 ± 0.93 ml and of 3.39 ± 1.05 vs 1.53 ± 0.78 ml. CONCLUSIONS: IRECT using the newly developed system proved to be effective and provided significantly larger ablation volumes compared with IRE alone. However, ECG triggering is a necessary prerequisite to allow a safe application of the system. KEY POINTS: • Working on the geometry of the IRE applicator in terms of expandable electrodes may overcome the current limitations of IRE resulting from the placement of multiple electrodes. • Efficacy of IRE ablations can be enhanced by the interstitial application of chemotherapy in the periphery of ablation areas.

Loco-regional hepatocellular carcinoma treatment services as a bridge to liver transplantation.

BACKGROUND: Liver transplantation remains the main curative treatment option for hepatocellular carcinoma (HCC) patients. In the Eurotransplant area Milan criteria are used to assign priority extra points (exceptional MELD, exMELD) for patients on the waiting list. To prevent patients from tumor progression, loco-regional (neoadjuvant) treatment (LRT) is used. For patients unlikely to timely receive an organ via primary allocation, "extended critera donor (ECD) organs" are used. The present study aimed to investigate the survival after LT with a strategy of minimizing waiting list dropouts by using LRT for bridging and transplanting ECD organs if possible and necessary. METHODS: Between October 2010 and May 2015, 50 liver transplants for HCC were included in this retrospective study. Of those, 42 (84%) met the Milan criteria according to the preoperative radiological examination. Forty-one patients (82%) received LRT. The waiting time was analyzed according to LRT. Kaplan-Meier curves with log-rank statistics were used for survival analyses. RESULTS: One- and five-year overall survival within Milan criteria was 94.3% and 83.7% compared with 91.7% and 67.9% beyond Milan criteria, though statistical significance was not reached (P = 0.487). LRT had no impact on overall survival (P = 0.629). Median waiting time was shorter if no LRT was performed (4.6 months vs. 1.5 months, P = 0.006) and there were no cases of waiting list dropouts. Using ECD organs had no impact on overall survival (P = 0.663). CONCLUSIONS: Patients with an expected waiting time to transplantation of >6 months could be successfully treated with LRT as a bridge to transplant. Overall and disease-free survival for patients within and beyond Milan criteria was comparable and the use of ECD organs in this cohort of HCC patients proved to be a safe option.

IL-6 and IL-8 Serum Levels Predict Tumor Response and Overall Survival after TACE for Primary and Secondary Hepatic Malignancies.

While surgical resection represents the standard potentially curative therapy for liver cancer, transarterial chemoembolization (TACE) has evolved as a standard therapy for intermediate-stage hepatocellular carcinoma (HCC) as well as liver metastases. However, it is still not fully understood which patients particularly benefit from TACE. Cytokines represent a broad category of signaling molecules that might reflect concomitant inflammation as an adverse prognostic factor. Here, we evaluated the role of interleukin (IL)-6, IL-8, and CC-chemokine ligand (CCL)22 as biomarkers in the context of TACE treatment. Cytokine serum levels were analyzed by multiplex immunoassay in 54 patients (HCC: n = 44, liver metastases: n = 10) undergoing TACE as well as 51 healthy controls. Patients with primary and secondary liver cancer showed significantly elevated levels of IL-6 and IL-8 but not CCL22 compared to healthy controls. Interestingly, low pre-interventional levels of IL-6 and IL-8 were predictors for an objective response after TACE in binary logistic regression. In contrast, patients with high pre-interventional IL-6 and IL-8 serum levels not only poorly responded to TACE but had a significantly impaired overall survival. Serum levels of IL-6 and IL-8 represent promising biomarkers for patients undergoing TACE and might help to pre-interventionally identify patients who particularly benefit from TACE regarding objective treatment response and overall survival.

Preoperative Embolization of the Celiac Axis or Common Hepatic Artery before Distal Pancreatectomy with Resection of the Celiac Axis.

Patients with locally advanced pancreatic cancer who undergo distal pancreatectomy with resection of the celiac axis (CA) are at risk for postoperative hepatic or gastric ischemia if collateral blood flow from the superior mesenteric artery (SMA) via the gastroduodenal artery is insufficient. This study presents a technique for preoperative angiographic evaluation of these collateral vessels by using an AMPLATZER Vascular Plug to temporarily occlude the CA or common hepatic artery while simultaneously performing digital subtraction angiography of the SMA. If collateral vessels are deemed sufficient, the plug can subsequently be released for permanent occlusion with the intent to enhance the blood flow in these collateral vessels.

Endovascular placement of an extraluminal arterial bypass graft - in vitro feasibility study.

PURPOSE: A novel approach for arterial bypass grafting using exclusively endovascular techniques was established in-vitro in a phantom model. MATERIAL AND METHODS: The experimental setting consisted of a gel-wax phantom with two embedded parallel fluid-filled silicon tubes simulating the superficial femoral vessels. Through an 8-French sheath, a re-entry catheter (OUTBACK®, Cordis) was placed in the simulated artery and used to puncture the vascular wall. Then a 0.014-inch guide wire was advanced into the extravascular space. With the curved needle of the re-entry catheter, the guide wire was steered on a course parallel to the vessel wall in the extravascular space for 5-10 cm. At the desired reentry site, the re-entry catheter was used to puncture the vascular wall again in order to regain access to the endovascular space. Once the tip of the guide wire had safely been placed in the vascular lumen, a self-expandable stent graft (VIABAHN® GORE®) was deployed to complete the extraluminal bypass. RESULTS: Endovascular placement of an extraluminal bypass was successfully achieved in 20 attempts. The mean duration of the procedure amounted to 14:58 (minutes: seconds) (SD ± 3:56). CONCLUSIONS: This in-vitro study suggests that endovascular placement of an extraluminal arterial bypass graft is technically feasible.

Targeted multiparametric magnetic resonance imaging/transrectal ultrasound-guided (mpMRI/TRUS) fusion prostate biopsy versus systematic random prostate biopsy: A comparative real-life study.

BACKGROUND: Patients with suspected prostate cancer usually undergo transrectal ultrasound-guided (TRUS) systematic biopsy, which can miss relevant prostate cancers and lead to overtreatment. AIMS: The aim of this study was to evaluate the detection rate for prostate cancer in MR-guided targeted biopsy (TB) and systematic biopsy (SB) in comparison with mpMRI of the prostate. METHODS AND RESULTS: Three hundred and eight men who underwent mpMRI due to elevated PSA values between 2015 and 2020 were studied at university hospital Aachen, Germany. MRI-images were divided into cohorts with suspicious findings (PI-RADS ≥ 3) and negative findings (PI-RADS < 3). In patients with PI-RADS ≥ 3 TB combined with SB was performed. A part of this group underwent RP subsequently. In patients with PI-RADS < 3 and clinical suspicion SB was performed. In the PI-RADS ≥ 3 group (n = 197), TB combined with SB was performed in 194 cases. Three cases were lost to follow-up. Biopsy yielded 143 positive biopsies and 51 cases without carcinoma. TB detected 71% (102/143) and SB 98% (140/143) of the overall 143 carcinoma. Overall, 102 carcinomas were detected by TB, hereof 66% (67/102) clinically significant (Gleason ≥ 3+4) and 34% (35/102) clinically insignificant carcinoma (Gleason 3+3). SB detected 140 carcinomas, hereof 64% (90/140) csPCA and 36% (50/140) nsPCA. Forty-one of the overall 143 detected carcinoma were only found by SB, hereof 46% (19/41) csPCA and 54% (22/41) nsPCA. Tumor locations overlapped in 44% (63/143) between TB and SB. In 25% (36/143), SB detected additional tumor foci outside the target lesions. 70/143 patients subsequently underwent RP. The detection of tumor foci was congruent between mpMRI and prostatectomy specimen in 79% (55/70) of cases. Tumor foci were mpMRI occult in 21% (15/70) of cases. In the group with negative mpMRI (n = 111), biopsy was performed in 81 cases. Gleason ≥ 3+4 carcinoma was detected in 7% and Gleason 3+3 in 24% cases. CONCLUSION: There was a notable number of cases in which SB detected tumor foci that were mpMRI occult and could have been missed by TB alone. Therefore, additional systematic random biopsy is still required. A supplemental random biopsy should be considered depending on the overall clinical suspicion in negative mpMRI.

Reduction of ADC bias in diffusion MRI with deep learning-based acceleration: A phantom validation study at 3.0 T.

PURPOSE: Further acceleration of DWI in diagnostic radiology is desired but challenging mainly due to low SNR in high b-value images and associated bias in quantitative ADC values. Deep learning-based reconstruction and denoising may provide a solution to address this challenge. METHODS: The effects of SNR reduction on ADC bias and variability were investigated using a commercial diffusion phantom and numerical simulations. In the phantom, performance of different reconstruction methods, including conventional parallel (SENSE) imaging, compressed sensing (C-SENSE), and compressed SENSE acceleration with an artificial intelligence deep learning-based technique (C-SENSE AI), was compared at different acceleration factors and flip angles using ROI-based analysis. ADC bias was assessed by Lin's Concordance correlation coefficient (CCC) followed by bootstrapping to calculate confidence intervals (CI). ADC random measurement error (RME) was assessed by the mean coefficient of variation (CV¯) and non-parametric statistical tests. RESULTS: The simulations predicted increasingly negative bias and loss of precision towards lower SNR. These effects were confirmed in phantom measurements of increasing acceleration, for which CCC decreased from 0.947 to 0.279 and CV¯ increased from 0.043 to 0.439, and of decreasing flip angle, for which CCC decreased from 0.990 to 0.063 and CV¯ increased from 0.037 to 0.508. At high acceleration and low flip angle, C-SENSE AI reconstruction yielded best denoised ADC maps. For the lowest investigated flip angle, CCC = {0.630, 0.771 and 0.987} and CV¯={0.508, 0.426 and 0.254} were obtained for {SENSE, C-SENSE, C-SENSE AI}, the improvement by C-SENSE AI being significant as compared to the other methods (CV: p = 0.033 for C-SENSE AI vs. C-SENSE and p < 0.001 for C-SENSE AI vs. SENSE; CCC: non-overlapping CI between reconstruction methods). For the highest investigated acceleration factor, CCC = {0.479,0.926,0.960} and CV¯={0.519,0.119,0.118} were found, confirming the reduction of bias and RME by C-SENSE AI as compared to C-SENSE (by trend) and to SENSE (CV: p < 0.001; CCC: non-overlapping CI). CONCLUSION: ADC bias and random measurement error in DWI at low SNR, typically associated with scan acceleration, can be effectively reduced by deep-learning based C-SENSE AI reconstruction.

Encrypted federated learning for secure decentralized collaboration in cancer image analysis.

Artificial intelligence (AI) has a multitude of applications in cancer research and oncology. However, the training of AI systems is impeded by the limited availability of large datasets due to data protection requirements and other regulatory obstacles. Federated and swarm learning represent possible solutions to this problem by collaboratively training AI models while avoiding data transfer. However, in these decentralized methods, weight updates are still transferred to the aggregation server for merging the models. This leaves the possibility for a breach of data privacy, for example by model inversion or membership inference attacks by untrusted servers. Somewhat-homomorphically-encrypted federated learning (SHEFL) is a solution to this problem because only encrypted weights are transferred, and model updates are performed in the encrypted space. Here, we demonstrate the first successful implementation of SHEFL in a range of clinically relevant tasks in cancer image analysis on multicentric datasets in radiology and histopathology. We show that SHEFL enables the training of AI models which outperform locally trained models and perform on par with models which are centrally trained. In the future, SHEFL can enable multiple institutions to co-train AI models without forsaking data governance and without ever transmitting any decryptable data to untrusted servers.

Denoising diffusion probabilistic models for 3D medical image generation.

Recent advances in computer vision have shown promising results in image generation. Diffusion probabilistic models have generated realistic images from textual input, as demonstrated by DALL-E 2, Imagen, and Stable Diffusion. However, their use in medicine, where imaging data typically comprises three-dimensional volumes, has not been systematically evaluated. Synthetic images may play a crucial role in privacy-preserving artificial intelligence and can also be used to augment small datasets. We show that diffusion probabilistic models can synthesize high-quality medical data for magnetic resonance imaging (MRI) and computed tomography (CT). For quantitative evaluation, two radiologists rated the quality of the synthesized images regarding "realistic image appearance", "anatomical correctness", and "consistency between slices". Furthermore, we demonstrate that synthetic images can be used in self-supervised pre-training and improve the performance of breast segmentation models when data is scarce (Dice scores, 0.91 [without synthetic data], 0.95 [with synthetic data]).

Reliability as a Precondition for Trust-Segmentation Reliability Analysis of Radiomic Features Improves Survival Prediction.

Machine learning results based on radiomic analysis are often not transferrable. A potential reason for this is the variability of radiomic features due to varying human made segmentations. Therefore, the aim of this study was to provide comprehensive inter-reader reliability analysis of radiomic features in five clinical image datasets and to assess the association of inter-reader reliability and survival prediction. In this study, we analyzed 4598 tumor segmentations in both computed tomography and magnetic resonance imaging data. We used a neural network to generate 100 additional segmentation outlines for each tumor and performed a reliability analysis of radiomic features. To prove clinical utility, we predicted patient survival based on all features and on the most reliable features. Survival prediction models for both computed tomography and magnetic resonance imaging datasets demonstrated less statistical spread and superior survival prediction when based on the most reliable features. Mean concordance indices were Cmean = 0.58 [most reliable] vs. Cmean = 0.56 [all] (p < 0.001, CT) and Cmean = 0.58 vs. Cmean = 0.57 (p = 0.23, MRI). Thus, preceding reliability analyses and selection of the most reliable radiomic features improves the underlying model's ability to predict patient survival across clinical imaging modalities and tumor entities.

Possible Alterations of Imaging Patterns in Computed Tomography for Delta-VOC of SARS-CoV-2.

BACKGROUND: So far, typical findings for COVID-19 in computed tomography (CT) have been described as bilateral, multifocal ground glass opacities (GGOs) and consolidations, as well as intralobular and interlobular septal thickening. On the contrary, round consolidations with the halo sign are considered uncommon and are typically found in fungal infections, such as invasive pulmonary aspergillosis. The authors recently observed several patients with COVID-19 pneumonia presenting with round, multifocal consolidations accompanied by a halo sign. As this may indicate alterations of CT morphology based on the virus variant, the aim of this study was to investigate this matter in more detail. METHODS: 161 CT scans of patients with confirmed SARS-CoV-2 infection (RT-PCR within 2 days of CT) examined between January 2021 and September 15, 2021 were included. Follow-up examinations, patients with invasive ventilation at the time of CT, and patients with insufficient virus typing for variants of concern (VOC) were excluded. CT scans were assessed for vertical and axial distribution of pulmonary patterns, degree of involvement, uni- vs. bilaterality, reticulations, and other common findings. The mean density of representative lesions was assessed in Hounsfield units. Results were compared using Mann-Whitney U-tests, Student's t-rests, descriptive statistics, and Fisher's exact tests. RESULTS: 75 patients did not meet the inclusion criteria. Therefore, 86/161 CT scans of unique patients were analyzed. PCR VOC testing confirmed manifestation of the Delta-VOC SARS-CoV-2 in 22 patients, 39 patients with Alpha-VOC and the remaining 25 patients with Non-VOC SARS-CoV-2 infections. Three patients with the Delta-VOC demonstrated multiple pulmonary masses or nodules with surrounding halo sign, whereas no patients with either Alpha-VOC (p = 0.043) or non-VOC (p = 0.095) demonstrated these findings. All three patients were admitted to normal wards and had no suspicion of a pulmonary co-infection. Patients with Delta-VOC were less likely to have ground glass opacities compared to Alpha-VOC (7/22 or 31.8 % vs. 4/39 or 10.3 %; p < 0.001), whereas a significant difference has not been observed between Delta-VOC and non-VOC (5/25 or 20 %; p = 0.348). The mean representative density of lesions did not show significant differences between the studied cohorts. CONCLUSION: In this study 3 out of 22 patients (13.6 %) with Delta-VOC presented with bilateral round pulmonary masses or nodules with surrounding halo signs, which has not been established as a notable imaging pattern in COVID-19 pneumonia yet. Compared to the other cohorts, a lesser percentage of patients with Delta-VOC presented with ground glass opacities. Based on these results Delta-VOC might cause a divergence in CT-morphologic phenotype. KEY POINTS: · Until recently, CT-morphologic signs of COVID-19 pneumonia have been presumed to be uncontroversially understood. Yet, recently the authors observed diverging pulmonary alterations in patients infected with Delta-VOC.. · These imaging alterations included round pulmonary masses or nodules with surrounding halo sign.. · These imaging alterations have not yet been established as typical for COVID-19 pneumonia, yet.. · Based on these results, Delta-VOC could impose a divergence of CT-morphologic phenotype.. CITATION FORMAT: · Yüksel C, Sähn M, Kleines M et al. Possible Alterations of Imaging Patterns in Computed Tomography for Delta-VOC of SARS-CoV-2 . Fortschr Röntgenstr 2022; 194: 1229 - 1241.

Automated major psoas muscle volumetry in computed tomography using machine learning algorithms.

PURPOSE: The psoas major muscle (PMM) volume serves as an opportunistic imaging marker in cross-sectional imaging datasets for various clinical applications. Since manual segmentation is time consuming, two different automated segmentation methods, a generative adversarial network architecture (GAN) and a multi-atlas segmentation (MAS), as well as a combined approach of both, were investigated in terms of accuracy of automated volumetrics in given CT datasets. MATERIALS AND METHODS: The bilateral PMM was manually segmented by a radiologist in 34 abdominal CT scans, resulting in 68 single 3D muscle segmentations as training data. Three different methods were tested for their ability to generate automated image segmentations: a GAN- and MAS-based approach and a combined approach of both methods (COM). Bilateral PMM volume (PMMV) was calculated in cm3 by each algorithm for every CT. Results were compared to the corresponding ground truth using the Dice similarity coefficient (DSC), Spearman's correlation coefficient and Wilcoxon signed-rank test. RESULTS: Mean PMMV was 239 ± 7.0 cm3 and 308 ± 9.6 cm3, 306 ± 9.5 cm3 and 243 ± 7.3 cm3 for the CNN, MAS and COM, respectively. Compared to the ground truth the CNN and MAS overestimated the PMMV significantly (+ 28.9% and + 28.0%, p < 0.001), while results of the COM were quite accurate (+ 0.7%, p = 0.33). Spearman's correlation coefficients were 0.38, 0.62 and 0.73, and the DSCs were 0.75 [95%CI: 0.56-0.88], 0.73 [95%CI: 0.54-0.85] and 0.82 [95%CI: 0.65-0.90] for the CNN, MAS and COM, respectively. CONCLUSION: The combined approach was able to efficiently exploit the advantages of both methods (GAN and MAS), resulting in a significantly higher accuracy in PMMV predictions compared to the isolated implementations of both methods. Even with the relatively small set of training data, the segmentation accuracy of this hybrid approach was relatively close to that of the radiologist.