Tumor classification of gastrointestinal liver metastases using CT-based radiomics and deep learning

OBJECTIVES

The goal of this study is to demonstrate the performance of radiomics and CNN-based classifiers in determining the primary origin of gastrointestinal liver metastases for visually indistinguishable lesions.

METHODS

In this retrospective, IRB-approved study, 31 pancreatic cancer patients with 861 lesions (median age [IQR]: 65.39 [56.87, 75.08], 48.4% male) and 47 colorectal cancer patients with 435 lesions (median age [IQR]: 65.79 [56.99, 74.62], 63.8% male) were enrolled. A pretrained nnU-Net performed automated segmentation of 1296 liver lesions. Radiomics features for each lesion were extracted using pyradiomics. The performance of several radiomics-based machine-learning classifiers was investigated for the lesions and compared to an image-based deep-learning approach using a DenseNet-121. The performance was evaluated by AUC/ROC analysis.

RESULTS

The radiomics-based K-nearest neighbor classifier showed the best performance on an independent test set with AUC values of 0.87 and an accuracy of 0.67. In comparison, the image-based DenseNet-121-classifier reached an AUC of 0.80 and an accuracy of 0.83.

CONCLUSIONS

CT-based radiomics and deep learning can distinguish the etiology of liver metastases from gastrointestinal primary tumors. Compared to deep learning, radiomics based models showed a varying generalizability in distinguishing liver metastases from colorectal cancer and pancreatic adenocarcinoma.

Unchaining Collective Intelligence for Science, Research, and Technology Development by Blockchain-Boosted Community Participation

Since its launch just over a decade ago by the cryptocurrency Bitcoin, the distributed ledger technology (DLT) blockchain has followed a breathtaking trajectory into manifold application spaces. This study aper analyses how key factors underpinning the success of this ground-breaking “Internet of value” technology, such as staking of collateral (“skin in the game”), competitive crowdsourcing, crowdfunding, and prediction markets, can be applied to substantially innovate the legacy organization of science, research, and technology development (RTD). Here, we elaborate a highly integrative, community-based strategy where a token-based crypto-economy supports finding best possible consensus, trust, and truth by adding unconventional elements known from reputation systems, betting, secondary markets, and social networking. These tokens support the holder’s formalized reputation and are used in liquid-democracy style governance and arbitration within projects or community-driven initiatives. This participatory research model serves as a solid basis for comprehensively leveraging collective intelligence by effectively incentivizing contributions from the crowd, such as intellectual property work, validation, assessment, infrastructure, education, assessment, governance, publication, and promotion of projects. On the analogy of its current blockbusters like peer-to-peer structured decentralized finance (“DeFi”), blockchain technology can seminally enhance the efficiency of science and RTD initiatives, even permitting to fully stage operations as a chiefless decentralized autonomous organization (DAOs).

Combined non-invasive assessment of endothelial shear stress and molecular imaging of inflammation for the prediction of inflamed plaque in hyperlipidaemic rabbit aortas

AIMS

To evaluate the incremental value of low endothelial shear stress (ESS) combined with high-resolution magnetic resonance imaging (MRI)- and computed tomography angiography (CTA)-based imaging for the prediction of inflamed plaque.

METHODS AND RESULTS

Twelve hereditary hyperlipidaemic rabbits underwent quantitative analysis of plaque in the thoracic aorta with 256-slice CTA and USPIO-enhanced (ultra-small superparamagnetic nanoparticles, P904) 1.5-T MRI at baseline and at 6-month follow-up. Computational fluid dynamics using CTA-based 3D reconstruction of thoracic aortas identified the ESS patterns in the convex and concave curvature subsegments of interest. Subsegments with low baseline ESS exhibited significant increase in wall thickness and plaque inflammation by MRI, in non-calcified plaque burden by CTA, and developed increased plaque size, lipid and inflammatory cell accumulation (high-risk plaque features) at follow-up by histopathology. Multiple regression analysis identified baseline ESS and inflammation by MRI to be independent predictors of plaque progression, while receiver operating curve analysis revealed baseline ESS alone or in combination with inflammation by MRI as the strongest predictor for augmented plaque burden and inflammation (low ESS at baseline: AUC = 0.84, P < 0.001; low ESS and inflammation by molecular MRI at baseline: AUC = 0.89, P < 0.001).

CONCLUSION

Low ESS predicts progression of plaque burden and inflammation as assessed by non-invasive USPIO-enhanced MRI. Combined non-invasive assessment of ESS and imaging of inflammation may serve to predict plaque with high-risk features.

Radiological Protection in Cone Beam Computed Tomography (CBCT). ICRP Publication 129

The objective of this publication is to provide guidance on radiological protection in the new technology of cone beam computed tomography (CBCT). Publications 87 and 102 dealt with patient dose management in computed tomography (CT) and multi-detector CT. The new applications of CBCT and the associated radiological protection issues are substantially different from those of conventional CT. The perception that CBCT involves lower doses was only true in initial applications. CBCT is now used widely by specialists who have little or no training in radiological protection. This publication provides recommendations on radiation dose management directed at different stakeholders, and covers principles of radiological protection, training, and quality assurance aspects. Advice on appropriate use of CBCT needs to be made widely available. Advice on optimisation of protection when using CBCT equipment needs to be strengthened, particularly with respect to the use of newer features of the equipment. Manufacturers should standardise radiation dose displays on CBCT equipment to assist users in optimisation of protection and comparisons of performance. Additional challenges to radiological protection are introduced when CBCT-capable equipment is used for both fluoroscopy and tomography during the same procedure. Standardised methods need to be established for tracking and reporting of patient radiation doses from these procedures. The recommendations provided in this publication may evolve in the future as CBCT equipment and applications evolve. As with previous ICRP publications, the Commission hopes that imaging professionals, medical physicists, and manufacturers will use the guidelines and recommendations provided in this publication for implementation of the Commission's principle of optimisation of protection of patients and medical workers, with the objective of keeping exposures as low as reasonably achievable, taking into account economic and societal factors, and consistent with achieving the necessary medical outcomes.

The retrobulbar sinus is superior to the lateral tail vein for the injection of contrast media in small animal cardiac imaging

Cardiac perfusion studies using computed tomography are a common tool in clinical practice. Recent technical advances and the availability of dedicated small animal scanners allow the transfer of these techniques to the preclinical sector in general and to mouse models of cardiac diseases in particular. This necessitates new requirements for contrast injection techniques as a rapid transport of contrast media from the intravenous access to the animal heart. Clinical contrast agents containing high iodine concentrations are used within small animal studies although they exhibit a high viscosity which might limit their transport within the vasculature. The authors provide a comparison of the transport of contrast media following an injection into the lateral tail vein and an injection into the retrobulbar sinus and discuss the anatomy involved. The temporal evolution of a contrast bolus and its in vivo distribution is visualized. It is demonstrated that injecting contrast agents into the lateral tail vein of mice results in a retrograde blood flow to the liver veins and therefore does not deliver a detectable contrast bolus to the heart, and thus it cannot be used for cardiac perfusion studies. By contrast, boli injected into the retrobulbar sinus are rapidly transported to the heart and provide ventricular contrast enabling perfusion studies similar to those in human patients. The results demonstrate that an injection into the retrobulbar sinus is superior to an injection into the lateral tail vein for the delivery of contrast boli to the animal heart, while all drawbacks of an injection into the lateral tail vein are overcome.

Anti-CD4-targeted gold nanoparticles induce specific contrast enhancement of peripheral lymph nodes in X-ray computed tomography of live mice

Antibody-conjugated gold nanoparticles have been applied as a biologically targeted contrast agent in live mice for one of the most widely used medical imaging methods, X-ray computed tomography. Such nanoprobes directed toward the CD4 receptor lead to distinctly enhanced X-ray contrast of peripheral lymph nodes. This study demonstrates the general feasibility of biologically specific X-ray imaging in living animals and discusses basic requirements for the use of nanoparticles as a targeted X-ray contrast agent.

Resistance to antiangiogenic therapy is directed by vascular phenotype, vessel stabilization, and maturation in malignant melanoma

Angiogenesis is not only dependent on endothelial cell invasion and proliferation, it also requires pericyte coverage of vascular sprouts for stabilization of vascular walls. Clinical efficacy of angiogenesis inhibitors targeting the vascular endothelial growth factor (VEGF) signaling pathway is still limited to date. We hypothesized that the level of vessel maturation is critically involved in the response to antiangiogenic therapies. To test this hypothesis, we evaluated the vascular network in spontaneously developing melanomas of MT/ret transgenic mice after using PTK787/ZK222584 for anti-VEGF therapy but also analyzed human melanoma metastases taken at clinical relapse in patients undergoing adjuvant treatment using bevacizumab. Both experimental settings showed that tumor vessels, which are resistant to anti-VEGF therapy, are characterized by enhanced vessel diameter and normalization of the vascular bed by coverage of mature pericytes and immunoreactivity for desmin, NG-2, platelet-derived growth factor receptor β, and the late-stage maturity marker α smooth muscle actin. Our findings emphasize that the level of mural cell differentiation and stabilization of the vascular wall significantly contribute to the response toward antiangiogenic therapy in melanoma. This study may be useful in paving the way toward a more rational development of second generation antiangiogenic combination therapies and in providing, for the first time, a murine model to study this.

A robot-guided minimally invasive approach for cochlear implant surgery: preliminary results of a temporal bone study

PURPOSE

The aim of this study was to create an access canal to the inner ear, by drilling, and perform the cochleostomy for cochlear implant surgery using robot guidance.

METHODS

A robot, a surgical drill and an Image-Guided Surgery (IGS) system were combined in a closed-loop setup. Ten temporal bones were scanned at the planning stages of the procedure. The robot guided the drill along the preplanned trajectory and created the approach. Postoperative scans were obtained.

RESULTS

The cochleostomy was performed completely in nine out of ten cases. This did not prove possible for one of the specimens, the target site selected being in too superficial a location in relation to the round window. No violation of the facial nerve took place, although the chorda tympani nerve was violated in one case and the stapes in two. It was obvious during preoperative planning that these structures would be violated, but this was accepted in order to maintain a safety margin from the facial nerve. No other unforeseen damage occurred.

CONCLUSIONS

This preliminary study suggests that robot-guided drilling of a minimally invasive approach to the cochlea might be feasible, but further improvements are necessary before any clinical application becomes possible. Where the width of the facial recess is less than 2.5 mm, the chorda tympani nerve and the ossicles are at risk.

Temporal bone imaging: comparison of flat panel volume CT and multisection CT

BACKGROUND AND PURPOSE

A recent development in radiology is the use of flat panel detectors in CT to obtain higher-resolution images. This technique is known as flat panel volume CT (fpVCT). We sought to compare the image quality and diagnostic value of 2 different flat panel detector-equipped scanners: one is a prototype fpVCT scanner, and the other is a so-called flat panel digital volume tomography (fpDVT) scanner, which is routinely used in clinical setup with current state-of-the-art multisection CT (MSCT) scanners.

MATERIALS AND METHODS

Five explanted temporal bones and 2 whole-head cadaveric specimens were scanned with fpVCT, fpDVT, and MSCT scanners. The image series were blindly evaluated by 3 trained observers who rated 38 anatomic structures with regard to their delineation/appearance.

RESULTS

Although the image quality obtained with fpVCT and fpDVT was rated significantly better compared with MSCT on isolated temporal bones, the differences were not significant when whole cadaveric heads were scanned.

CONCLUSIONS

Theoretic and practical advantages exist for flat panel detector-equipped scanners, including improved image quality. However, when imaging whole cadaveric heads, no significant difference could be demonstrated between them and standard-of-care MSCT.

Imaging anti-angiogenic treatment response with DCE-VCT, DCE-MRI and DWI in an animal model of breast cancer bone metastasis

As current classification systems for the assessment of treatment response in bone metastasis do not meet the needs of oncologists, new imaging biomarkers are desirable. Therefore, the diagnostic impact of dynamic contrast enhanced (DCE)-volumetric computed tomography (VCT) (descriptive analysis), DCE-MRI (two-compartment model) and diffusion weighted imaging (DWI) for monitoring anti-angiogenic therapy effects of the VEGF antibody bevacizumab in breast cancer bone metastases in rats was studied. Nude rats (n=8 animals treated with bevacizumab and n=9 untreated control rats) with site-specific osteolytic bone metastasis of the hind leg were imaged with a 1.5T clinical MRI-scanner in an animal coil as well as in a volumetric CT-scanner at days 30, 40, 50 and 60 after inoculation of MDA-MB-231 human breast cancer cells. From these data, osteolytic lesion size (OLS), peak enhancement (PE), area under the curve (AUC), amplitude (A), exchange rate constant (k(ep)) and apparent diffusion coefficient (ADC) were determined in bone metastases. Prior to changes in OLS (p< or =0.05 at days 50 and 60) there was already a significant decrease in PE, AUC and A (p< or =0.05 at days 40-60) in treated animals compared to controls. However, for k(ep) and ADC there were no significant differences between the groups at any time point (p>0.05 at days 40-60). In conclusion, anti-angiogenic treatment response in osteolytic breast cancer bone metastases can be assessed early with surrogate markers of vascularization, while DWI appears to be insensitive.

[Advances in the staging of renal cell carcinoma with high-resolution imaging]

Modern imaging modalities such as computed tomography (CT) and magnetic resonance imaging (MRI) allow high-resolution imaging of the abdomen. Modern scanners made high temporal as well as high spatial resolution available. Therapeutic approaches to the treatment of renal cell carcinoma have been improved over the recent years. Besides conventional and open laparoscopic tumor nephrectomy and nephron sparing, surgical approaches such as local tumor cryotherapy and radiofrequency ablation (RF) are ablative modalities and are used increasingly. Improved anesthesiological methods and new surgical approaches also allow curative treatment in extended tumors. Prerequisites for preoperative imaging modalities include visualization of the kidney tumor as well as its staging. Tumor-related infiltration of the renal pelvis or invasion of the perinephric fat and the renal hilus has to be excluded prior to nephron sparing surgery. In cases with extended tumors with infiltration of the inferior vena cava, it is necessary to visualize the exact extension of the tumor growth towards the right atrium in the vena cava. The radiologist should be informed about the diagnostic possibilities and limitations of the imaging modalities of CT and MRI in order to support the urologist in the planning and performance of surgical therapeutical approaches.

3D-Darstellung des Mittelohres mittels computergestützter Nachverarbeitung helikaler Mehrschicht-CT-Daten

Post-processing of CT-data allows non-invasive 3D-Visualisation of the middle ear for diagnosis and surgical planning. In this study different post-processing techniques and the clinical application of a 3D-postprocessing algorithm in a large number of patients are presented. 20 normal patients, 6 dissected temporal bones and 213 patients with suspected middle ear pathology were examined using a low-dosage Multi-Slice CT protocol. Virtual endoscopic views of the middle ear and 3D-images of the ossicles were generated using a standardised algorithm. Evaluation of the image quality was performed. The virtual views of the dissected temporal bones were compared to real views. In 32 patients high-quality 3D-models of the individual anatomical structures were generated and displayed using different visualisation techniques. The standardised and evaluated method enabled visualisation of the normal middle ear anatomy. Assessment of different pathologies, especially malformation, trauma, implants and postoperative alterations, was facilitated. The high-quality 3D-models allowed precise imaging of the anatomical structures. 3D-Visualisation of the middle ear using CT-data is beneficial for radiological diagnosis and surgical planning in cases of complex middle ear pathology as a complementary examination technique.

3D visualisation of the middle ear and adjacent structures using reconstructed multi-slice CT datasets, correlating 3D images and virtual endoscopy to the 2D cross-sectional images

The 3D imaging of the middle ear facilitates better understanding of the patient's anatomy. Cross-sectional slices, however, often allow a more accurate evaluation of anatomical structures, as some detail may be lost through post-processing. In order to demonstrate the advantages of combining both approaches, we performed computed tomography (CT) imaging in two normal and 15 different pathological cases, and the 3D models were correlated to the cross-sectional CT slices. Reconstructed CT datasets were acquired by multi-slice CT. Post-processing was performed using the in-house software "3D Slicer", applying thresholding and manual segmentation. 3D models of the individual anatomical structures were generated and displayed in different colours. The display of relevant anatomical and pathological structures was evaluated in the greyscale 2D slices, 3D images, and the 2D slices showing the segmented 2D anatomy in different colours for each structure. Correlating 2D slices to the 3D models and virtual endoscopy helps to combine the advantages of each method. As generating 3D models can be extremely time-consuming, this approach can be a clinically applicable way of gaining a 3D understanding of the patient's anatomy by using models as a reference. Furthermore, it can help radiologists and otolaryngologists evaluating the 2D slices by adding the correct 3D information that would otherwise have to be mentally integrated. The method can be applied to radiological diagnosis, surgical planning, and especially, to teaching.

Virtual endoscopy of the middle ear: experimental and clinical results of a standardised approach using multi-slice helical computed tomography

Virtual endoscopy (VE) enables non-invasive 3D endoluminal imaging of the middle ear by post-processing of CT data. To optimise the clinical application a standardised approach was evaluated in normal and pathologic cases. Data acquisition was performed using multi-slice helical CT in 20 normal patients and 15 patients with malformation or trauma. Virtual endoscopy of the tympanic cavity and 3D images of the ossicles were generated using surface and volume rendering. Qualitative assessment of the representation of anatomical structures was performed in normal patients. In 15 pathological cases the diagnostic benefit was evaluated by comparing the 3D images to the 2D images and intra-operative findings. In all 35 cases 3D imaging was possible using the standardised approach. The ossicular chain as well as the bony and soft tissue structures of the tympanic cavity were visualised in 20 normal patients. In 7 of 8 patients with malformation and 1 of 7 patients with trauma the original diagnosis was changed by 3D imaging. Standardisation and evaluation of the method in normal patients is essential as it enhances the diagnostic reliability. Virtual endoscopy facilitates understanding of the complex anatomy of the middle ear. In cases of suspected malformation and confirmed trauma it is helpful for diagnosis and surgical planning.

Expression of an Erwinia pectate lyase in three species of Aspergillus

Transgenic filamentous fungi of the species Aspergillus niger, A. nidulans and A. awamori expressing and secreting Erwinia carotovora subsp. atroseptica pectate lyase 3 (PL3) were generated. Correct processing of the pre-enzyme was achieved using the A. niger pectin lyase A (PEL A) signal peptide. With the prepro-peptide of A. niger polygalacturonase II, secreted enzymes still possessed the 6- aa pro-sequence, indicating the importance of the conformation of the precursor protein for correct cleavage of the signal sequence. PL3 expression was markedly increased in media optimized for limited protease activity, and reached 0.4, 0.8 and 2.0 mg/l for expression in A. niger, A. awamori and A. nidulans, respectively. Glycans attached to the PL3 enzymes exhibited species-specific differences, and an increase of molecular mass coincided with reduced specific activities of the enzymes.

Synergism between Erwinia pectate lyase isoenzymes that depolymerize both pectate and pectin

Phytopathogenic Erwinia bacteria cause tissue maceration by secretion of pectinolytic enzymes such as pectate lyase (PL). Sequencing of overlapping genomic fragments from Erwinia carotovora subsp. atroseptica established the organization of a 7.5 kbp region encoding PL isoenzymes. Two intergenic regions of 656 and 645 bp separate three enzyme coding regions of 1125 bp exhibiting approximately 80% positional identity. The promoters of each of the three genes contain a segment with high homology to the binding sequence of the E. chrysanthemi KdgR transcription repressor, implying similar mechanisms of gene regulation in the two bacterial species. Separate expression of the pel genes in the Escherichia coli-pT7-7 system and purification of their products yielded PLs at 7-33 mg (I culture)-1 with greater than 95% purity. Availability of the recombinant enzymes allowed determination of the kinetic differences amongst the PL isoforms, PL1, PL2 and PL3. The results show that PL is not strictly confined to depolymerization of pectate since each isoenzyme more readily degrades 31% esterified pectin. Addition of isoenzyme combinations revealed no synergism with respect to degradation of pectate or 31% esterified pectin. However, addition of enzyme combinations containing PL3 enhanced the activity towards 68% esterified pectin, against which individual PL activities were low, by up to 64%. These data suggest that the combination of PL isoenzymes extends the range of pectic substrates which the bacterium can degrade.