Simple Classification Research Articles

Indications of a changing winter through the lens of lake mixing in Earth’s largest freshwater system

Abstract Global surface freshwater primarily resides in lakes, with the overwhelming majority found in Earth’s largest lakes, thus understanding potential climate change effects in these large lakes is critical. In dimictic lakes, climate change has extended the duration of summer thermal stratification and reduced the length of the ice season. These changes are relatively straightforward to evaluate in smaller, inland lakes. However in large lakes, such as the North American Great Lakes, temporally intermittent and spatially heterogeneous ice cover, and spatial thermal heterogeneity limit the utility of simple ice on-off or mixing classifications; therefore, assessing how climate change is impacting winter conditions in large lakes is challenging. Here, we use in-situ and satellite-derived surface water temperature observations from the North American Great Lakes to overcome these limitations and show that warming air temperatures are driving reductions in the number of winter days, collectively those with either ice cover or inverse thermal stratification, in favor of increases in isothermal conditions for the period 1995 to 2023. We find that on average the Great Lakes are experiencing a loss of 14 winter days per decade. Our results demonstrate how climate change has yielded disproportionate changes in the annual thermal cycle and mixing conditions of Earth’s largest freshwater system and signals the potential for fundamental ecosystem shifts due to a loss of winter.

The CHD severity classification system: development of a tool to assist with disease stratification for CHD research.

Complexity stratification for CHD is an integral part of clinical research due to its heterogenous clinical presentation and outcomes. To support our ongoing research efforts into CHD requiring disease severity stratifications, a simplified CHD severity classification system was developed and verified, with potential utility for clinical researchers without specialist CHD knowledge or access to clinical/medical records. A two-tiered analysis approach was undertaken. First-tier analysis included the audit of a comprehensive system based on: i) timing of intervention, ii) cardiac morphology, and iii) cardiovascular physiology using real patient data (n = 30), across 10 common CHD lesions. Second-tier analysis allowed for a simplified version of the classification system using morphology as a stand-alone predictor. Twelve clinicians of varying specialities involved in CHD care ranked 10 common lesions from least to most severe based on typical presentation and clinical course. First-tier analysis identified that cardiac morphology was the principal driver of complexity. Second-tier analysis largely confirmed the ranking and classification of the lesions into the broad CHD severity groups, although some variation was noted, specifically among non-cardiac specialists. This simplified version of the classicisation system, with morphology as a stand-alone predictor of severity, allowed for effective stratification for the purposes of analysis. The findings presented here support this comprehensive and simple CHD severity classification system with broad utility in CHD research, particularly among clinicians and researchers with limited knowledge of CHD. The model may be applied to produce locally relevant research tools.

Optimization of a Hydrogen-Fueled Parametric Strut Injector Using an Automated Workflow Computational Fluid Dynamics Method

Abstract A strut injector for burning hydrogen has been optimized using an automated workflow system. In order to choose the optimal set of injectors to cover the design space, an optimized design of experiments (DOE) method was used to automatically choose the parameters that best spanned the design space. One hundred candidate designs were chosen, and a script was used to generate a series of stereolithography (STL) files for each design. The STL files were then uploaded to a supercomputer for computational fluid dynamics analysis. For each of the 100 designs, a four step process was followed to generate the required data, and this included an automated mesh generation step, a field initialization step, a mesh adaptation step, and finally an large eddy simulation all within the yales2 numerical framework. In order to reduce the time required for postprocessing and the amount of data required, the simulations relied heavily on an on-the-fly postprocessing methodology, which reduced the complex time-unsteady flow fields to a small number of quantified outputs of interest that measured the suitability of each design such as the pressure drop across the injector and the efficiency of the mixing process. At the conclusion of these simulations, automated scripts translated these outputs into a smaller set of parameters that could be used to compare each design and allow subsequent optimization and surrogate modeling. Several surrogate modeling methods were attempted with mixed results; however, a simple classification methodology quickly identified the parameters of interest.

Simple fuzzy classification of metal qualities of Ona River, Ibadan, Nigeria, and its implication for fish production and other uses

Elevated concentrations of metals in rivers can pose significant risks to human health. Consequently, it is essential to monitor metal levels in river water to minimize human exposure, particularly through consumption of harvested fish and other uses. This study examined the water quality of the Ona River in Ibadan, Nigeria. Seven potentially toxic elements (PTEs), namely Fe, Cu, Mn, Zn, Pb, Cd, Cr and four macronutrients: Mg, Na, K, Ca were investigated. Surface and pore water samples were collected from six sampling points during the peak of the dry and the rainy seasons. These were analyzed according to standard procedures and processed using the Simple Fuzzy Classification (SFC) model with four classes. The results revealed that the PTEs: Fe, Pb, Cd, Cr, and Mn were detected above guideline limits in several sampling points, indicating possible contamination from anthropogenic sources. Simple fuzzy classification puts the quality of water in the investigated portion of the river in the range of 67.4–88.5% in the C1 (pristine) category during the dry season and 61.9–88.6% in C1 (pristine) during rainy season. However, there were membership functions of up to 58.7 and 63.2; 61.2 and 75.4 as well as 44.3 and 43.5% in C2, C3, and C4 categories (contaminated, polluted and extremely polluted, respectively) during the first and second sampling, respectively. Due to the high concentrations of toxic and bioaccumulative metals in the river, efforts should be made to control the discharge of pollutants into the river. These will prevent human health exposure risks tendencies for various uses, including fish production.

Proposal of a revised 3-tier "2mm" risk classification model for postoperative pancreatic fistula in robotic pancreatoduodenectomy.

The recent International Study Group for Pancreatic Surgery (ISGPS) risk classification for postoperativepancreatic fistula (grade B/C) was developed based on data from open and mixed minimally invasive pancreatoduodenectomy. The ISGPS risk classification model has not been validated specifically for POPF after robotic pancreatoduodenectomy (RPD). We calculated the rate of POPF (ISGPS 2016 definition, grade B/C) by analyzing consecutive patients after RPD by surgeons after their learning curves (80 RPDs per surgeon). The validation of the ISGPS 4-tier and the simplified 3-tier risk classification was conducted using the area under the receiver operating curve (AUC). From 2019 to 2023, 187 patients after RPD were included. Neither the ISGPS 4-tier nor the simplified 3-tier classification model showed robust discrimination (AUC: 0.696 and 0.685, respectively). Moreover, both risk classifications failed to differentiate the rates of POPF and major complications among subgroups. Multivariate analysis suggested that soft pancreatic texture and pancreatic duct ≤ 2mm were independent risk factors for POPF after RPD. After adjusting the duct size's cutoff from 3 to 2mm, the revised 4-tier "2mm" classification model showed no significant difference between risk categories B and C (6.7% vs. 9.4%, P = 0.063). The revised 3-tier "2mm" classification model stratified patients into A (n = 54), B (n = 68), and C (n = 65) groups, with corresponding POPF rates of 0.0%, 8.8%, and 23.1% (P < 0.001), and major complication rates of 5.6, 14.7, and 24.6% (P = 0.014), respectively. Compared to the simplified 3-tier classification model, the revised 3-tier "2mm" classification model showed improved discrimination (AUC: 0.753 vs. 0.685, P = 0.034) and clinical utility. The current ISGPS 4-tier and the simplified 3-tier classification models lacked sufficient discrimination in patients after RPD. We propose a revised 3-tier "2mm" risk classification model for RPD with a robust discrimination, which requires further international validation with prospectively obtained data.

A sound recognition algorithm to filter non construction site noise

Sixense Engineering have developed a new algorithm to correct the noise levels measured around construction sites, by identifying and filtering out sounds that are generated by activities outside the site (emergency services sirens, traffic horns, etc.). The idea is to go beyond the simple classification of audio files by applying the classification algorithm to all the noise events that contribute to the noise dose LAeq,T. In this innovative process, the noise dose is calculated from elementary LAeq,1min,. When noise events are detected (in terms of LAeq,1min), the corresponding audio recording (.wav file) are analysed by the sound recognition algorithm. When the algorithm identifies that the noise source is not construction related, with a high degree of confidence, the corresponding LAeq,1min is excluded from the noise dose calculation. In other words, the measured noise dose can be filtered by removing non construction related noise events produced by external sources like sirens, horns, motorbikes, birds, music, etc.

Detailed biogeographical mapping as a useful novel tool for the conservation of endemic taxa: a case of study for Iberian orchids

Background Iberian orchids occur in plant communities designated as habitats of interest in the European Union, particularly in Mediterranean forests. Their ecological importance highlights the need for a deeper understanding of their distribution and ecological requirements in order to develop effective conservation and management policies. Methods This study focuses on the biogeographical mapping and characterisation of five Iberian and Balearic endemic orchid species at district level. Distribution data were collected from existing biodiversity databases and integrated into a Geographic Information System (GIS). In order to assign the correct biogeographical epithet to each taxon, a set of rules and criteria was developed to ensure an objective, simple and universal classification based on the species' distribution areas. Results For each orchid species, the study identified its phytosociological assignment, bioclimatic range and current conservation status. Detailed maps were produced, providing insights into the biogeographical, bioclimatic and phytosociological attributes of these taxa. Conclusions The biogeographical mapping and ecological classification presented in this study provide a basis for effective decision-making regarding the conservation and management of these orchid species. In addition, the results can be used to update their conservation status to better reflect their ecological needs and threats.

Effect of Different Injury Morphology of the Endplate on Intervertebral Disc Degeneration: Retrospective Cohort Study.

To describe a simplified classification scheme for endplate injury morphology based on 3D CT and to examine possible associations between endplate injury morphology and vertebral space and other variables such as type of fracture and disc degeneration in a group of patients with thoracolumbar fractures. This study was a retrospective cohort study. We collected patients with thoracolumbar fractures admitted from January 2015 to August 2020 and divided them into three groups based on the morphology of endplate injury (45 cases of mild endplate injury, 54 cases of moderate endplate injury, and 42 cases of severe endplate injury, SEI). Data of vertebral body and intervertebral space height and angle, the Pfirrmann grade, endplate healing morphology were collected during preoperative, postoperative, and long-term follow-up of patients in each group. One-way analysis of variance (ANOVA), chi-squared test, and repeated measurement ANOVA were used to compare and analyze the influence of endplate injury morphology on patient prognosis. Most moderate injuries to the endplate (fissure-type injury) and severe injuries (irregular depression-type injury, Schmorl's node-type injury) resulted in significant disc degeneration in the long-term transition. This study also showed significant differences in the height of the anterior margin of the injured spine and the intervertebral space height index during this process. The current study suggests that although the region of injury in endplate fissure-type injury is small preoperatively, it may be a major factor in leading to severe disc degeneration, loss of intervertebral height, and Cobb angle in the long term. The results of our study therefore may allow surgeons to predict the prognosis of patients with thoracolumbar fractures and guide their treatment.

A complex neural network model by Hilbert Transform

The phase information of the optical wave plays a vital role in processing wave-related signals. In deep learning fields, complex-valued neural networks are put forward on the concept of complex amplitudes for full utilization of phase information. To build a complex-valued neural network, the common way is to exploit Fourier Transform of the observed signal to extract amplitude and phase information. However, this will lead to spectrum waste for a real-valued signal by introducing negative frequencies that have no physical meaning. To this end, we attempt to use Hilbert Transform as an alternative to yield a single sideband spectrum and avoid negative frequencies from interacting with positive ones. On the other hand, Fourier transform is a global analysis thus it tells nothing about the time domain. As our key insight, we further explore the usage of instantaneous frequency calculated by Hilbert Transform and propose a new method of constructing complex input from a time–frequency angle. Simple pixel-wise classification experiments are carried out on two hyperspectral datasets and MNIST dataset. Experimental results have demonstrated that Hilbert Transform with instantaneous frequency performs better by a large margin than Fourier Transform owing to the additional time information.

Multi-reference source aircraft classification network based on unknown class prototype estimation

Abstract Gathering complete aircraft types for close-set tasks is challenging and costly in fine-grained aircraft classification, resulting in encountering unknown class aircraft images in real-world models. To address this problem, we propose a network called a multi-reference source fine-grained aircraft classification network (MRSN) to explicitly and implicitly distinguish known class boundaries and embed unknown classes into the sample space. Specifically, we propose an embedding module (UCPE) to synthesize an unknown class prototype to facilitate the modeling of unknown class data distribution. Besides, we introduce a concatenated multi-reference source module (CMS) that combines distance measurement and probability estimation to classify whether testing images are of unknown class or not. Our proposed MRSN innovatively extends the open-set learning paradigm to the fine-grained classification of geospatial objects rather than simple scene classification. Besides, it shows excellent performance in our constructed dataset, which demonstrates the effectiveness and progressiveness of our method.

Web-Based Expert System Application for Early Diagnosis of HIV/AIDS Using the Naive Bayes Method

AIDS is a progressive decrease in the immune system so that opportunistic infections can appear and end in death, therefore the author created an early diagnosis system for HIV/AIDS using the website-based Naïve Bayes algorithm. Naïve Bayes is a simple probability classification that can calculate all possibilities by combining a number of combinations and frequencies of a value from the database obtained.the results of the research obtainedThe naïve Bayes algorithm can be implemented for early diagnosis of HIV/AIDS by means that the existing HIV/AIDS symptom data is adjusted to the patient's symptom data processed using the naïve Bayes algorithm and then it is concluded what the symptoms are and What is the solution.

Adnexal Masses in Pregnancy: A Single-Centre Prospective Observational Cohort Study

Objective: To prospectively determine the nature of adnexal masses diagnosed during pregnancy and investigate whether ultrasound was a reliable means of assessing these. Methods: A single-centre prospective observational cohort study was conducted in a large tertiary referral hospital in London. Pregnant women with an adnexal mass detected at or prior to the 12-week routine ultrasound received a detailed ultrasound by a level II ultrasound practitioner at the time of detection; at 12 weeks; 20 weeks; and 6 weeks postpartum. The following outcomes were recorded: subjective impression of the mass; International Ovarian Tumor Analysis simple rules classification; resolution and intervention rate; and the incidence of complications related to the mass. Results: A total of 28,683 pregnant patients were scanned and an adnexal masses was detected in 277 patients, yielding an incidence of 1%. 274 participants were included in the analysis. Subjective impression was as follows: simple 75.9%; dermoid 29.1%; endometrioma 6.6%; haemorrhagic 3.3%; para-ovarian 2.6%; torted simple 0.7%; decidualized endometrioma 0.4%; fibroma 0.4%; theca luteal 0.4%; and borderline ovarian tumour: 0.7%. There was a significant reduction in the volume at each scan (p &lt; 0.0001). Approximately 74.2% of the masses resolved spontaneously. Surgery was performed in 14/274: 2 antenatally, 6 at caesarean section, and 6 postpartum. In 5/247 (2%), there were complications due to the mass. Using IOTA simple rules, 272/274 (99.3%) (p &lt; 0.0001) were classifiable. Only 1/274 (0.4%) had malignant features as per IOTA (p = 0.05). As there were no confirmed malignancies, the accuracy of IOTA simple rules could not be calculated. Conclusions: Adnexal masses in pregnancy are uncommon and the majority spontaneously resolve. Malignancy is rare, as are complications. In the absence of concerns regarding malignancy or cyst accident, there is no need for additional monitoring of these masses during pregnancy.

The Value of Pretherapeutic Basal Calcitonin Cut-Offs for the Therapeutic Strategy and Prediction of Long-Term Outcome of Patients with Medullary Thyroid Cancer—A 30-Year Single-Center Experience

Background: The clinical relevance of clearly defined pretherapeutic basal calcitonin (bCt) cut-offs for predicting lymph node metastases (LNMs) and long-term outcomes (LOs) has so far not been tested in a large cohort of patients with medullary thyroid cancer included in a Ct screening program during the initial diagnostic workup of thyroid nodules. Material and Methods: Female (f) patients with a bCt level of ≤23 pg/mL and male (m) patients with a level of ≤43 pg/mL were assigned to Group 1 (minimal oncologic risk), patients with a bCt between 24 and 84 pg/mL (f) and 44–99 pg/mL (m) to Group 2 (low oncologic risk), and those with a bCt of ≥85 pg/mL (f) and ≥100 pg/mL (m) to Group 3 (high oncologic risk). All patients underwent surgery applying a uniform surgical protocol. The median follow-up was 100 months. Results: The study included 306 patients. In 3/115 (2.6%) patients in Group 1 and in 3/50 (6.0%) in Group 2, LNM in the central but not lateral neck and no distant metastases (DMet) were documented. In both groups, the biochemical long-term cure rate was 95.7% and the disease-specific-survival (DSS) rate was 100% at 10, 15 and 20 years. Lateral LNM and DMet were diagnosed only in Group 3. The bCt levels of N0 and N1 patients showed broadly overlapping ranges, thus impeding the differentiation between those patients through bCt. Both the cure rate and DSS were significantly worse in Group 3. The overall biochemical long-term cure rate was 78.2%. Conclusions: Within a Ct screening program, grouping patients upon pretherapeutic bCt provides a simple risk classification system for indicating surgery, predicting LN involvement, and LOs.

Comparative Analysis of Classification and Segmentation Performance of Different Dog Breeds Based on Mask R-CNN

The classification of different image data is a common task in real life, which has been achieved as early as convolutional neural networks. With the emergence of R-CNN (Region CNN), simple classification tasks began to develop towards object detection tasks, followed by various segmentation tasks. After Faster R-CNN [4], Mask R-CNN [3] achieved instance segmentation while detecting objects. This algorithm model needs to distinguish different target objects in the image panel and learn a large number of characteristics to represent the details of each target object. In dog data images, there are differences and similarities in the characteristics of dog species, such as fur color, texture, and body shape. So in this article, the author selected and created two different dog datasets, representing targets with significant differences and similarities in features. Compare and analyze the performance of Mask R-CNN object detection and instance segmentation using these two dog datasets.

A new proposal of simplified classification of non-small cell lung cancer oligometastases for easy applicability through systematic literature analysis and meta-analysis validation

IntroductionOligometastasis (OM) exhibits wide range of prognosis, which necessitates appropriate classification for optimal therapeutic decision-making. Complementing the ESTRO-EORTC classification which lacked prognostic differentiation and was rather complex, we propose a new and simpler classification based on systematic literature analysis and meta-analysis validation. MethodThe databases were searched up to April 2024. Inclusion criteria were (1) ≥ 10 Non-small cell lung cancer OM patients, (2) local ablative treatment (LAT) versus control (systemic/supportive treatment), (3) reporting progression free survival (PFS) or overall survival (OS), respectively. A simpler classification was proposed through systematic reviews evaluating outcomes based on OM characteristics. According to this new classification, the LAT benefit and pooled 2-year OS and 1-year PFS percentiles were validated through meta-analysis. ResultsIn overall meta-analysis, LAT was correlated with enhanced 1-year PFS (odds ratio (OR):3.487, p < 0.001) and 2-year OS (OR:2.984, p < 0.001), respectively. According to simplified classification, LAT benefit of 1-year PFS was differentiated with ORs of 5.631 (p < 0.001), 3.484 (p < 0.001), and 1.702 (p = 0.067) for Synchronous (Syn), OPS (Oligopersistence), and OPR (Oligoprogression/recurrence) subgroups, respectively. Inter-subgroup comparisons showed significant differences as well. For 2-year OS, ORs of LAT benefit were 3.366 (p < 0.001), 3.355 (p < 0.001), and 1.821 (p = 0.127) in Syn, OPS, and OPR subgroups, respectively; LAT benefit was significant in Syn and OPS, but not significant in OPR. In pooled percentile comparison, 1-year pooled PFS was significantly lower in the OPR group than others, both in the LAT and control arms. ConclusionBased on a systematic literature analysis and meta-analysis validation, we developed a simpler three-step OM classification: Syn, OPS, and OPR. We would propose this new classification that is simpler and more applicable to clinical decisions than the currently available classification.

The underlying mechanisms of alignment in error backpropagation through arbitrary weights

Understanding the mechanisms by which plasticity in millions of synapses in the brain is orchestrated to achieve behavioral and cognitive goals is a fundamental question in neuroscience. In this regard, insights from learning methods in artificial neural networks (ANNs) and in particular supervised learning using backpropagation (BP) seem inspiring. However, the implementation of BP requires exact matching between forward and backward weights, which is unrealistic given the known connectivity pattern in the brain (known as the “weight transport problem”). Notably, it has been shown that under certain conditions, error BackPropagation Through Arbitrary Weights (BP-TAW) can lead to a partial alignment between forward and backward weights (weight alignment or WA). This learning algorithm, which is also known as feedback alignment (FA), can result in surprisingly good degrees of accuracy in simple classification tasks. However, the underlying mechanisms and mathematical basis of WA are not thoroughly understood. In this work, we demonstrate the mathematical basis of WA and answer the question of why and in what conditions WA occurs. We show that the occurrence of WA in ANNs is induced by statistical properties of the output and error signals of neurons, such as autocorrelation and cross-correlation, and can happen even in the absence of learning or reduction of the loss function. Moreover, we show that WA can be improved significantly by limiting the norm of input weights to neurons and that such a weight normalization (WN) method can improve the classification accuracy of BP-TAW. The findings presented can be used to further improve the performance of BP-TAW and open new ways for exploring possible learning mechanisms in biological neural networks without exact matching between forward and backward weights.

Simple and accurate genomic classification model for distinguishing between human and pig Staphylococcus aureus

Staphylococcus aureus (S. aureus) can cause various infections in humans and animals, contributing to high morbidity and mortality. To prevent and control cross-species transmission of S. aureus, it is necessary to understand the host-associated genetic variants. We performed a two-stage genome-wide association study (GWAS) including initial screening and further validation to compare genomic differences between human and pig S. aureus, aiming to identify host-associated determinants. Our multiple GWAS analyses found six consensus significant k-mers associated with host species, providing novel genetic evidence for distinguishing human from pig S. aureus. The best k-mer predictor achieved a high classification accuracy of 98.12% on its own and had extremely high resolution similar to the SNPs-based phylogeny, offering a very simple target for predicting the cross-species transmission risk of S. aureus. The final k-mer model revealed that 90% of S. aureus isolates from farm workers were predicted as livestock origin, suggesting a high risk of cross-species transmission. Bayesian inference revealed different cross-species transmission directions, with the human-to-pig transmission for ST5 and the pig-to-human transmission for ST398. Our findings provide a simple and accurate k-mer model for identifying and predicting the cross-species transmission risk of S. aureus.

Gemini-Assisted Deep Learning Classification Model for Automated Diagnosis of High-Resolution Esophageal Manometry Images.

Background/Objectives: To develop a deep learning model for esophageal motility disorder diagnosis using high-resolution manometry images with the aid of Gemini. Methods: Gemini assisted in developing this model by aiding in code writing, preprocessing, model optimization, and troubleshooting. Results: The model demonstrated an overall precision of 0.89 on the testing set, with an accuracy of 0.88, a recall of 0.88, and an F1-score of 0.885. It presented better results for multiple categories, particularly in the panesophageal pressurization category, with precision = 0.99 and recall = 0.99, yielding a balanced F1-score of 0.99. Conclusions: This study demonstrates the potential of artificial intelligence, particularly Gemini, in aiding the creation of robust deep learning models for medical image analysis, solving not just simple binary classification problems but more complex, multi-class image classification tasks.

Orthophoto-Based Vegetation Patch Analyses—A New Approach to Assess Segmentation Quality

The following paper focuses on evaluating the quality of image prediction in the context of searching for plants of a single species, using the example of Heracleum sosnowskyi Manden, in a given area. This process involves a simplified classification that ends with a segmentation step. Because of the particular characteristics of environmental data, such as large areas of plant occurrence, significant partitioning of the population, or characteristics of a single individual, the use of standard statistical measures such as Accuracy, the Jaccard Index, or Dice Coefficient does not produce reliable results, as shown later in this study. This issue demonstrates the need for a new method for assessing the betted prediction quality adapted to the unique characteristics of vegetation patch detection. The main aim of this study is to provide such a metric and demonstrate its usefulness in the cases discussed. Our proposed metric introduces two new coefficients, M+ and M−, which, respectively, reward true positive regions and penalise false positive regions, thus providing a more nuanced assessment of segmentation quality. The effectiveness of this metric has been demonstrated in different scenarios focusing on variations in spatial distribution and fragmentation of theoretical vegetation patches, comparing the proposed new method with traditional metrics. The results indicate that our metric offers a more flexible and accurate assessment of segmentation quality, especially in cases involving complex environmental data. This study aims to demonstrate the usefulness and applicability of the metric in real-world vegetation patch detection tasks.

Classifying cancer-associated fibroblasts—The good, the bad, and the target

Cancer-associated fibroblasts (CAFs) are heterogeneous and ubiquitous stromal cells within the tumor microenvironment (TME). Numerous CAF types have been described, typically using single-cell technologies such as single-cell RNA sequencing. There is no general classification system for CAFs, hampering their study and therapeutic targeting. We propose a simple CAF classification system based on single-cell phenotypes and spatial locations of CAFs in multiple cancer types, assess how our scheme fits within current knowledge, and invite the CAF research community to further refine it.