Receptive Field Size Research Articles

Deep learning approaches based improved light weight U-Net with attention module for optic disc segmentation.

Glaucoma is a major cause of blindness worldwide, and its early detection is essential for the timely management of the condition. Glaucoma-induced anomalies of the optic nerve head may cause variation in the Optic Disc (OD) size. Therefore, robust OD segmentation techniques are necessary for the screening for glaucoma. Computer-aided segmentation has become a promising diagnostic tool for the early detection of glaucoma, and there has been much interest in recent years in using neural networks for medical image segmentation. This study proposed an enhanced lightweight U-Net model with an Attention Gate (AG) to segment OD images. We also used a transfer learning strategy to extract relevant features using a pre-trained EfficientNet-B0 CNN, which preserved the receptive field size and AG, which reduced the impact of gradient vanishing and overfitting. Additionally, the neural network trained using the binary focal loss function improved segmentation accuracy. The pre-trained Attention U-Net was validated using publicly available datasets, such as DRIONS-DB, DRISHTI-GS, and MESSIDOR. The model significantly reduced parameter quantity by around 0.53 M and had inference times of 40.3 ms, 44.2 ms, and 60.6 ms, respectively.

LAE-Net: A locally-adaptive embedding network for low-light image enhancement

In the low-light enhancement task, one of the major challenges lies in how to balance the image enhancement properties of light intensity, detail presentation and color fidelity. In natural scenes, the multi-distribution of frequency and illumination characteristics in the spatial domain makes the balance more difficult. To solve this problem, we propose a Locally-Adaptive Embedding Network, namely LAE-Net, to realize high-quality low-light image enhancement with locally-adaptive kernel selection and feature adaptation for multi-distribution issues. Specifically, for the frequency multi-distribution, we rethink the spatial-frequency characteristic of human eyes, experimentally explore the relationship among the receptive field size, the image spatial frequency and the light enhancement properties, and propose an Entropy-Inspired Kernel-Selection Convolution, where each neuron can adaptively adjust the receptive field size according to its spatial frequency characterized by information entropy. For the illumination multi-distribution, we propose an Illumination Attentive Transfer subnet, where the neurons can simultaneously sense global consistency and local details, and accordingly hint where to focus the efforts on, thereby adjusting the refined features. Extensive experiments with ablation analysis show the effectiveness of our method and the proposed method outperforms many related state-of-the-art techniques on four benchmark datasets: MEF, LIME, NPE and DICM.

Cell type-specific calcium imaging of central sensitization in mouse dorsal horn

Allodynia is a state in which pain is elicited by innocuous stimuli. Capsaicin applied to the skin results in an allodynia that extends to a broad region beyond the application site. This sensitization is thought to be mediated by spinal networks; however, we do not have a clear picture of which spinal neurons mediate this phenomenon. To address this gap, we used two-photon calcium imaging of excitatory interneurons and spinal projection neurons in the mouse spinal dorsal horn. To distinguish among neuronal subtypes, we developed CICADA, a cell profiling approach to identify cell types during calcium imaging. We then identified capsaicin-responsive and capsaicin-sensitized neuronal populations. Capsaicin-sensitized neurons showed emergent responses to innocuous input and increased receptive field sizes consistent with psychophysical reports. Finally, we identified spinal output neurons that showed enhanced responses from innocuous input. These experiments provide a population-level view of central sensitization and a framework with which to model somatosensory integration in the dorsal horn.

Cortical mechanisms of visual brightness.

SUMMARYThe primary visual cortex signals the onset of light and dark stimuli with ON and OFF cortical pathways. Here, we demonstrate that both pathways generate similar response increments to large homogeneous surfaces and their response average increases with surface brightness. We show that, in cat visual cortex, response dominance from ON or OFF pathways is bimodally distributed when stimuli are smaller than one receptive field center but unimodally distributed when they are larger. Moreover, whereas small bright stimuli drive opposite responses from ON and OFF pathways (increased versus suppressed activity), large bright surfaces drive similar response increments. We show that this size-brightness relation emerges because strong illumination increases the size of light surfaces in nature and both ON and OFF cortical neurons receive input from ON thalamic pathways. We conclude that visual scenes are perceived as brighter when the average response increments from ON and OFF cortical pathways become stronger.

Dynamic changes in spatial representation within the posterior parietal cortex in response to visuomotor adaptation.

Recent studies used functional magnetic resonance imaging (fMRI)population receptive field (pRF) mapping to demonstrate that retinotopic organization extends from the primary visual cortex to ventral and dorsal visual pathways, by quantifying visual field maps, receptive field size, and laterality throughout multiple areas. Visuospatial representation in the posterior parietal cortex (PPC) is modulated by attentional deployment, raising the question of whether spatial representation in the PPC is dynamic and flexible, and whether this flexibility contributes to visuospatial learning. To answer this question, changes in spatial representation within the PPC and early visual cortex were recorded with pRF mapping before and after prism adaptation (PA)-a well-established visuomotor technique that modulates visuospatial attention according to the direction of the visual displacement. As predicted, results showed that adaptation to left-shifting prisms increases pRF size in left PPC, while leaving space representation in the early visual cortex unchanged. This is the first evidence that PA drives a dynamic reorganization of response profiles in the PPC. These findings show that spatial representations in the PPC not only reflect changes driven by attentional deployment but dynamically change in response to modulation of external factors such as manipulation of the visuospatial input during visuomotor adaptation.

Activating parvalbumin-expressing interneurons produces iceberg effects in mouse primary visual cortex neurons

In the primary visual cortex (V1) inhibitory interneurons form a local circuit with excitatory pyramidal cells to produce distinct receptive field properties. Parvalbumin-expressing interneurons (Pvalb+) are the most common subclass of V1 interneurons, and studies of orientation tuning indicate they shape pyramidal stimulus selectivity by balancing excitation with inhibition relative to the spike threshold. The iceberg effect, where subthreshold responses have broader tuning than spiking responses, predicts that other receptive field properties besides orientation tuning should also be affected by this balance mediated by Pvalb+ cells. To test this, we measured receptive field size and visual latency of pyramidal cells while Pvalb+ activity was optogenetically increased. We found that amplifying Pvalb+ input to pyramidal cells significantly increased their latency and decreased their receptive field size, which corroborates the proposed role of Pvalb+ interneurons in sculpting pyramidal tuning by controlling cortical gain.

An Unsupervised Monocular Visual Odometry Based on Multi-Scale Modeling.

Unsupervised deep learning methods have shown great success in jointly estimating camera pose and depth from monocular videos. However, previous methods mostly ignore the importance of multi-scale information, which is crucial for pose estimation and depth estimation, especially when the motion pattern is changed. This article proposes an unsupervised framework for monocular visual odometry (VO) that can model multi-scale information. The proposed method utilizes densely linked atrous convolutions to increase the receptive field size without losing image information, and adopts a non-local self-attention mechanism to effectively model the long-range dependency. Both of them can model objects of different scales in the image, thereby improving the accuracy of VO, especially in rotating scenes. Extensive experiments on the KITTI dataset have shown that our approach is competitive with other state-of-the-art unsupervised learning-based monocular methods and is comparable to supervised or model-based methods. In particular, we have achieved state-of-the-art results on rotation estimation.

Model-based characterization of the selectivity of neurons in primary visual cortex.

Statistical models are increasingly being used to understand the complexity of stimulus selectivity in primary visual cortex (V1) in the context of complex time-varying stimuli, replacing averaging responses to simple parametric stimuli. Although such models often can more accurately reflect the computations performed by V1 neurons in more natural visual environments, they do not by themselves provide insight into V1 neural selectivity to basic stimulus features such as receptive field size, spatial frequency tuning, and phase invariance. Here, we present a battery of analyses that can be directly applied to encoding models to link complex encoding models to more interpretable aspects of stimulus selectivity. We apply this battery to nonlinear models of V1 neurons recorded in awake macaque during random bar stimuli. In linking model properties to more classical measurements, we demonstrate several novel aspects of V1 selectivity not available to simpler experimental measurements. For example, this approach reveals that individual spatiotemporal elements of the V1 models often have a smaller spatial scale than the neuron as a whole, resulting in nontrivial tuning to spatial frequencies. In addition, we propose measures of nonlinear integration that suggest that classical classifications of V1 neurons into simple versus complex cells will be spatial-frequency dependent. In total, rather than obfuscate classical characterizations of V1 neurons, model-based characterizations offer a means to more fully understand their selectivity, and link their classical tuning properties to their roles in more complex, natural, visual processing.NEW & NOTEWORTHY Visual neurons are increasingly being studied with more complex, natural visual stimuli, and increasingly complex models are necessary to characterize their response properties. Here, we describe a battery of analyses that relate these more complex models to classical characterizations. Using such model-based characterizations of V1 neurons furthermore yields several new insights into V1 processing not possible to capture in more classical means to measure their visual selectivity.

Triple-branch ternary-attention mechanism network with deformable 3D convolution for hyperspectral image classification

ABSTRACT In recent years, the classification of hyperspectral images (HSI) has received extensive research attention. As compared with traditional HSI classification, which only uses spectral information, it is found that spatial information is also essential in HSI classification. To effectively utilize the spectral and spatial information of HSI, this paper proposes a triple-branch ternary-attention mechanism network with deformable 3D convolution (D3DTBTA). In D3DTBTA, three branches, i.e. the spectral, spatial-X, and spatial-Y branches, are combined with the attention mechanism in three directions, which can better capture the vector features of three dimensions in HSI. Furthermore, considering the adaptation of scale and receptive field size in the convolution operation, our method uses deformable convolution to enable D3DTBTA to enhance feature extraction. Our experimental results show that the framework outperforms existing algorithms on four hyperspectral datasets, especially when the training samples are limited.

An Attention Encoder-Decoder Network Based on Generative Adversarial Network for Remote Sensing Image Dehazing

Remote sensing image dehazing is a difficult problem for its complex characteristics. It can be regarded as the preprocessing of high-level tasks of remote sensing images. To remove haze from the hazy remote sensing image, an encoder-decoder based on generative adversarial network is proposed. It first learns the low-frequency information of the image, and then learns the high-frequency information of the image. The skip connection is also added in the network to avoid losing information. To further improve the ability of learning more useful information, a multi-scale attention module is proposed. Meanwhile, a CBlock module is also designed to extract more feature information. It can capture different size of receptive fields. In order to reduce the computational pressure of the network, a distillation module is used in the network. Inspired by multi-scale network, an enhance module is designed and introduced it in the end of the network to further improve the dehazing ability of the network by integrating context information on multi-scale. We compared with five methods and our proposed method on RICE dataset. Experimental results show that our method achieves the best effect, both qualitatively and quantitatively.

Phytochemical quercetin alleviates hyperexcitability of trigeminal nociceptive neurons associated with inflammatory hyperalgesia comparable to NSAIDs.

Quercetin is a flavonoid that is widely found in fruits and vegetables. Quercetin inhibits cyclooxygenase-2 and modulates voltage-gated ion channels, however, its effect on nociceptive neuron-associated inflammatory hyperalgesia remains unknown. The present study investigated under in vivo conditions whether systemic administration of quercetin attenuates the inflammation-induced hyperexcitability of trigeminal spinal nucleus caudalis (SpVc) neurons associated with mechanical hyperalgesia and compared its effect to the non-steroidal anti-inflammatory drug, diclofenac. Complete Freund’s adjuvant was injected into the whisker pads of rats to induce inflammation, and then mechanical stimulation was applied to the orofacial area to assess the threshold of escape. The mechanical threshold was significantly lower in inflamed rats compared to uninjected naïve rats, and this lowered threshold returned to control levels 2 days after administration of quercetin or diclofenac. The mean discharge frequency of SpVc wide-dynamic range (WDR) neurons to both non-noxious and noxious mechanical stimuli in inflamed rats was significantly decreased after quercetin or diclofenac administration under combination of three anesthetic agents (medetomidine, midazolam and butorphanol). In addition, the increased mean spontaneous discharge of SpVc WDR neurons in inflamed rats significantly decreased after quercetin or diclofenac administration. Similarly, quercetin or diclofenac restored the expanded mean receptive field size in inflamed rats to control levels. In this study, the combination of three anesthetic agents did not result in any obvious “noxious pinch-evoked after discharges” in CFA inflamed day 2 rat as described previously in pentobarbital-anesthetized rats. Together, these results suggest that administration of quercetin attenuates inflammatory hyperalgesia associated with hyperexcitability of nociceptive SpVc WDR neurons via inhibition of the peripheral cyclooxygenase-2 signaling cascade and voltage-gated ion channels. These findings support the proposed potential of quercetin as a therapeutic agent in complementary alternative medicine strategies for preventing trigeminal inflammatory mechanical hyperalgesia.

IENet: a robust convolutional neural network for EEG based brain-computer interfaces

Objective. Brain-computer interfaces (BCIs) based on electroencephalogram (EEG) develop into novel application areas with more complex scenarios, which put forward higher requirements for the robustness of EEG signal processing algorithms. Deep learning can automatically extract discriminative features and potential dependencies via deep structures, demonstrating strong analytical capabilities in numerous domains such as computer vision and natural language processing. Making full use of deep learning technology to design a robust algorithm that is capable of analyzing EEG across BCI paradigms is our main work in this paper. Approach. Inspired by InceptionV4 and InceptionTime architecture, we introduce a neural network ensemble named InceptionEEG-Net (IENet), where multi-scale convolutional layer and convolution of length 1 enable model to extract rich high-dimensional features with limited parameters. In addition, we propose the average receptive field (RF) gain for convolutional neural networks (CNNs), which optimizes IENet to detect long patterns at a smaller cost. We compare with the current state-of-the-art methods across five EEG-BCI paradigms: steady-state visual evoked potentials (VEPs), epilepsy EEG, overt attention P300 VEPs, covert attention P300 visual-EPs and movement-related cortical potentials. Main results. The classification results show that the generalizability of IENet is on par with the state-of-the-art paradigm-agnostic models on test datasets. Furthermore, the feature explainability analysis of IENet illustrates its capability to extract neurophysiologically interpretable features for different BCI paradigms, ensuring the reliability of algorithm. Significance. It can be seen from our results that IENet can generalize to different BCI paradigms. And it is essential for deep CNNs to increase the RF size using average RF gain.

A Fast Multi-Scale Generative Adversarial Network for Image Compressed Sensing.

Recently, deep neural network-based image compressed sensing methods have achieved impressive success in reconstruction quality. However, these methods (1) have limitations in sampling pattern and (2) usually have the disadvantage of high computational complexity. To this end, a fast multi-scale generative adversarial network (FMSGAN) is implemented in this paper. Specifically, (1) an effective multi-scale sampling structure is proposed. It contains four different kernels with varying sizes so that decompose, and sample images effectively, which is capable of capturing different levels of spatial features at multiple scales. (2) An efficient lightweight multi-scale residual structure for deep image reconstruction is proposed to balance receptive field size and computational complexity. The key idea is to apply smaller convolution kernel sizes in the multi-scale residual structure to reduce the number of operations while maintaining the receptive field. Meanwhile, the channel attention structure is employed for enriching useful information. Moreover, perceptual loss is combined with MSE loss and adversarial loss as the optimization function to recover a finer image. Numerous experiments show that our FMSGAN achieves state-of-the-art image reconstruction quality with low computational complexity.

Population Imaging of Central Sensitization

Capsaicin applied locally to the skin causes central sensitization that results in allodynia, a state in which pain is elicited by innocuous stimuli. Here, we used two-photon calcium imaging of neurons in the dorsal spinal cord to visualize central sensitization across excitatory interneurons and spinal projection neurons. To distinguish among excitatory neuron subtypes, we developed CICADA, a cell profiling approach that leverages the expression of distinct Gq-coupled receptors. We then identified capsaicin-responsive and capsaicin-sensitized neuronal populations. Capsaicin-sensitized neurons showed emergent responses to low threshold input and increased receptive field sizes consistent with the psychophysical phenomenon that allodynia is observed across an extended secondary zone. Finally, we identified spinal projection neurons that showed a shift in tuning toward low threshold input. These experiments provide a population-level view of central sensitization and a framework with which to model somatosensory integration in the dorsal horn. Grant support from NINDS.

PHENOTYPIC PARAMETER EXTRACTION FOR WHEAT EARS BASED ON AN IMPROVED MASK-RCNN ALGORITHM

The acquisition of traditional wheat ear phenotypic parameters is labour intensive and subjective, and some trait parameters are difficult to measure, which greatly limits the progress of wheat ear research. To obtain the phenotypic parameters of wheat ears in batches at a low cost, this paper proposed a convenient and accurate method for extracting phenotypic parameters of wheat ears. First, three improvement directions were proposed based on the Mask Region-Convolutional Neural Network (Mask-RCNN) model. 1) To extract the multiscale features of wheat ears, a hierarchical residual link was constructed in a single residual block of the backbone network ResNet101 to obtain information on different sizes of receptive fields. 2) The feature pyramid network (FPN) was improved to increase the recognition accuracy of wheat ear edges through multiple two-way information flow sampling. 3) The mask evaluation mechanism was improved, specific network blocks were used to learn and predict the quality of the mask, and the detection of wheat ears and grains was performed by precise segmentation; an automatic extraction algorithm was designed for wheat ear phenotypic parameters based on the segmentation results to extract 22 phenotypic parameters. The experiments showed that the improved Mask-RCNN was superior to the existing model in the segmentation accuracy of wheat ears and grains; the parameters of wheat ear length, width, and number of grains extracted by the automatic extraction algorithm were close to the manual measurement values. This research meets the demand for automatic extraction of wheat ear phenotype data for large-scale quality testing and commercial breeding and has strong practicability.

ILU-Net: Inception-Like U-Net for retinal vessel segmentation

High-precision segmentation of retinal vessels is a critical procedure for clinicians to analyze retinal diseases. However, previous U-Net based segmentation methods are lost in the preservation of capillaries and vascular structures of complex bifurcation. To solve these problems, this paper proposes a novel U-Net based network, named Inception-Like U-Net (ILU-Net), including the following key points: First, two different types of inception-like block (Down-sampling Inception Blocks (DIB) in encoder and up-sampling Inception Blocks (UIB) in decoder) replace the two convolution operations in U-Net to learn the characteristics from different sizes of receptive field. Second, a novel skip connection strategy is proposed and can better connect low-level features to high-level features for reproducing more micro-scale vascular structures. Third, the segmentation performance is further improved by using the combination of two activation functions (ReLU and ELU). Qualitative and quantitative results of experiments on the public DRIVE dataset demonstrate that the proposed ILU-Net approach is effective and even better than the state-of-the-arts in preserving capillaries and complex bifurcation structures.

An effective modular approach for crowd counting in an image using convolutional neural networks

Abrupt and continuous nature of scale variation in a crowded scene is a challenging task to enhance crowd counting accuracy in an image. Existing crowd counting techniques generally used multi-column or single-column dilated convolution to tackle scale variation due to perspective distortion. However, due to multi-column nature, they obtain identical features, whereas, the standard dilated convolution (SDC) with expanded receptive field size has sparse pixel sampling rate. Due to sparse nature of SDC, it is highly challenging to obtain relevant contextual information. Further, features at multiple scale are not extracted despite some inception-based model is not used (which is cost effective). To mitigate theses drawbacks in SDC, we therefore, propose a hierarchical dense dilated deep pyramid feature extraction through convolution neural network (CNN) for single image crowd counting (HDPF). It comprises of three modules: general feature extraction module (GFEM), deep pyramid feature extraction module (PFEM) and fusion module (FM). The GFEM is responsible to obtain task independent general features. Whereas, PFEM plays a vital role to obtain the relevant contextual information due to dense pixel sampling rate caused by densely connected dense stacked dilated convolutional modules (DSDCs). Further, due to dense connections among DSDCs, the final feature map acquires multi-scale information with expanded receptive field as compared to SDC. Due to dense pyramid nature, it is very effective to propagate the extracted feature from lower dilated convolutional layers (DCLs) to middle and higher DCLs, which result in better estimation accuracy. The FM is used to fuse the incoming features extracted by other modules. The proposed technique is tested through simulations on three well known datasets: Shanghaitech (Part-A), Shanghaitech (Part-B) and Venice. Results justify its relative effectiveness in terms of selected performance.

KiU-Net: Overcomplete Convolutional Architectures for Biomedical Image and Volumetric Segmentation.

Most methods for medical image segmentation use U-Net or its variants as they have been successful in most of the applications. After a detailed analysis of these "traditional" encoder-decoder based approaches, we observed that they perform poorly in detecting smaller structures and are unable to segment boundary regions precisely. This issue can be attributed to the increase in receptive field size as we go deeper into the encoder. The extra focus on learning high level features causes U-Net based approaches to learn less information about low-level features which are crucial for detecting small structures. To overcome this issue, we propose using an overcomplete convolutional architecture where we project the input image into a higher dimension such that we constrain the receptive field from increasing in the deep layers of the network. We design a new architecture for im- age segmentation- KiU-Net which has two branches: (1) an overcomplete convolutional network Kite-Net which learns to capture fine details and accurate edges of the input, and (2) U-Net which learns high level features. Furthermore, we also propose KiU-Net 3D which is a 3D convolutional architecture for volumetric segmentation. We perform a detailed study of KiU-Net by performing experiments on five different datasets covering various image modalities. We achieve a good performance with an additional benefit of fewer parameters and faster convergence. We also demonstrate that the extensions of KiU-Net based on residual blocks and dense blocks result in further performance improvements. Code: https://github.com/jeya-maria-jose/KiU-Net-pytorch.

Urban forestry detection by deep learning method with GaoFen-2 remote sensing images

Urban forestry is regarded as the green infrastructure of cities, the analysis of which is of great significance for research on statistical analysis, such as greening, carbon sink, etc. on the basis of remote sensing detection. However, current urban forestry detection algorithms are focused on typical machine learning, whose results are limited by the manually selected features. Therefore, an innovative detection algorithm for analyzing urban forestry is proposed. First, the simple linear iterative clustering algorithm is adopted to segment the dataset, and the edge information of the object is preserved as much as possible. Second, a new network structure is constructed on the basis of Mask-RCNN to refine the detection results: (1) dilated convolution is adopted to replace the normal convolution of C1 in backbone (ResNet101) to increase the size of receptive field; (2) convolutional block attention module is introduced in C2 to give more weight to the shallow features such as edges and colors; and (3) atrous spatial pyramid pooling (ASPP) is added to branch of C3, enhancing the ability of multiscale fusion. With the effect of ASPP on detection results at different positions of backbone compared, the results show that the more backward the ASPP is, the more obvious the trend of negative correlation between network convergence speed and detection accuracy (C3: 89.2%, C4: 84.3%, and C5: 81.6%) is. The accuracy of random forest and support vector machines is also compared, with the former of 69.4% and the latter of 71.2%. In addition, the method of transfer learning to unmanned aerial vehicle data is used in this research, with the accuracy of 81.6%. The result shows that the improved segmentation model features better ability of learning shallow features than the original, able to accurately extract urban forestry from remote sensing images, which provides the reference for the further research on urban forestry detection, mapping, as well as the multisource remote sensing data fusion.

Development of a Piezoelectric Actuated Tactile Stimulation Device for Population Receptive Field Mapping in Human Somatosensory Cortex With fMRI.

Multichannel tactile stimulation devices is need to investigate human finger population receptive field (pRF) characteristics in the primary somatosensory cortex during functional magnetic resonance imaging (fMRI). To accurately characterize right-hand somatosensory representation based on the Bayesian pRF model. Prospective. A water phantom and six healthy participants (four males, mean 23.8 years old). T1-weighted magnetization-prepared rapid gradient-echo, T2*-weighted echo planar imaging at 3T. The piezoelectric actuated tactile stimulation device consisted of execution unit and control unit. The output performance of the device was measured by a laser displacement sensor. The effect of the device on images' signal-to-noise ratio (SNR) was measured by phantom experiments. The activation representation arrangement order, relative volumes, and receptive field size of the right hand were assessed during the along-digits and cross-digits paradigms. The normality of the data was tested by the Shapiro-Wilk method. A paired-sample t test was performed to test pRF characteristics for all digit pairings. The significance level was set to P=0.05 (false discovery rate [FDR] correct). Percussive stimulation provided by the piezoelectric actuated tactile stimulator had a stable displacement (2.64 mm) over a wide range of vibration frequencies (0-30 Hz). The output delay of the device was 1millisecond. The device did not affect the image's SNR (without the device: SNR=138.24 ± 7.87, temporal SNR [TSNR]=440.03 ± 52.08. With the device: SNR=138.06 ± 8.44, TSNR=438.52 ± 56.38. PSNR =0.88, PTSNR =0.46). Representations of right-hand fingers showed the same arrangement order in both experiments (D1-D5 arranged along the central sulcus). However, the relative volumes of D3 showed significant differences in S1 (P=0.003). Among four subareas, the relative volumes of D3 were significantly different in area 1 (P=0.047). This developed stimulator, through experimental verification, could play a role in pRF mapping exploration. 2 TECHNICAL EFFICACY STAGE: 1.