Radiomics-Based Machine Learning for Outcome Prediction in a Multicenter Phase II Study of Programmed Death-Ligand 1 Inhibition Immunotherapy for Glioblastoma.

BACKGROUND AND PURPOSE: Imaging assessment of an immunotherapy response in glioblastoma is challenging due to overlap in the appearance of treatment-related changes with tumor progression. Our purpose was to determine whether MR imaging radiomics-based machine learning can predict progression-free survival and overall survival in patients with glioblastoma on programmed death-ligand 1 inhibition immunotherapy. MATERIALS AND METHODS: Post hoc analysis was performed of a multicenter trial on the efficacy of durvalumab in glioblastoma (n = 113). Radiomics tumor features on pretreatment and first on-treatment time point MR imaging were extracted. The random survival forest algorithm was applied to clinical and radiomics features from pretreatment and first on-treatment MR imaging from a subset of trial sites (n = 60-74) to train a model to predict long overall survival and progression-free survival and was tested externally on data from the remaining sites (n = 29-43). Model performance was assessed using the concordance index and dynamic area under the curve from different time points. RESULTS: The mean age was 55.2 (SD, 11.5) years, and 69% of patients were male. Pretreatment MR imaging features had a poor predictive value for overall survival and progression-free survival (concordance index = 0.472-0.524). First on-treatment MR imaging features had high predictive value for overall survival (concordance index = 0.692-0.750) and progression-free survival (concordance index = 0.680-0.715). CONCLUSIONS: A radiomics-based machine learning model from first on-treatment MR imaging predicts survival in patients with glioblastoma on programmed death-ligand 1 inhibition immunotherapy.

Predicting response to cancer immunotherapy using noninvasive radiomic biomarkers.

INTRODUCTION: Immunotherapy is regarded as one of the major breakthroughs in cancer treatment. Despite its success, only a subset of patients responds-urging the quest for predictive biomarkers. We hypothesize that artificial intelligence (AI) algorithms can automatically quantify radiographic characteristics that are related to and may therefore act as noninvasive radiomic biomarkers for immunotherapy response. PATIENTS AND METHODS: In this study, we analyzed 1055 primary and metastatic lesions from 203 patients with advanced melanoma and non-small-cell lung cancer (NSCLC) undergoing anti-PD1 therapy. We carried out an AI-based characterization of each lesion on the pretreatment contrast-enhanced CT imaging data to develop and validate a noninvasive machine learning biomarker capable of distinguishing between immunotherapy responding and nonresponding. To define the biological basis of the radiographic biomarker, we carried out gene set enrichment analysis in an independent dataset of 262 NSCLC patients. RESULTS: The biomarker reached significant performance on NSCLC lesions (up to 0.83 AUC, P < 0.001) and borderline significant for melanoma lymph nodes (0.64 AUC, P = 0.05). Combining these lesion-wide predictions on a patient level, immunotherapy response could be predicted with an AUC of up to 0.76 for both cancer types (P < 0.001), resulting in a 1-year survival difference of 24% (P = 0.02). We found highly significant associations with pathways involved in mitosis, indicating a relationship between increased proliferative potential and preferential response to immunotherapy. CONCLUSIONS: These results indicate that radiographic characteristics of lesions on standard-of-care imaging may function as noninvasive biomarkers for response to immunotherapy, and may show utility for improved patient stratification in both neoadjuvant and palliative settings.

Ising model with conserved magnetization on the human connectome: Implications on the relation structure-function in wakefulness and anesthesia.

Dynamical models implemented on the large scale architecture of the human brain may shed light on how a function arises from the underlying structure. This is the case notably for simple abstract models, such as the Ising model. We compare the spin correlations of the Ising model and the empirical functional brain correlations, both at the single link level and at the modular level, and show that their match increases at the modular level in anesthesia, in line with recent results and theories. Moreover, we show that at the peak of the specific heat (the critical state), the spin correlations are minimally shaped by the underlying structural network, explaining how the best match between the structure and function is obtained at the onset of criticality, as previously observed. These findings confirm that brain dynamics under anesthesia shows a departure from criticality and could open the way to novel perspectives when the conserved magnetization is interpreted in terms of a homeostatic principle imposed to neural activity.

Quantification of internal carotid artery flow with digital subtraction angiography: validation of an optical flow approach with Doppler ultrasound.

BACKGROUND AND PURPOSE: Digital subtraction angiography is the reference standard technique to evaluate intracranial vascular anatomy and used on the endovascular treatment of vascular diseases. A dedicated optical flow-based algorithm was applied to DSA to measure arterial flow. The first quantification results of internal carotid artery flow validated with Doppler sonography are reported. MATERIALS AND METHODS: We included 22 consecutive patients who underwent endovascular procedures. To assess the sensitivity of the algorithm to contrast agent-blood mixing dynamics, we acquired high-frame DSA series (60 images/s) with different injection rates: 1.5 mL/s (n = 19), 2.0 mL/s (n = 18), and 3.0 mL/s (n = 13). 3D rotational angiography was used to extract the centerline of the vessel and the arterial section necessary for volume flow calculation. Optical flow was used to measure flow velocities in straight parts of the ICAs; these data were further compared with Doppler sonography data. DSA mean flow rates were linearly regressed on Doppler sonography measurements, and regression slope coefficient bias from value 1 was analyzed within the 95% confidence interval. RESULTS: DSA mean flow rates measured with the optical flow approach significantly matched Doppler sonography measurements (slope regression coefficient, b = 0.83 ± 0.19, P = .05) for injection rate = 2.0 mL/s and circulating volumetric blood flow <6 mL/s. For injection rate = 1.5 mL/s, volumetric blood flow <3 mL/s correlated well with Doppler sonography (b = 0.67 ± 0.33, P = .05). Injection rate = 3.0 mL/s failed to provide DSA-optical flow measurements correlating with Doppler sonography because of the lack of measurable pulsatility. CONCLUSIONS: A new model-free optical flow technique was tested reliably on the ICA. DSA-based blood flow velocity measurements were essentially validated with Doppler sonography whenever the conditions of measurable pulsatility were achieved (injection rates = 1.5 and 2.0 mL/s).

A DSA-based method using contrast-motion estimation for the assessment of the intra-aneurysmal flow changes induced by flow-diverter stents.

BACKGROUND AND PURPOSE: Flow-diverter stents modify intra-aneurysmal blood flow and induce the progressive thrombosis of intracranial aneurysms followed by stable vascular reconstruction. The aim of this study was to report a new method for the appraisal of intracranial blood flow from DSA performed during endovascular treatment procedures. MATERIALS AND METHODS: A cohort of 24 patients with unruptured IAs who underwent FDS implantation was prospectively recruited. Pre- and post-DSA sequences in combination with 3D rotational angiography were acquired. The quantification of arterial and intra-aneurysmal flow was accomplished by using an optical flow approach. Flow reduction was assessed by using a new metric termed the mean aneurysm flow amplitude ratio. The correlation between the MAFA ratio and the incidence of aneurysm thrombosis was assessed by using receiver operating characteristic analysis and the Fisher exact test when the optimum Youden index was found. RESULTS: The quantification of flow was successfully achieved in 21 of 24 patients (87.5%). On the imaging follow-up, 18 aneurysms developed complete thrombosis (87.5%) and 3 displayed residual circulation (12.5%). The threshold analysis of the MAFA ratio significantly predicted thrombosis at 12 months below a threshold of 1.03 (P=.035). There was no significant correlation between the time for complete occlusion of the aneurysm and contrast stagnation inside the aneurysm after treatment (P>.05). CONCLUSIONS: The MAFA ratio based on DSA flow quantification appears to be a reliable predictor for the assessment of stent treatment outcomes in this small study. These results open the door for perioperative flow quantification and provide indices that may help clinicians make appropriate intraprocedural decisions.

System identification theory in pharmacokinetic modeling of dynamic contrast-enhanced MRI: influence of contrast injection.

Optimization of experimental settings of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), like the contrast administration protocol, is of great importance for reliable quantification of the microcirculatory properties, such as the volume transfer-constant K(trans). Using system identification theory and computer simulations, the confounding effects of volume, rate and multiplicity of a contrast injection on the reliability of K(trans) estimation was assessed. A new tracer-distribution model (TDM), based on in vivo data from rectal cancer patients, served to describe the relationship between the contrast agent injection and the blood time-course. A pharmacokinetic model (PKM) was used to describe the relation between the blood and tumor tissue time-courses. By means of TDM and PKM in series, the tissue-transfer function of the PKM was analyzed. As both the TDM and PKM represented low-frequency-pass filters, the energy-density at low frequencies of the blood and tissue time-courses was larger than at high frequencies. The simulations, based on measurements in humans, predict that the K(trans) is most reliable with a high injection volume administered in a single injection, where high rates only modestly improve K(trans).

Elevated plasma chemokine CCL18/PARC in beta-thalassemia.

Plasma CCL18/PARC, a member of the CC chemokine family, has been found to be several ten-fold increased in symptomatic Gaucher type I patients. Elevated plasma chitotriosidase levels are a well-known abnormality in Gaucher patients, however, its diagnostic use is limited by the frequent genetic deficiency in the protein. Like the situation in Gaucher disease, lipids accumulate in macrophages of patients suffering from beta-thalassemia, and, in both conditions, increased chitotriosidase levels occur. We here report that plasma CCL18/PARC is also significantly increased in patients with beta-thalassemia major (range 76.8-4977.8, median=650.8 ng/ml, n=36 and control range 10-72, median=33 ng/ml n=36 respectively, P<0.001). The CCL18/PARC levels are lower than in Gaucher patients (range 174.8-10798.7, median 2538.2 ng/ml, n=28, P<0.001). In our cohort of beta-thalassemic patients, CCL18/PARC showed a significant negative correlation to iron chelation therapy and a significant positive correlation to ferritin and chitotriosidase levels, the latter only in the patients with the wild type genotype for the enzyme. Our study demonstrates that beta-thalassemic patients have increased CCL18/PARC levels that could be of value in monitoring iron overload and compliance to therapy.

Soluble CD163: a marker molecule for monocyte/macrophage activity in disease.

By immunoprecipitation we have identified a soluble plasma form of CD163 (sCD163), the IL-6 inducible macrophage-receptor for clearing haptoglobin-haemoglobin complexes. A sandwich ELISA for measuring sCD163 was established and used to determine the sCD163 levels in normal subjects and patients with inflammatory and myeloproliferative diseases. In normal subjects, the concentration of sCD163 was high (median 1.9 mg/l) with low intra-individual variation. Highly increased levels were seen in patients with sepsis, myeloid leukaemia and in patients with Gaucher disease characterized by accumulation of tissue macrophages. Although the physiological role of sCD163 remains unknown, our present data suggest that sCD163 might prove to be a valuable marker molecule in infectious and myeloproliferative diseases.

Dixon quantitative chemical shift imaging is a sensitive tool for the evaluation of bone marrow responses to individualized doses of enzyme supplementation therapy in type 1 Gaucher disease.

Type 1 Gaucher disease can be effectively treated with enzyme supplementation therapy. Bone disease is a debilitating feature of the disorder and results from infiltration of the bone marrow by Gaucher cells. The effect of treatment on bone marrow infiltration is difficult to measure, necessitating the development of sensitive techniques to allow adequate dosing. Dixon quantitative chemical shift imaging (Dixon-QCSI) is a MRI technique to measure displacement of fatty marrow by Gaucher cells. Low bone marrow fat fractions have been found in Gaucher disease. We studied the effect of individualized low doses of enzyme therapy on the fat fractions of the lumbar spine in 12 adult Gaucher disease patients before and during treatment and in 9 untreated Gaucher controls. Fat fractions were decreased in 9/12 patients (median 0.20, range 0.08-0.40) and equally low in the untreated Gaucher controls compared to age-matched healthy volunteers (normal values 0.27-0.43, P < 0.01). During treatment, fat fractions increased significantly already after 1 year in 11/12 patients (P = 0.007). After 4 to 5 years, fat fractions normalized in 11/12 patients. Fat fractions remained low in the untreated Gaucher controls (P = 0.5 and 0.6 at 1 and 2 years, respectively). Six of 11 patients had a dose increase, which did not clearly affect fat fractions. Dixon-QCSI is a sensitive tool for the measurement of the response of bone marrow to enzyme therapy.

The natural course of Gaucher disease in The Netherlands: implications for monitoring of disease manifestations.

This retrospective study in 20 untreated type I Gaucher disease patients shows that in Dutch patients clinical manifestations of Gaucher disease type I are progressive in the majority of patients, children as well as adults. This is in contrast with studies among Jewish patients. Our results emphasize the need for a regular follow-up to enable timely initiation of enzyme therapy.

3D Rotational Angiography: Recent Experience in the Evaluation of Cerebral Aneurysms for Treatment.

SUMMARY: Three dimensional (3D) reconstruction techniques providing volume rendered 3D images from rotational angiography data now exist.We report the design and early experience with one such system. 237 aneurysms were studied. Information was obtained on the morphology of the aneurysm itself and the vascular architecture in and around the aneurysm. 218 (92%) aneurysms went on to have endovascular treatment. The 3D images provided valuable information on aneurysmal anatomy, including relationships with the parent and adjacent vessels. This technique allowed fast and safe decision-making regarding the feasibility of endovascular or surgical treatment and provided useful information for performing the chosen treatment.

Intracellular transport of acid beta-glucosidase and lysosome-associated membrane proteins is affected in Gaucher's disease (G202R mutation).

Gaucher's disease (GD) is caused by an inherited deficiency of acid beta-glucosidase with storage of glucosylceramides in the lysosomes of macrophages. This study identifies a G202R mutation in the acid beta-glucosidase gene in an infant with severe neuronopathic (type 2) GD and only slightly reduced acid beta-glucosidase activity. Western blot analysis, pulse chase experiments, and the thin frozen section immunogold method were used to analyse the implications of this mutation on the pathogenesis, clinical heterogeneity and diagnostic evaluation of GD. The results show that acid beta-glucosidase persists in the patient's fibroblasts as a mannose-rich polypeptide in the endoplasmic reticulum and is not transported to the lysosomes. By contrast, high expression of the lysosome-associated membrane proteins LAMP-1 and LAMP-2, saposin C, and cathepsin D was observed in the patient's lysosomes. Immunogold labelling of the integral membrane proteins LAMP-1 and LAMP-2 increases significantly at the cell surface of Kupffer cells and fibroblasts as well as at the apical membrane of hepatocytes. In addition, LAMP-1 and LAMP-2 associate with the bilayer of stored glucosylceramide. It is concluded that defective intracellular transport of mutant acid beta-glucosidase from the endoplasmic reticulum to lysosomes leads to a more severe clinical phenotype than the residual enzyme activity may indicate. Furthermore, the detection of LAMP in the tubular bundles of undigested glucosylceramides, as well as their increased concentration at the surfaces of the affected cells, suggests that these proteins play a role in the storage or removal of substrate in GD. Intracellular targeting of acid beta-glucosidase and LAMP contributes to the broad phenotypic heterogeneity of GD.

Three-dimensional reconstruction of high contrast objects using C-arm image intensifier projection data.

The reconstruction of three-dimensional (3D) objects from 2D X-ray cone-beam projections using a circular source path is most commonly done with an algorithm according to Feldkamp et al. [Feldkamp LA, Davis LC, Kress JW. Practical cone-beam algorithms. J Opt Soc Am A 1984;6:612-619]. An adaptation of this so-called Feldkamp method to cone-beam projections acquired with a C-arm system is presented here. In a phantom study, reconstruction results obtained along real source-detector trajectories of a C-arm system are compared to reconstruction results obtained from projections acquired from a full-circular trajectory and from one consisting of two full orthogonal circles, which fulfills Tuy's sufficiency condition. The straightforward application of Feldkamp's method adapted to projection data obtained with a C-arm system illustrates the 3D imaging potential of image intensifier based cone-beam computed tomography. Reconstruction results from projection data of different patients acquired with a motorized C-arm system such as vessel structures filled with contrast agent and bones are presented.

Expression of mutated glucocerebrosidase alleles in human cells.

Gaucher disease is a heterogeneous disease characterized by impaired activity of the lysosomal enzyme glucocerebrosidase. This heterogeneity is attributed to a large number of mutations in the corresponding gene. In order to test the biochemical properties of some mutations prevalent among Israeli populations, the normal human glucocerebrosidase cDNA and cDNAs carrying mutations N370S, L444P, D409H, recTL, recNcil, P415R and 84GG were coupled to the T7 RNA polymerase promoter in a vaccinia virus-derived expression vector (pTM-1). Recombinant viruses were produced and used to infect human tissue culture cells. RNA and protein stability, recognition by anti-glucocerebrosidase monoclonal antibodies and intracellular enzymatic activity were measured. The results demonstrated that the D409H allele directed synthesis of cytoplasmic RNA with decreased stability compared with its normal counterpart or other mutated forms. The D409H and L444P mutated proteins had lower stability than that of their normal counterpart, while the recNcil-mutated protein was more stable. Only glucocerebrosidase forms harboring leucine at position 444 were recognized by the anti-glucocerebrosidase monoclonal antibodies used (8E4 and 2C7). Measurements of enzymatic activity of the recombinant proteins in cells loaded with a fluorescent glucosylceramide demonstrated that the N370S mutated enzyme had activity similar to that of the normal enzyme. The other mutated enzymes exhibited varying degrees of activities, generally corresponding to the phenotypes with which they are associated. The results presented demonstrate the use of the vaccinia virus-derived expression system and of loading living cells with fluorescent substrate as efficient tools for studying mutants in Gaucher disease and in other lysosomal diseases.

Overexpression of human glucocerebrosidase containing different-sized leaders.

Gaucher disease results from impaired activity of the lysosomal enzyme glucocerebrosidase. Aiming at overexpressing the human glucocerebrosidase and testing the efficiency of the two in-frame ATGs of its gene in directing synthesis of an active enzyme, it was coupled to the T7 RNA polymerase promoter in a vaccinia virus-derived expression vector (pTM-1). cDNAs containing either one or both ATGs of the glucocerebrosidase mRNA were linked to the T7 polymerase promoter. Recombinant viruses were produced and used for infecting human cells in tissue culture. The results demonstrated that both ATGs directed translation of active glucocerebrosidase, resulting in a 10-fold increase in enzymic activity. Most of the protein remained sensitive to endoglycosidase H. The active enzyme represented a small fraction of the expressed glucocerebrosidase. The recombinant enzyme had the same Km and optimal pH towards the artificial substrate 4-methylumbelliferyl glucopyranoside as the authentic endogenous human enzyme. Measurements of intracellular enzymic activity directed by the cDNAs with either one or both ATGs in cells loaded with a fluorescent glucosylceramide demonstrated a 30% increase in activity directed by the cDNAs containing the first ATG over that containing the second ATG. This indicates that the protein synthesized from the first ATG, with a 38 amino acid leader, is translocated through the endoplasmic reticulum more readily than its counterpart directed by the second ATG, with a 19 amino acid leader. The elevation in glucocerebrosidase activity and the reproducibility of the data leads us to propose the use of the vaccinia virus-derived expression system as a tool for studying glucocerebrosidase mutants in Gaucher disease.

HLA-DM and MHC class II molecules co-distribute with peptidase-containing lysosomal subcompartments.

MHC class II molecules associate with peptides in the endocytic pathway. Different endosomal locations for peptide loading of class II molecules, varying from early endosomes (EE) to lysosomes, have been assigned on the basis of subcellular fractionation experiments. We have determined the intracellular location of HLA-DM, a molecule that supports peptide loading of class II molecules, by separating vesicles from the melanoma cell line Mel JuSo on the basis of buoying density and surface charge. In both fractionations, HLA-DM co-fractionated with a lysosomal compartment containing beta-hexosaminidase (beta-hex) activity and not with endosomes. Further analysis showed that HLA-DM mainly co-fractionated with a sub-lysosomal structure characterized by a relative low density and containing both pro- and mature cathepsin D and MHC class II molecules. Fluid phase markers first enter this compartment before entering high-density lysosomes that contain exclusively mature cathepsin D, some HLA-DM and no detectable MC class II molecules. Finally we determined the intracellular location of neutral and acidic peptidases. Whereas neutral peptidase activity was detected in the endoplasmic reticulum and/or plasma membrane fractions, acidic peptidase activity exclusively migrated at the position of HLA-DM containing lysosomal vesicles. Our results show that class II molecules co-migrate with HLA-DM, pro- and mature cathepsin D, beta-hex and acidic peptidase activity. HLA-DM, cathepsin d and class II molecules were not observed at the position of EE. Our data suggest that HLA-DM-mediated peptide loading of class II molecules occurs in a lysosomal subcompartment.

Enhanced lysosomal acidification leads to increased chloroquine accumulation in CHO cells expressing the pfmdr1 gene.

Expression of the pfmdr1-encoded Pgh1 protein of Plasmodium falciparum in CHO cells confers a phenotype of increased sensitivity to chloroquine due to an increased Pgh1-mediated accumulation of this antimalarial. Pgh1 carrying amino acid substitutions associated with chloroquine resistance in P. falciparum does not confer this phenotype. Here, we present studies on the underlying mechanism of Pgh1 mediated chloroquine influx into CHO cells. First, we measured intralysosomal pH using FITC-labelled dextran and found the intralysosomal pH in Pgh1 expressing cells to be decreased. A decreased lysosomal pH was not observed in cells expressing Pgh1 carrying the S1034C and N1042D double substitution found in some chloroquine-resistant P. falciparum parasites. Secondly, Pgh1-mediated uptake of chloroquine was abolished in the presence of bafilomycin A1, a specific inhibitor of vacuolar [H+]ATPases and was nearly abrogated in the presence of NH4Cl. Finally, cells expressing wild-type Pgh1 showed increased uptake of both (+)- and (-)[3H]chloroquine enantiomers, indicating that Pgh1-mediated uptake of chloroquine is not enantioselective and in agreement with a pH-driven process. We conclude from these studies that Pgh1 does not transport chloroquine, but instead influences chloroquine accumulation by modulating the pH of acidic organelles. This function is abolished in Pgh1 carrying amino acid substitutions S1034C and N1042D. We speculate that the pfmdr1 gene encodes a vacuolar chloride channel.

Subcellular fractionation of cultured normal human melanocytes: new insights into the relationship of melanosomes with lysosomes and peroxisomes.

In order to obtain information on the disputed nature of melanosomes a comparison was made between the localization of melanosomal markers with those of other well-defined subcellular organelles such as lysosomes and peroxisomes. The distribution of marker enzymes was studied using two different density gradient systems, i.e., Percoll and Nycodenz. Furthermore, the subcellular localization of various types of antigens was analyzed using indirect immunofluorescence and immuno-electron microscopy. All methods revealed the existence of partial co-localization of melanosomal and lysosomal proteins and different localization of peroxisomal markers. The results suggest that melanosomes may share a common origin with lysosomal structures.

Type 1 Gaucher disease: molecular, biochemical, and clinical characterization of patients from northern Portugal.

We report the study of 16 catholic type 1 Gaucher disease patients originating from a well-defined region in the north of Portugal where a relatively high incidence is observed. The patients were screened for mutations: 3060G-->A, 5841A-->G, 5976C-->G, and 6433T-->C, which enabled the identification of 27 of the 32 mutated alleles. Four different genotypes were identified, namely 5841G/6433C (n = 6), 5841G/5841G (n = 5), 5841G/? (n = 4), and 6433C/? (n = 1). All but one of the patients carried at least one 5841G mutated allele, making its frequency 62.5%, which is similar to that described for Ashkenazi Jewish patients. The 5841G homozygotes presented an overall milder clinical profile, whereas no clear genotype/phenotype correlation could be established for heterozygous patients. On the basis of residual glucocerebrosidase activity, no distinction could be made between 5841G homozygotes and 5841G/6433C compound heterozygotes. Patients that had at least one 5841G allele (encoding the Ser 370 mutated enzyme) all presented a cell-type-specific residual glucocerebrosidase activity as well as an increased molecular activity when measured in the presence of the physiological activators.