Processing Steps Research Articles

Demonstrating the Effectiveness of an Alternative to Triton X-100 for Detergent-Mediated Viral Inactivation in Biomanufacturing.

Detergent-mediated viral inactivation is an important process step for ensuring viral safety of parenteral biotherapeutics, including plasma proteins and monoclonal antibodies (mAb). The conventional Triton X-100 detergent has ecological toxicity concerns and REACH classification that mandate replacement in the biopharmaceutical industry. Criteria for a replacement detergent include viral inactivation efficacy, acceptable safety and biodegradation profile, process removal, and quality suitable for parenteral drug product manufacturing. A non-ionic, C11-15 secondary alcohol ethoxylate, Deviron 13-S9 detergent, has been demonstrated to meet the necessary requirements for detergent performance. Benchmarking studies with Triton X-100 detergent demonstrate comparable performance with a panel of enveloped viruses in multiple matrices, including human IgG, clarified cell culture harvest, and fractionated plasma. Deviron 13-S9 detergent demonstrated viral inactivation efficiency comparable to or better than Triton X-100 detergent, achieving > 5 log reduction values. Critical micelle concentration was determined across different temperatures and media. Deviron 13-S9 detergent was demonstrated to be readily biodegradable according to OECD 301B guidelines. The absence of detergent binding to typical chromatography resins used in downstream purification was confirmed. The process removal of Deviron 13-S9 detergent from a protein-containing matrix was demonstrated using a protein A resin. These findings support Deviron 13-S9 detergent as a viable alternative to Triton X-100 detergent, ensuring robust viral inactivation, environmental compatibility, and alignment with regulatory requirements.

Transitioning from wet lab to artificial intelligence: a systematic review of AI predictors in CRISPR.

The revolutionary CRISPR-Cas9 system leverages a programmable guide RNA (gRNA) and Cas9 proteins to precisely cleave problematic regions within DNA sequences. This groundbreaking technology holds immense potential for the development of targeted therapies for a wide range of diseases, including cancers, genetic disorders, and hereditary diseases. CRISPR-Cas9 based genome editing is a multi-step process such as designing a precise gRNA, selecting the appropriate Cas protein, and thoroughly evaluating both on-target and off-target activity of the Cas9-gRNA complex. To ensure the accuracy and effectiveness of CRISPR-Cas9 system, after the targeted DNA cleavage, the process requires careful analysis of the resultant outcomes such as indels and deletions. Following the success of artificial intelligence (AI) in various fields, researchers are now leveraging AI algorithms to catalyze and optimize the multi-step process of CRISPR-Cas9 system. To achieve this goal AI-driven applications are being integrated into each step, but existing AI predictors have limited performance and many steps still rely on expensive and time-consuming wet-lab experiments. The primary reason behind low performance of AI predictors is the gap between CRISPR and AI fields. Effective integration of AI into multi-step CRISPR-Cas9 system demands comprehensive knowledge of both domains. This paper bridges the knowledge gap between AI and CRISPR-Cas9 research. It offers a unique platform for AI researchers to grasp deep understanding of the biological foundations behind each step in the CRISPR-Cas9 multi-step process. Furthermore, it provides details of 80 available CRISPR-Cas9 system-related datasets that can be utilized to develop AI-driven applications. Within the landscape of AI predictors in CRISPR-Cas9 multi-step process, it provides insights of representation learning methods, machine and deep learning methods trends, and performance values of existing 50 predictive pipelines. In the context of representation learning methods and classifiers/regressors, a thorough analysis of existing predictive pipelines is utilized for recommendations to develop more robust and precise predictive pipelines.

Best Practices for Conducting Systematic Reviews: Perspectives of Experienced Systematic Review Researchers in Educational Sciences

Abstract Systematic reviews have been gaining attention as a research methodology and are among the most frequently cited sources in educational sciences. However, best practices for conducting systematic reviews in educational sciences are still evolving. We conducted N = 12 qualitative interviews to learn from experienced systematic review researchers across various educational disciplines and geographic locations. The interviews focused on the different steps of the systematic review process (i.e., designing, including/excluding, screening, coding, analyzing, reporting), benefits, challenges, team collaboration, ethical considerations, and technologies while conducting systematic reviews. Several themes were identified, providing best practices for conducting systematic reviews and highlighting the importance of a systematic, comprehensive, and transparent research process. The findings have implications for researchers who wish to conduct systematic reviews, instructors who teach students on conducting systematic reviews, as well as editors and reviewers of journals who publish systematic reviews.

Lived Experiences of Older Adults with Advanced CKD and Their Caregivers: A Qualitative Study Uncovering Risk and Resilience Dynamics.

To date, very little is known about the lived experiences of families impacted by chronic kidney disease (CKD), especially regarding the adaptive coping strategies these families use to successfully cope with the chronic stress they must face due to CKD. An exploratory qualitative descriptive study was conducted by recruiting a sub-sample of adults with advanced CKD participating in the Canadian Frailty Observation and Interventions Trial (CanFIT) study and some of their caregivers. As part of this ongoing larger study, 12 adults with advanced CKD and seven of their caregivers (N = 19) completed one focus group discussion that explored topics related to their unique lived experiences of individuals impacted by CKD. Narrative data was analyzed using a 3-step inductive thematic analysis process. Three themes that portray participants' lived experiences were identified, including 1) experiencing chronic stress due to CKD; 2) coping successfully with the stress caused by CKD; and 3) recommendations to improve family well-being. Social and health services for families impacted by CKD may be more effective in promoting the health, well-being, and quality of life of both adults with CKD and their caregivers if they acknowledge the chronic stressors these families face daily and provide support strategies that help them successfully cope with such stressors.

Direct observations of nucleation and early-stage growth of Au-catalyzed GaAs nanowires on Si(111)

Developing a reliable procedure for the growth of III-V nanowires (NW) on silicon (Si) substrates remains a significant challenge, as current methods rely on trial-and-error approaches with varying interpretations of critical process steps such as sample preparation, Au-Si alloy formation in the growth reactor, and NW alignment. Addressing these challenges is essential for enabling high-performance electronic and optoelectronic devices that combine the superior properties of III-V NW semiconductors with the well-established Si-based technology. Combining conventional scalable growth methods, such as metalorganic chemical vapor deposition (MOCVD) within situcharacterization using environmental transmission electron microscopy (ETEM-MOCVD) enables a deeper understanding of the growth dynamics, if that knowledge is transferable to the scalable processes. We report on successful epitaxial growth of Au-catalyzed GaAs NWs on Si(111) substrates using micro-electromechanical system chips with monocrystalline Si-cantilevers in both conventional MOCVD and ETEM-MOCVD systems. The conventional MOCVD provided a framework for initial parameter tuning, while ETEM-MOCVD offered valuable insights into early nucleation and catalyst-substrate interactions. Our findings show that nucleation is significantly influenced by the removal of native oxide layers and the initial formation of the Au-Si alloy. Ourin situstudies revealed different NW-substrate interfaces, essential for optimizing the epitaxial growth process. We identified three typical configurations of NW 'roots', each impacted by growth conditions and preparation steps, affecting the structural and potentially the optical properties of the NWs. Similarly, doping from the Si-substrate may affect both optical and electrical properties; however, compositional analysis revealed no traces of Si in NWs post-nucleation and a small amount in the catalytic droplet. Our research highlights the importance ofin situstudies for a comprehensive understanding of nucleation mechanisms, paving the way for optimizing III-V NW growth on Si substrates and developing high-performance III-V/Si devices.

Magnetic Pulse Powder Compaction

Powder metallurgy (PM) offers several advantages over conventional melt metallurgy, including improved homogeneity, fine grain size, and pseudo-alloying capabilities. Transitioning from conventional methods to PM can result in significant enhancements in material properties and production efficiency by eliminating unnecessary process steps. Dynamic compaction techniques, such as impulse and explosive compaction, aim to achieve higher powder density without requiring sintering, further improving PM efficiency. Among these techniques, magnetic pulse compaction (MPC) has gained notable interest due to its unique process mechanics and distinct advantages. MPC utilizes the rapid discharge of energy stored in capacitors to generate a pulsed electromagnetic field, which accelerates a tool to compress the powder. This high-speed process is particularly well-suited for compacting complex geometries and finds extensive application in industries such as powder metallurgy, welding, die forging, and advanced material manufacturing. This paper provides an overview of recent advancements and applications of MPC technology, highlighting its capabilities and potential for broader integration into modern manufacturing processes.

Differentiable pixel-super-resolution lensless imaging

Conventional lensless imaging systems require complex phase diversity measurements and sequential processing steps, limiting their practical application despite their compact design. We present a differentiable end-to-end pixel-super-resolution (dPSR) technique that unifies PSR hologram synthesis, autofocusing, and complex-field reconstruction within a single optimization framework. By jointly optimizing these traditionally separate processes, our method eliminates both phase diversity requirements and error accumulation from sequential processing. Our method achieves superior position estimation accuracy (mean error 0.0282 pixels versus 0.1172 pixels with conventional methods), delivering precise autofocusing with accuracy better than 0.3 µm, and enabling a twofold resolution enhancement beyond the sensor’s native pixel size. This robust performance is validated through both simulated and experimental results, including challenging phase objects and label-free cell imaging, establishing dPSR as a practical solution for high-resolution microscopy applications.

Recovery of Silicon Kerf Through Oxidative Cleaning and Drying Process

High quantity of slurries, kerf, are generated during ingot wafering and represent up to 40% of the silicon in weight. The highly pure silicon contained in kerf is contaminated with organics, metal impurities, water and an oxide layer. The recovery of the silicon could be an efficient way to reduce costs in the production of solar panels through material and energy savings. At high temperatures, organic contamination causes the formation of silicon carbide species that are difficult to segregate during kerf remelting into a pure silicon ingot. Also, during the fusion of silicon, temperatures up to 1500°C are reached and the mixture of silicon and water forms oxides and gases that reduce production yields and affect melting conditions. For these reason, purification and drying steps are required to ensure proper silicon raw material for remelting into ingots. So far, approaches using strong acidic or organic solvent are used to reduce contamination by organic compounds. In this study, more environmentally friendly processes were applied on authentic kerf waste to purify and dry silicon, to make it suitable as raw material for silicon ingot preparation. The amount of water and organic compounds were monitored along the process steps. Using aqueous oxidative processes in combination with filtration, kerf was cleaned and a total carbon reduction of at least 30% was observed. The control of humidity enables the convenient conditioning of purified kerf and effective drying reduced the moisture content to less than 5% wt.

An Integrated Model for Mass Transport, Corrosion Propagation, and Cracking in Offshore Reinforced Concrete Structures

The corrosion of steel reinforcements substantially degrades the longevity of reinforced concrete structures, particularly in marine settings. This investigation introduces a comprehensive model that simulates the processes involved in moisture and chloride ion transport, rebar corrosion, and the consequent cracking of concrete. The model reveals that the transport dynamics of chloride ions are primarily dictated by their penetration rates through the solution. The sensitivity of the steel to corrosion is a function of the concentrations of water and chloride ions, whereas the rate of corrosion predominantly depends on the availability of oxygen at the corrosive site. Oxygen diffusion is the rate-limiting step in the entire process of the electrochemical reactions of the rebar. And the peak corrosion rates are observed at the interface between the solution and the gas phase. The model calculates the stress and strain in the concrete resulting from volumetric expansion due to oxidization of the steel bars. By accurately reproducing the progression of corrosion-related damage, this model provides crucial insights for predicting the service life of offshore concrete structures and enhancing durability against aggressive environmental conditions.

A robust, deep learning-based analysis of time-domain signals for NMR spectroscopy

When analyzing the Free Induction Decay (FID) signal produced by nuclear magnetic resonance (NMR) spectroscopy, Fourier transforms (FT) are used to decompose time-domain signals arising from nuclear interactions. This transformation enables the extraction of frequency-domain information, allowing for the recognition of patterns within the generated NMR spectra. Most modern NMR processing software applies FT to generate the final spectra. Researchers process FID using various techniques, such as phase correction, windowing, and FT, to enhance the interpretation of the obtained spectra. This processing step requires careful consideration of the characteristics of the original data and can also be influenced by the researchers' experience, often making it time-consuming to produce reliable results. However, recent advancements in artificial intelligence, particularly deep learning, have demonstrated superior pattern recognition capabilities compared to humans in complex scenarios. These developments have been successfully applied to various aspects of NMR spectroscopy. In this study, we demonstrate that neural networks can replace FT in NMR spectroscopy, enabling robust and rapid prediction of spectra and peak lists from FID signals. Our results confirm that deep learning can efficiently process NMR data to generate final spectra. As a proof of concept, we present the resulting spectra, along with peak lists predicted by supplying only FID input to the deep learning algorithm. The generated peak lists can be considered as spectra with infinite resolution.

Methodology for quality risk prediction for milk powder production plants with domain-knowledge-involved serial neural networks.

In dairy enterprises, predicting product quality attributes that are influenced by operating parameters is a major task. To reduce quality loss in production, a prediction-based quality control method is proposed in this study. In particular, a serial neural network was designed, and an innovative quality risk prediction methodology based on the integration of SNN and domain knowledge was created. The methodology involves three steps: (1) the processing steps at each unit operation are mapped to a layer of a back propagation network, (2) the branch networks are connected by key quality attributes, and (3) the model is trained with preprocessed data. The experiment was conducted based on milk powder production, demonstrating that the proposed methodology has a higher accuracy and shorter response time compared with those of existing methods. In addition, the practical value of the prediction methodology in actual dairy companies was discussed.

Differentiating low-carbon waste management strategies for bio-based and biodegradable plastics under various energy decarbonization scenarios.

Bio-based and biodegradable (bio-)plastics are heralded as a key solution to mitigate plastic pollution and reduce CO2 emissions. Yet, their end-of-life treatments embodies complex energy and material interactions, potentially leading to emissions through incineration or recycling. This study investigates the cradle-to-grave, emphasizing the waste management stage, carbon footprint for several types of bio-plastics, leveraging both GWP100a and CO2 uptake methods to explore the carbon reduction benefits of recycling over disposal. Our findings indicate that in scenarios characterized by carbon-intensive electricity, using polylactic acid (PLA) as an example, incineration with energy recovery (-1.6316kg CO2-eq/kg, PLA) yields a more favorable carbon footprint compared to chemical recycling (-1.5317kg CO2-eq/kg, PLA). In contrast, in environments with a high proportion of renewable energy, chemical recycling is a superior method, and compared to incineration (-1.4087kg CO2-eq/kg, PLA), the carbon footprint of chemical recycling (-2.0406kg CO2-eq/kg, PLA) are significantly reduced. While mechanical recycling presents considerable environmental benefits, its applicability is constrained by the waste quality, especially in the case of biodegradable plastics like PLA. In addition, the degradation of biodegradable plastics such as PLA was modeled during compost and anaerobic digestion processes. This enables us to quantify the specific biogenic carbon emissions released during these processing steps, revealing the direct emissions with dynamic degradation. This study highlights the importance of tailoring bio-plastic waste management strategies to support global energy decarbonization while understanding their life-cycle carbon metabolism to effectively tackle plastic pollution and climate change.

Combining steric exclusion with anion exchange - development of a universal and scalable adeno-associated virus downstream process.

Adeno-associated viruses (AAV) are among the leading vectors for in vivo gene therapy. The purification of AAV remains a bottleneck as it typically requires multiple individual process steps, often resulting in product loss and high costs. Current downstream processes are usually serotype-specific and rely primarily on expensive affinity resins. To address these limitations, we developed a serotype-independent purification method using steric exclusion chromatography (SXC) that can be combined with a subsequent anion exchange full/empty separation step. This alternative approach eliminates the need for intermediate concentration and buffer exchange, thereby reducing the number of process steps required while achieving high-purity full AAV particles. SXC conditions were optimized using a design of experiments approach. Isocratic separation of full and empty AAV resulted in further purification of the sample. The overall process achieved a viral genome recovery of 51.7 %, along with impurity depletions of 99.9 % for DNA and 99.8 % for protein. The process was successfully adapted to different AAV serotypes and genes of interest, demonstrating its robustness and versatility. In addition, the scalability of SXC was demonstrated, highlighting its potential for large-scale manufacturing. This streamlined, universal, and scalable process provides a robust and efficient alternative to traditional AAV purification processes, addressing critical challenges in gene therapy production and paving the way for broader implementation in research and manufacturing.

Nanocellulose synthesis via synergistic application of solid acid and cellulase.

Nanocellulose stands out in numerous applications due to its excellent properties. Yet, achieving its preparation in a cost-effective, efficient, and environmentally benign manner remains challenging. This study introduces a green synthesis approach by employing a non-polluting solid acid, combined with a cellulase enzyme, for nanocellulose production. We explored the impact of varying the addition sequence of solid acid and cellulase on nanocellulose yield. Experimental results showed that under optimal cellulase hydrolysis conditions, solid acid hydrolysis yielded 41.3% nanocellulose, whereas cellulase hydrolysis resulted in a 27.1% yield. In co-treatment experiments, with the following process steps: 1) sequential hydrolysis with solid acid, followed by cellulose (SA-E); 2) simultaneous hydrolysis with solid acid and cellulase (SA+E); 3) cellulase followed by solid acid hydrolysis (E-SA). the nanocellulose yields were: 73.37%, 70.43%, and 57.21%, respectively. These results proved a synergistic effect between solid acid and cellulase in both SA-E and SA+E scenarios. In addition, the synergistic mechanism between solid acid and cellulase was proposed. This approach presents a highly promising strategy for achieving high yields of nanocellulose.

Perspectives on goal setting: Video-reflexive ethnography with speech-language therapists and clients.

Goal setting is an essential step in the clinical reasoning process of speech and language therapists (SLTs) who provide care for children, adolescents and adults with communication disorders. In the light of person-centred care, shared or collaborative goal setting between the SLT and client is advised in (inter)national guidelines. SLTs face challenges in implementing (shared) goal setting as theoretical frameworks and practical interventions are scarce and less applicable to use with a wide range of communication vulnerable populations. A first step in developing theory and practical interventions is to explore first-hand experiences of SLTs and clients about day-to-day goal-setting practice. This study was guided by the following research question: What are the perspectives and needs of SLTs and persons with communication disorders regarding (shared) goal setting in routine SLT services? The qualitative study was carried out in the setting of routine speech-language therapy services in community practices, primary education and neurological rehabilitation in the Netherlands. Data collection followed the principles of video-reflexive ethnography, using video footage of goal-setting conversations to facilitate semi-structured, reflexive interviews. Data analysis was based on reflexive thematic analysis. A total of 12 interviews were conducted with client-SLT dyads, covering perspectives from children, parents and adults with a range of communication difficulties and their SLTs. Data analysis resulted in four themes, of which two contain subthemes. Each theme represents a central organizing concept found in SLT and client interviews. The themes were identified as: (1) goal setting is a complex process; (2) goal talk needs to be communication accessible; (3) communicative participation goals are hard to grasp; and (4) the importance of relationships. Topics such as power imbalance, communication vulnerability, effective communication strategies, and motivation and trust are explored under these themes. SLTs are encouraged to view shared goal setting as a process that needs to be explicitly planned and communicated with clients regardless of their age or communication vulnerability. SLTs have expert knowledge and skills when it comes to supporting communication and applying these skills during goal talks might strengthen shared goal setting and foster a therapeutic relationship. There is a need to concretely conceptualize and embed shared goal setting in policy and clinical guidelines. The themes reported have tentative clinical implications for developing such policy, and shared goal-setting interventions for SLT practice, under the condition that SLTs and people with communication disorders are continuously involved. What is already known on the subject SLTs want to set meaningful goals together with their patients but lack theory and resources to effectively shape the goal-setting process. Few studies have directly reported on the perspectives and needs of SLTs and patients regarding goal setting. Patients generally perceive goal setting as a vague activity in which they minimally participate. SLTs want to involve patients in the goal-setting process and describe the potential benefits, but they also want to report barriers on the systemic and professional competence level. What this paper adds to the existing knowledge SLTs and patients perceive shared goal setting as a multifaceted process, rather than a one-off conversation. This process holds potential vulnerabilities for SLTs and patients alike, and the themes in this study propose potentially helpful ingredients to mediate this vulnerability and shape the goal-setting process. What are the potential or clinical implications of this work? To take the first steps towards effective shared goal setting, SLTs should embrace the element of discovery in goal setting and apply their expert knowledge in supporting communication. To develop practical interventions for SLTs, shared goal setting needs to be further conceptualized and embedded in policy and clinical guidelines.

AI-based plastic waste sorting method utilizing object detection models for enhanced classification.

The export ban on plastic waste by China has brought domestic plastic recycling to the forefront of environmental concerns, with sorting being a crucial step in the recycling process. This study assessed the performance of advanced AI models, Mask R-CNN, and YOLO v8, in enhancing plastic waste sorting. The models were evaluated in terms of accuracy, mean average precision (mAP), precision, recall, F1 score, and inference time, with hyperparameter tuning performed through grid search. Mask R-CNN, with an accuracy of 0.912 and mAP of 0.911, outperformed YOLO v8 in tasks requiring detailed segmentation, despite a longer inference time of 200-350ms. Conversely, YOLO v8, with an accuracy of 0.867 and mAP of 0.922, excelled in real-time applications owing to its shorter inference time of 80-160ms. This study underscores the importance of selecting the appropriate model based on specific application requirements.

AI-based methods for biomolecular structure modeling for Cryo-EM.

Cryo-electron microscopy (Cryo-EM) has revolutionized structural biology by enabling the determination of macromolecular structures that were challenging to study with conventional methods. Processing cryo-EM data involves several computational steps to derive three-dimensional structures from raw projections. Recent advancements in artificial intelligence (AI) including deep learning have significantly improved the performance of these processes. In this review, we discuss state-of-the-art AI-based techniques used in key steps of cryo-EM data processing, including macromolecular structure modeling and heterogeneity analysis.

Biomass-Derived Hard Carbon Materials for High-Performance Sodium-Ion Battery

Using biomass-derived hard carbon materials as anode materials for sodium-ion batteries has facilitated resource recycling and brought significant economic benefits. However, the main obstacles to the large-scale application of these materials are the low Coulombic efficiency and high irreversible capacity of hard carbon materials. This study used waste moso bamboo as a carbon source to prepare and pre-oxidize hard carbon through a stepped temperature sintering process. The introduction of oxygen atoms into the carbon layers has been shown to increase the spacing between the carbon layers, which facilitates the insertion of sodium-ions into them. Moreover, the presence of oxygen-containing groups increases the number of edge and vacancy defects in the carbon skeleton, thereby enhancing the actual capacity of the material. Studies have indicated that different pre-oxidation times have varying impacts on the electrochemical properties of hard carbon materials. This study used discarded moso bamboo as the raw material, and the optimal pre-oxidation duration of bamboo-based hard carbon was determined to be 4.5 h through a series of comparative experiments. A high-performance biomass-derived hard carbon material was prepared via a stepwise sintering process. It exhibited a specific capacity of 301.4 mAh·g−1 at 0.1 C and a first-cycle Coulombic efficiency of 87%.

3D Printed pH-responsive Colonic Capsules for the Delivery of Live Aqueous Bacterial Suspensions

Delivering live bacterial therapeutics to the colon orally is challenging due to the harsh gastrointestinal (GI) conditions and/or the manufacturing processes involved in the production of dry formulations, which can drastically decrease cell viability. In previous work, we evaluated the performance of a 3D printed pH-responsive capsule capable of encapsulating aqueous cargos. We herein evaluate its ability to encapsulate live bacterial suspensions with limited processing steps. The capsules maintained their integrity in conditions simulating the upper GI tract (stomach and proximal intestine) and only released their contents in the environment of the lower intestine, i.e. ileum and colon. The mean viability of individual or mixed selected strains remained above 75% during a full simulated GI transit to the colon. In beagle dogs, genomic DNA of 2 out of the 3 delivered strains was detected in the feces, and DNA copy levels did not differ between the capsules and the control suspension of non-encapsulated bacteria. These results could be attributed to the differing physiological conditions of fasted beagle dogs vs. the simulated environments, or possibly to a non-optimal assessment of bacterial colonization. A follow-up study after capsule treatment, incorporating sampling from various colonic tissues and fluids, could provide key insights into bacterial colonization.

Recognition of autism in subcortical brain volumetric images using autoencoding-based region selection method and Siamese Convolutional Neural Network

BackgroundAutism Spectrum Disorder (ASD) is a neurodevelopmental condition that affects social interactions and behavior. Accurate and early diagnosis of ASD is still challenging even with the improvements in neuroimaging technology and machine learning algorithms. It’s challenging because of the wide range of symptoms, delayed appearance of symptoms, and the subjective nature of diagnosis. In this study, the aim is to enhance ASD recognition by focusing on brain subcortical regions, which are critical for understanding ASD pathology. MethodologyFirst, subcortical structures were extracted from a collection of brain MRI datasets using sophisticated processing steps. Next, a 3D autoencoder was trained on these 3D images to help identify brain regions related to ASD. Two distinct feature selection methods were then applied to the features extracted from the encoder. The highest-ranked features were iteratively selected and increased to reconstruct a specific percentage of the brain that represents the most relevant parts for ASD. Finally, a Siamese Convolutional Neural Network (SCNN) was employed as the classifier model. ResultsThe 3D autoencoder stage helped in identifying and reconstructing the significant subcortical regions related to ASD. Based on the studied dataset, high agreement in regions like the Putamen and Pallidum indicated the critical nature of these structures in distinguishing Autism from controls cases. Subsequently, applying SCNN on these selected subcortical regions yielded promising results. For example, using the classifier on the output regions identified by the Mutual Information (MI) features selection method achieved the highest accuracy of 0.66. ConclusionsThis study shows that using a two-stage model involving autoencoder and SCNN can notably improve the classification of ASD from brain MRI volumetric images. Applying an iterative feature extraction approach allowed to achieve a more accurate identification of ASD-related brain areas. This two-stage approach not only improved classification performance but also enhanced the interpretability of the neuroimaging data.