Using Artificial Intelligence to Predict Intracranial Hypertension in Patients After Traumatic Brain Injury: A Systematic Review.

Intracranial hypertension (IH) is a key driver of secondary brain injury in patients with traumatic brain injury. Lowering intracranial pressure (ICP) as soon as IH occurs is important, but a preemptive approach would be more beneficial. We systematically reviewed the artificial intelligence (AI) models, variables, performances, risks of bias, and clinical machine learning (ML) readiness levels of IH prediction models using AI. We conducted a systematic search until 12-03-2023 in three databases. Only studies predicting IH or ICP in patients with traumatic brain injury with a validation of the AI model were included. We extracted type of AI model, prediction variables, model performance, validation type, and prediction window length. Risk of bias was assessed with the Prediction Model Risk of Bias Assessment Tool, and we determined the clinical ML readiness level. Eleven out of 399 nonduplicate publications were included. A gaussian processes model using ICP and mean arterial pressure was most common. The maximum reported area under the receiver operating characteristic curve was 0.94. Four studies conducted external validation, and one study a prospective clinical validation. The prediction window length preceding IH varied between 30 and 60 min. Most studies (73%) had high risk of bias. The highest clinical ML readiness level was 6 of 9, indicating "real-time model testing" stage in one study. Several IH prediction models using AI performed well, were externally validated, and appeared ready to be tested in the clinical workflow (clinical ML readiness level 5 of 9). A Gaussian processes model was most used, and ICP and mean arterial pressure were frequently used variables. However, most studies showed a high risk of bias. Our findings may help position AI for IH prediction on the path to ultimate clinical integration and thereby guide researchers plan and design future studies.

New method for quantification of the relative severity and (a)symmetry of isolated metopic synostosis.

Trigonocephaly is the result of premature fusion of the metopic suture and its severity can vary widely. However, there is no gold standard for quantification of the severity. This study was performed to quantify severity using the Utrecht Cranial Shape Quantifier (UCSQ) and to assess forehead symmetry. Nineteen preoperative non-syndromic trigonocephaly patients (age ≤1 year) were included for the analysis of severity and symmetry. Severity according to the UCSQ was based on the following combined variables: forehead width and relative skull elongation. The UCSQ was compared to the most established quantification methods. A high correlation was found between the UCSQ and visual score (r=0.71). Moderate to negligible correlation was found between the UCSQ and frontal angle, binocular distance, inter-ocular distance, and frontal stenosis. Additionally, correlation between the visual score and these established quantification methods was negligible. Assessment of the frontal peak (a)symmetry (ratio of right to left triangle area in the curve) showed a mean right versus left triangle area ratio of 1.4 (range 0.9-2.4). The results suggest that the UCSQ is appropriate for the quantification of severity based on the high correlation with clinical judgement. Furthermore, a larger triangle area right than left was unexpectedly found, indicating forehead asymmetry.

Imaging heterogeneity of peptide delivery and binding in solid tumors using SPECT imaging and MRI.

BACKGROUND: As model system, a solid-tumor patient-derived xenograft (PDX) model characterized by high peptide receptor expression and histological tissue homogeneity was used to study radiopeptide targeting. In this solid-tumor model, high tumor uptake of targeting peptides was expected. However, in vivo SPECT images showed substantial heterogeneous radioactivity accumulation despite homogenous receptor distribution in the tumor xenografts as assessed by in vitro autoradiography. We hypothesized that delivery of peptide to the tumor cells is dictated by adequate local tumor perfusion. To study this relationship, sequential SPECT/CT and MRI were performed to assess the role of vascular functionality in radiopeptide accumulation. METHODS: High-resolution SPECT and dynamic contrast-enhanced (DCE)-MRI were acquired in six mice bearing PC295 PDX tumors expressing the gastrin-releasing peptide (GRP) receptor. Two hours prior to SPECT imaging, animals received 25 MBq (111)In(DOTA-(ßAla)2-JMV594) (25 pmol). Images were acquired using multipinhole SPECT/CT. Directly after SPECT imaging, MR images were acquired on a 7.0-T dedicated animal scanner. DCE-MR images were quantified using semi-quantitative and quantitative models. The DCE-MR and SPECT images were spatially aligned to compute the correlations between radioactivity and DCE-MRI-derived parameters over the tumor. RESULTS: Whereas histology, in vitro autoradiography, and multiple-weighted MRI scans all showed homogenous tissue characteristics, both SPECT and DCE-MRI showed heterogeneous distribution patterns throughout the tumor. The average Spearman's correlation coefficient between SPECT and DCE-MRI ranged from 0.57 to 0.63 for the "exchange-related" DCE-MRI perfusion parameters. CONCLUSIONS: A positive correlation was shown between exchange-related DCE-MRI perfusion parameters and the amount of radioactivity accumulated as measured by SPECT, demonstrating that vascular function was an important aspect of radiopeptide distribution in solid tumors. The combined use of SPECT and MRI added crucial information on the perfusion efficiency versus radiopeptide uptake in solid tumors and showed that functional tumor characteristics varied locally even when the tissue appeared homogenous on current standard assessment techniques.

Optimization of combined temozolomide and peptide receptor radionuclide therapy (PRRT) in mice after multimodality molecular imaging studies.

BACKGROUND: Successful treatments of patients with somatostatin receptor (SSTR)-overexpressing neuroendocrine tumours (NET) comprise somatostatin-analogue lutetium-177-labelled octreotate ((177)Lu-TATE) treatment, also referred to as peptide receptor radionuclide therapy (PRRT), and temozolomide (TMZ) treatment. Their combination might result in additive effects. Using MRI and SPECT/CT, we studied tumour characteristics and therapeutic responses after different (combined) administration schemes in a murine tumour model in order to identify the optimal treatment schedule for PRRT plus TMZ. METHODS: We performed molecular imaging studies in mice bearing SSTR-expressing H69 (humane small cell lung cancer) tumours after single intravenous (i.v.) administration of 30 MBq (177)Lu-TATE or TMZ (oral 50 mg/kg daily for 14 days). Tumour perfusion was evaluated weekly by dynamic contrast-enhanced MRI (DCE-MRI), whereas tumour uptake of (111)In-octreotide was quantified using SPECT/CT until day 39 after treatment. Based on these results, seven different (177)Lu-octreotate and TMZ combination schemes were evaluated for therapy response, varying the order and time interval of the two therapies and compared with single treatments. RESULTS: PRRT and TMZ both resulted in tumour size reduction, accompanied by significant changes in MRI characteristics such as an enhanced tumour perfusion. Moreover, TMZ treatment also resulted in increased uptake of the SST analogue (111)In-octreotide until day 13. In the subsequent therapy study, 90 % of animals receiving (177)Lu-TATE at day 14 after TMZ treatment showed complete response, being the best anti-tumour results among groups. CONCLUSIONS: Molecular imaging studies indicated that PRRT after TMZ treatment could induce optimal therapeutic effects because of enhanced tumour uptake of radioactivity after TMZ, which was confirmed by therapy responses. Therefore, clinical translation of TMZ treatment prior to PRRT might increase tumour responses in NET patients as well.

Regional heterogeneity changes in DCE-MRI as response to isolated limb perfusion in experimental soft-tissue sarcomas.

Experimental evidence supports an association between heterogeneity in tumor perfusion and response to chemotherapy/radiotherapy, disease progression and malignancy. Therefore, changes in tumor perfusion may be used to assess early effects of tumor treatment. However, evaluating changes in tumor perfusion during treatment is complicated by extensive changes in tumor type, size, shape and appearance. Therefore, this study assesses the regional heterogeneity of tumors by dynamic contrast-enhanced MRI (DCE-MRI) and evaluates changes in response to isolated limb perfusion (ILP) with tumor necrosis factor alpha and melphalan. Data were acquired in an experimental cancer model, using a macromolecular contrast medium, albumin-(Gd-DTPA)45. Small fragments of BN 175 (a soft-tissue sarcoma) were implanted in eight brown Norway rats. MRI of five drug-treated and three sham-treated rats was performed at baseline and 1 h after ILP intervention. Properly co-registered baseline and follow-up DCE-MRI were used to estimate the volume transfer constant (K(trans) ) pharmacokinetic maps. The regional heterogeneity was estimated in 16 tumor sectors and presented in cumulative map-volume histograms. On average, ILP-treated tumors showed a decrease in regional heterogeneity on the histograms. This study shows that heterogenic changes in regional tumor perfusion, estimated using DCE-MRI pharmacokinetic maps, can be measured and used to assess the short-term effects of a potentially curative treatment on the tumor microvasculature in an experimental soft-tissue sarcoma model.

Heterogeneity in DCE-MRI parametric maps: a biomarker for treatment response?

This study aims to quantify the heterogeneity of tumour enhancement in dynamic contrast-enhanced MRI (DCE-MRI) using texture analysis methods. The suitability of the coherence and the fractal dimension to monitor tumour response was evaluated in 18 patients with limb sarcomas imaged by DCE-MRI pre- and post-treatment. According to the histopathology, tumours were classified into responders and non-responders. Pharmacokinetic (K(trans)) and heuristic model-based parametric maps (slope, max enhancement, AUC) were computed from the DCE-MRI data. A substantial correlation was found between the pharmacokinetic and heuristic model-based parametric maps: ρ = 0.56 for the slope, ρ = 0.44 for maximum enhancement, and ρ = 0.61 for AUC. From all four parametric maps, the enhancing fraction, and the heterogeneity features (i.e. coherence and fractal dimension) were determined. In terms of monitoring tumour response, using both pre- and post-treatment DCE-MRI, the enhancing fraction and the coherence showed significant differences between the response group and the non-response group (i.e. the highest sensitivity (91%) for K(trans), and the highest specificity (83%) for max enhancement). In terms of treatment prediction, using solely the pre-treatment DCE-MRI, the enhancing fraction and coherence discriminated between responders and non-responders. For prediction, the highest sensitivity (91%) was shared by K(trans), slope and max enhancement, and the highest specificity (71%) was achieved by K(trans). On average, tumours that responded showed a high enhancing fraction and high coherence on the pre-treatment scan. These results suggest that specific heterogeneity features, computed from both pharmacokinetic and heuristic model-based parametric maps, show potential as a biomarker for monitoring tumour response.

Dynamic contrast-enhanced MRI using macromolecular contrast media for monitoring the response to isolated limb perfusion in experimental soft-tissue sarcomas.

The objective of this study was to evaluate the potential of dynamic contrast-enhanced MRI for quantitative characterization of tumor microvessels and to assess the microvascular changes in response to isolated limb perfusion with TNF-alpha and melphalan. Dynamic contrast-enhanced MRI was performed in an experimental cancer model, using a macromolecular contrast medium, albumin-(Gd-DTPA)45. Small fragments of BN 175, a soft-tissue sarcoma, were implanted in 11 brown Norway (BN) rats. Animals were assigned randomly to a control (Haemaccel) or drug-treated group (TNF-alpha/melphalan). MRI was performed at baseline and 24 h after ILP. The transendothelial permeability (K(PS)) and the fractional plasma volume (fPV) were estimated from the kinetic analysis of MR data using a two-compartment bi-directional model. K(PS) and fPV decreased significantly in the drug-treated group compared to baseline (p<0.05). In addition, K(PS) post therapy was significantly lower (p<0.05) in the drug-treated group than in the control group. There was no significant difference in fPV between the drug-treated and the control group after therapy. Tumor microvascular changes in response to isolated limb perfusion can be determined after 24 h by dynamic contrast-enhanced MRI. The data obtained in this experimental model suggest possible applications in the clinical setting, using the appropriate MR contrast agents.

Texture analysis in radiographs: the influence of modulation transfer function and noise on the discriminative ability of texture features.

Tissue structures, represented by textures in radiographs, can be quantified using texture analysis methods. Different texture analysis methods have been used to discriminate between different aspects of various diseases in primarily x rays of chest, bone, and breasts. However, most of these methods have not specifically been developed for use on radiographs. Certain characteristics of the radiographic process, e.g., noise and blurring, influence the visible texture. In order for a texture analysis method to be able to discriminate between different underlying textures, it should not be too sensitive for such processes as image noise and blur. In this study, we investigated the sensitivity of four different texture analysis methods for image noise and blur. First, a baseline measurement was performed of the discriminative performance of the Spatial Gray-Level Dependence method, the Fourier Power Spectrum, the Fractal Dimension, and the Morphological Gradient Method on images, which were not affected by radiographic noise and blur. Two types of images were used: fractal and Brodatz. Whereas the Brodatz images represent very different textures, the differences between the fractal images are more gradual. We assume that the behavior of the different texture analysis methods on the fractal images is representative for their performance on radiologic textures. On these types of images we simulated the effect of four different noise levels and the effect of two different modulation transfer functions, corresponding with different screenfilm combinations. The influence on the discriminative performance of the four texture analysis methods was evaluated. The influence of noise on the discriminative performance is, as expected, dependent on the image type used; the discrimination of more gradually different images, such as the fractal images, is already lowered for relatively low noise levels. In contrast, when the images are more different, only high noise levels decrease the discriminative performance. The discriminative power of the Morphological Gradient Method is least affected by image blur. We conclude that the discriminative performance of the Morphological Gradient Method is superior to that of other methods in circumstances which mimic the conditions prevailing in radiographs.

Unraveling the role of structure and density in determining vertebral bone strength.

The strength of bone is determined not only by bone density but also by structure. Therefore, quantification of the structure in radiographs by texture parameters may result in a better prediction of fracture risk. Since in radiographs density and structure are strongly correlated, the predictive power of texture parameters should be corrected for the influence of BMD to determine the additional information conveyed by these parameters. In this study, we evaluated the predictive power of various texture parameters based on the Grey-Level Dependence Method and the Morphological Gradient Method. This study was performed on 67 vertebrae obtained from 20 male and 12 female human cadaver thoracolumbar spines. BMD and area of the vertebral body were determined from QCT images and texture parameters were derived from direct magnification (DIMA) radiographs. The fracture force, measured under conditions simulating the in vivo situation, was corrected with the area of the vertebra to yield the fracture stress (FS). Results of the study indicate that BMD correlates significantly with FS r = 0.82 (P < 0. 001, n = 24) and r = 0.94 (P < 0.001, n = 43) for female and male vertebrae, respectively. Correlation coefficients of the investigated texture parameters were as high as 0.80 (P < 0.001) and 0.67 (P < 0.001) for the female and male vertebrae, respectively. Multiple regression analysis showed that in female vertebrae, the addition of one texture parameter to BMD results in a better prediction of strength. The multiple correlation coefficient was 0. 87 (P < 0.001) in this case. In male vertebrae, BMD was the best predictor of fracture stress. These results suggest that texture parameters, as measured in magnification radiographs, can predict bone strength. Whereas in all cases BMD is the best single predictor of bone strength, for women texture parameters contain useful additional information.

Estimation of fractal dimension in radiographs.

In the last decade, the fractal dimension has become a popular parameter to characterize image textures. Also in radiographs, various procedures have been used to estimate the fractal dimension. However, certain characteristics of the radiographic process, e.g., noise and blurring, interfere with the straightforward application of these estimation methods. In this study, the influence of quantum noise and image blur on several estimation methods was quantified by simulating the effect of quantum noise and the effect of modulation transfer functions, corresponding with different screen-film combinations, on computer generated fractal images. The results are extrapolated to explain the effect of film-grain noise on fractal dimension estimation. The effect of noise is that, irrespective of the noise source, the fractal dimension is overestimated, especially for lower fractal dimensions. On the other hand, blurring results in an underestimation of the dimensions. The effect of blurring is dependent on the estimation method used; the dimension estimates by the power spectrum method are lowered with a constant value, whereas the underestimation by the methods working in the spatial domain is dependent on the given dimension. The influence of the MTF and noise on fractal dimension estimation seriously limits the comparability of fractal dimensions estimated from radiographs which differ in noise content or MTF. Only when the power spectrum method is used, it is possible to correct for the influence of different MTFs of screen-film combinations. It is concluded that only when using the same object-focus distance, the same exposure conditions, the same digitizer at the same resolution, can fractal dimensions as estimated in radiographs be reliably compared.