The research, carried out in the Ablekuma South District of the Greater Accra Metropolis during the first and last quarters of 2020, forms part of a broader effort funded by the Alliance for Health Policy and Systems Research through the RAISE project to adapt home visit guidelines for Ghana. Researchers from the University of Ghana and partner institutions conducted five focus group discussions with 36 participants, including community health nurses, community health officers and programme managers, alongside three stakeholder engagement meetings involving 46 people such as district health administrators, community volunteers, faith healers and herbal practitioners. All sessions were audio-recorded with consent, transcribed, and independently analysed by three qualitative scientists using the Braun and Clarke thematic approach, with findings validated back to stakeholders.

The results, published in Health Research Policy and Systems, paint a stark picture of physical deprivation. Nurses reported lacking the most basic tools of their trade: blood pressure apparatus, thermometers, weighing scales, cotton wool and first aid kits with essential medications. One nurse described her visiting bag as effectively empty, asking how she could responsibly administer family planning without first checking a client’s blood pressure. Protective clothing was equally absent, with no raincoats, umbrellas or footwear provided for staff who must walk long distances through hilly, rocky and densely populated terrain. When asked who should supply these resources, participants pointed to the government through the district health directorate, while noting that philanthropic support was irregular and that some nurses quietly paid out of their own pockets to fund essential care for clients.

Transportation emerged as a compounding physical barrier. Nurses are assigned large geographical catchment areas to cover repeatedly for months or years, and in rural postings motorbikes, training and fuel stipends are sometimes provided. In Greater Accra, however, heavy traffic and safety concerns mean motorbikes are withheld from community health nurses, who are predominantly women, and vehicles are often reassigned to male staff for non-visit duties. Participants who had worked in other regions described outreach vehicles and transport allowances there, contrasting sharply with the Accra facilities where nothing at all was provided. The inconsistency, they said, is poorly documented and leaves urban nurses dependent on unreliable stipends or their own feet.

Beyond logistics, the study identified a cluster of psychological challenges rooted in mistrust and fear. Clients seeking to avoid follow-up visits provide false addresses, unreachable phone numbers, or directions to locations that do not exist, such as a brown gate just around a corner that never materialises. Some families simply refuse entry, particularly when a newborn is in the house and grandparents wish to continue traditional practices that a nurse would object to. More alarmingly, nurses reported verbal and physical attacks from relatives of clients, stray dog attacks at gated homes, and even sexual assaults on unarmed young female staff by men in the communities. One nurse recounted how a husband threatened to organise boys to beat visiting staff; the visits stopped, and the woman he sought to shield from care later died.

Role conflict adds a professional dimension to these psychological strains. Community health nurses who care for pregnant women at home are expected to record their findings and medications, such as malaria prophylaxis, in antenatal booklets that midwives consider their own record space. Midwives queried nurses who wrote in these books, and nurses eventually stopped, creating fragmented documentation for women receiving care from both cadres. Participants argued that clearer job descriptions and defined boundaries of practice are needed so that midwives understand visiting nurses are supporting, not supplanting, their work.

Sociocultural interferences form the third major barrier. Many community members attribute illness to spiritual causes and exhaust herbal or faith-based remedies before considering medical treatment, with traditional healers referring patients to hospitals only after indigenous treatments fail. Sacred and festive days render entire communities inaccessible on scheduled days, forcing nurses to build their itineraries around fishing days and market days. Religious objections to modern family planning lead women to default on antenatal care and deliver at home without skilled supervision. Traditional newborn practices, including applying hot stones, saliva, ground chalk, gauze and towels to the umbilical cord, and keeping infants hidden until an out-dooring ceremony, directly restrict nurses’ access to the most vulnerable patients. Patriarchal household structures compound this, since male heads of family must often grant permission before health information can reach women and girls.

The solutions proposed by participants are strikingly practical. They called for well-stocked visiting bags, vaccine supplies, transport funding and expanded CHPS coverage, with some suggesting more male nurses on visiting teams so male clients can speak comfortably. For safety, they recommended community maps, walking boots, umbrellas and raincoats, a companion model pairing nurses with trusted community volunteers, walking in groups where no companion is available, and even defensive equipment such as pepper spray or tasers, possibly issued through police partnerships and returned at the end of each day. In-service training and, crucially, a dedicated Ghanaian home visit practice guideline specifying scope, boundaries, referral pathways, service set-up, logistics and human and financial resources were described as the foundation for everything else. The researchers note that the existing Disease Prevention Life Course Approach Model underpinning CHPS contains no clear-cut guidance on home visit activities, leaving staff uncertain and the practice difficult to evaluate.

The study’s authors argue that guidelines developed in high-income countries such as the United States and Canada, while useful, do not address the layered community, family and household gatekeepers, the absent addressing systems, the security realities and the cultural nuances of Ghana. Formative research to tailor training, public education campaigns, male Community Health Ambassadors recruited through men’s fellowship meetings, mixed-gender visiting teams, culturally sensitive uniforms and dedicated communication lines to the police are among the measures they propose. They caution that the findings, drawn from purposively selected participants in one district and vulnerable to recall and participation bias, cannot be generalised, but they believe the lessons apply broadly across low- and middle-income countries in Africa and Asia where patriarchal structures, poor road networks and weak addressing systems similarly constrain community health work. With dedicated, culturally grounded guidelines, they conclude, home visiting in Ghana could finally be prepared, protected and accountable enough to deliver on its century-old promise.

Subject of Research: Barriers to and solutions for nurse-led home visit practice in Ghana

Article Title: Challenges of home visit practice and perceived solutions for improvement in Ghanaian context: perspectives of healthcare professional and community informants

Article References: Ohene, L. A., Adjorlolo, S., Chandi, M. G., Aryeetey, C., Ansah- Ofei, A. M., Aikins, M., & Aziato, L. (2026). Challenges of home visit practice and perceived solutions for improvement in Ghanaian context: perspectives of healthcare professional and community informants. Health Research Policy and Systems, 24(S1), Article 72. https://doi.org/10.1186/s12961-026-01461-w

Image Credits: AI Generated

DOI: 10.1186/s12961-026-01461-w

Keywords: home visits, community health nurses, Ghana, primary healthcare, CHPS, practice guidelines, universal health coverage, sociocultural barriers, nurse safety, qualitative research, Challenges, home

The review, led by Yosra Didi together with Ahlem Walha and Ali Wali, does not simply catalog architectures and accuracy figures. Instead, the authors organize the literature through a structured five-dimensional qualitative analytical framework covering methodological rigor, clinical relevance, technical validity, reproducibility evidence and imaging-specific considerations. Crucially, the team is transparent that this framework is a lens for structured qualitative comparison rather than a quantitative scoring system, a deliberate choice that reflects how difficult it is to compare studies whose experimental designs differ so widely. By attending explicitly to methodological context, the review aims to explain why results that look comparable on paper may not be comparable in practice.

Technically, the field the authors survey is dominated by convolutional neural networks, the deep learning family that has become the workhorse of medical image analysis. The reviewed studies trace a clear evolutionary arc: early work relied on classic backbones such as VGG, ResNet, Inception and Xception, typically fine-tuned on ImageNet pre-trained weights through transfer learning, while more recent efforts embrace efficient architectures like MobileNet and EfficientNet for deployment-friendly inference, capsule networks for preserving spatial hierarchies, and vision transformers such as the Swin transformer that replace convolutional inductive biases with attention-based global context. Ensemble methods, which combine predictions from multiple networks to squeeze out additional accuracy, appear repeatedly, as do hybrid pipelines that pair deep feature extractors with classical classifiers such as support vector machines, k-nearest neighbors and random forests.

Alongside pure classification, the review documents a substantial strand of work built on segmentation networks, including U-Net and V-Net, that first delineate lungs, heart borders or lesion regions before classification proceeds. Studies of cardiomegaly detection, tuberculosis localization and COVID-19 severity assessment illustrate why this matters: segmenting the anatomical structure of interest can focus a model’s attention on clinically relevant tissue and reduce the risk that a network learns shortcuts from irrelevant image regions. Attention mechanisms and explainable AI overlays, such as probability maps and saliency-based visualizations, are increasingly bolted onto these pipelines in an effort to make black-box predictions legible to clinicians, though the review notes that interpretability claims are rarely tested against actual radiologist reasoning.

The data underpinning all of this is a familiar roster of public benchmark datasets: ChestX-ray8 and its successors, CheXpert, PadChest, MIMIC-CXR, VinDr-CXR, the Montgomery and Shenzhen tuberculosis collections, the Kaggle pneumonia archives, and a wave of COVID-19 repositories assembled during the pandemic, including the COVID-19 Radiography Database, BIMCV-COVID-19+, and the RSNA Ricord releases. The authors find that studies evaluated on a single dataset report systematically more optimistic performance than those subjected to multi-dataset evaluation regimes, a pattern consistent with models overfitting to the imaging characteristics, label conventions and patient populations of one institution. When the same architecture is tested across hospitals, scanners and countries, performance differences between datasets emerge that single-cohort experiments never reveal.

Underlying dataset limitations compound the problem. Many widely used chest X-ray corpora carry labels derived automatically from radiology reports rather than from independent expert adjudication, introducing label noise that caps achievable accuracy in ways rarely quantified. Class imbalance, uncertain and missing labels, and demographic and geographic skew all recur across the reviewed corpus, and the reviewers highlight that such weaknesses directly affect diagnostic reliability. Data augmentation, including generative approaches based on convolutional GANs and techniques such as CLAHE contrast enhancement, is frequently deployed to compensate, but the review observes that augmentation choices are rarely ablated systematically, leaving it unclear how much of a reported gain comes from the model and how much from the preprocessing recipe.

Perhaps the most damning findings concern experimental hygiene. Across a substantial share of the 81 studies, the reviewers document inconsistent experimental design choices, non-standardized preprocessing pipelines and limited ablation reporting, making it genuinely difficult to determine which technical innovations actually drive performance. Preprocessing steps ranging from resizing and normalization to lung cropping and denoising vary from paper to paper, sometimes within evaluations of the same architecture, and hyperparameter details are often underreported. The reproducibility picture is similarly uneven: code and trained weights are infrequently released, and the review points to established reporting standards, such as the CLAIM checklist for artificial intelligence in medical imaging, as tools that remain underused despite being designed precisely to address these gaps.

The scarcity of external clinical validation emerges as the review’s central concern. Very few of the surveyed systems have been tested on data from institutions absent from training, and fewer still have been evaluated in genuine clinical workflows with radiologists in the loop. The authors draw an implicit contrast with landmark analyses, such as the widely cited Nature Machine Intelligence study cataloging common pitfalls in COVID-19 machine learning research, which documented how data leakage and flawed study design produced models that collapsed outside their original datasets. Against that backdrop, the pattern the Tunisian team identifies, apparent performance differences between single-dataset and multi-dataset regimes paired with limited external validation, reads as a warning that laboratory excellence has not yet translated into deployable trust.

Still, the review is constructive rather than dismissive. Based on architectural cross-comparisons and a qualitative audit of reproducibility practices, the authors consolidate recommendations for the field: adopt multi-dataset and external validation as the default rather than the exception, standardize preprocessing and reporting so that results can be meaningfully compared, report ablations that isolate the contribution of each design choice, and design studies around clinical relevance from the outset, including the disease prevalence, co-morbidities and imaging conditions of the intended deployment setting. They also flag the open challenges that will define the next phase of the field, chief among them generalization across populations and scanners, evaluation standardization, and the practical integration of these models into radiology workflows.

The stakes could hardly be higher. Chest radiography remains one of the most performed imaging examinations in the world, and in many health systems, particularly in low- and middle-income countries where the review’s tuberculosis-focused literature is concentrated, a reliable AI triage tool could extend diagnostic capacity where radiologists are scarce. The review’s authors, who report no funding and no conflicts of interest, analyzed only publicly available published studies and generated no new datasets of their own. Their message to the community is ultimately one of disciplined optimism: the technical machinery for AI-assisted chest X-ray interpretation is mature and improving rapidly, but the evidence base must be rebuilt on foundations of external validation, transparent reporting and standardized evaluation before these systems can responsibly share the reading room with human experts.

Subject of Research: AI-driven thoracic disease classification in chest radiography

Article Title: AI-driven techniques for thoracic disease classification in chest radiography: A comprehensive review and critical analysis

Article References: Didi, Y., Walha, A., & Wali, A. (2026). AI-driven techniques for thoracic disease classification in chest radiography: A comprehensive review and critical analysis. Neural Computing and Applications, 38(17), Article 737. https://doi.org/10.1007/s00521-026-12457-6

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12457-6

Keywords: artificial intelligence, deep learning, chest X-ray, thoracic disease classification, convolutional neural networks, medical imaging, pneumonia detection, tuberculosis, COVID-19, transfer learning, clinical validation, reproducibility

The methodological design of the study is itself notable for its rigor in a field where such designs are often missing. The researchers used a sequential mixed-methods approach, labeled QUAN-qual, meaning the quantitative phase came first and directly informed the qualitative phase that followed. In the quantitative phase, the team surveyed 71 faculty members who held postgraduate training in medical education, drawn from institutions across Pakistan, and achieved a striking 93.4 percent response rate. Such a high rate matters because it reduces the risk that the survey reflects only the views of the most motivated or most frustrated respondents. The qualitative phase then used two focus group discussions, one with junior faculty and one with senior faculty, to dig beneath the survey numbers and understand the lived experience behind them.

The results reveal a generational split in the obstacles that matter most. Among junior faculty, the primary challenges in conducting medical education research were lack of time, lack of training, and lack of support. These are, in many respects, the classic barriers reported by early-career researchers everywhere, but in Pakistan they take on a particular intensity because medical educationists often carry heavy teaching and administrative loads with no protected time carved out for scholarship. For senior faculty, the picture shifted: their main challenges were lack of funding for publication, the absence of well-reputed local journals devoted to medical education research, and a pervasive culture of mistrust that undermines collaboration. In other words, even those who had survived the early career bottleneck found themselves facing a publishing ecosystem and a professional culture that offered little reward or scaffolding.

The qualitative analysis distilled these findings into overarching themes that applied to both junior and senior faculty. At the systemic level, participants described the absence of protected time and inadequate institutional support as structural features of their working lives, not temporary inconveniences. At the individual level, they acknowledged gaps in their own preparation, including insufficient training in research methods and in the theoretical frameworks that give educational scholarship its analytical backbone. This dual diagnosis is important because it suggests the problem cannot be fixed by training alone or by institutional reform alone; the barriers operate at multiple levels simultaneously and reinforce one another. A well-trained educationist with no protected time produces nothing, while an educationist with generous time but no methodological grounding produces work that journals will not accept.

The context in which these findings sit is a global one. Medical education research is overwhelmingly produced in high-income countries, with limited contribution from low- and middle-income countries such as Pakistan. This asymmetry matters for science as a whole because educational questions in Lahore or Karachi differ in funding realities, class sizes, student demographics, and cultural expectations from those in Boston or London. When the evidence base for how to teach medicine is built almost entirely in resource-rich settings, its findings travel poorly to the places where most of the world’s medical students actually study. Pakistan, as the study’s authors note, has made departments of medical education mandatory in every medical and dental college, creating a large pool of trained professionals whose insights are, in principle, exactly what the global literature lacks.

The study was conducted with formal ethical oversight, receiving approval from the Ethics Review Committee at Aga Khan University Hospital under study ID 2022-7738-23209, with informed consent obtained from all participants and the research conducted in accordance with the ethical principles of the Declaration of Helsinki. The researchers report that the work received no specific grant from any funding agency in the public, commercial, or non-profit sectors, a detail that is perhaps telling in itself: even a study about the lack of research funding had to proceed without it. The authors declare no competing interests, and the article is published open access under a Creative Commons license, making the full findings freely available to the very community of educationists whose circumstances it documents.

What makes the study’s conclusions actionable is the specificity of its recommendations. The authors argue that scholarship in medical education in Pakistan is hindered by individual, institutional, and systemic factors, and they direct their prescriptions accordingly. Institutions, they conclude, need to focus on providing protected research time, bolstering training in medical education research, and establishing robust mentorship and collaborative networks to strengthen the research culture. Protected time, in particular, is a recurring theme in faculty development research worldwide, but the Pakistani data give it local weight: when 93 percent of trained educationists in a national survey point to time as a barrier, the case for institutional scheduling reforms becomes difficult to ignore.

The mistrust that senior faculty identified as a barrier to collaboration deserves particular attention, because it points to a problem that no amount of funding can solve on its own. Collaborative research depends on norms of credit-sharing, transparent authorship, and confidence that intellectual contributions will be respected. Where those norms are weak, researchers retreat into silos, small studies go unpublished, and the cumulative building of knowledge stalls. Similarly, the absence of well-reputed local journals means Pakistani educationists face a choice between competing in high-income-country journals with rejection rates shaped by different priorities, or publishing in venues with limited visibility. A credible regional publication infrastructure, the findings imply, is not a vanity project but a structural prerequisite for a self-sustaining research community.

The implications extend well beyond Pakistan’s borders. The World Health Organization and international medical education bodies have repeatedly emphasized that health workforce education in low- and middle-income countries must be evidence-driven rather than imported wholesale from contexts that differ fundamentally. Studies like this one provide the diagnostic foundation for that ambition, identifying exactly where the pipeline from trained educationist to published researcher breaks down. The senior authors, including Naveed Yousuf, and co-author Syed Moin Ali of Aga Khan University’s Department for Educational Development, along with Rafay Iqbal of the university’s Department of Family Medicine, frame their work as a mapping exercise, and the map they produce is one that institutional leaders, policymakers, and international funders can all read. For junior faculty wondering whether their struggles are personal failings, and for deans wondering why their education departments generate so little publishable output, the answer documented here is neither indolence nor incapacity, but the accumulated weight of missing time, missing training, missing money, missing journals, and missing trust.

Beyond its specific findings, the study offers a useful illustration of how research capacity itself can be built in resource-constrained settings. The near-complete survey response suggests that Pakistani medical educationists were eager to have their professional difficulties documented, an appetite for engagement that institutions could harness. The sequential design also demonstrates a pragmatic model for LMIC researchers: a modest survey followed by focused group discussions can generate publishable, policy-relevant evidence without the large budgets that often deter early-career scholars in such settings.

The timing of the work is also worth noting. The article was released as an accepted manuscript ahead of its final version of record, a publishing route that shortens the lag between peer review and public availability. For a field whose participants explicitly identified slow, difficult publication as a barrier, faster dissemination models have practical relevance. The open access license further means the findings can circulate freely among the very institutions expected to act on them, including smaller colleges that may lack journal subscriptions.

There is also a measurement lesson embedded in the study. By separating junior and senior faculty experiences, the researchers avoided treating the workforce as a monolith, revealing that career stage changes which barriers dominate. That distinction has direct implications for intervention design: mentorship schemes and protected time may matter most early on, while funding access, journal infrastructure, and trust-building become decisive later. Capacity-building programs imported from high-income countries frequently assume a uniform faculty trajectory, and this evidence argues for tailoring support to career stage. The study thus functions both as a diagnosis of Pakistan’s medical education scholarship landscape and as a template for how other low- and middle-income countries might map their own.

Subject of Research: Mapping the landscape: current status, challenges, and perspectives on scholarship of medical educationists in Pakistan

Article Title: Mapping the landscape: current status, challenges, and perspectives on scholarship of medical educationists in Pakistan

Article References: Sabzwari, S. R., Ali, S. M., Iqbal, R., & Yousuf, N. (2026). Mapping the landscape: current status, challenges, and perspectives on scholarship of medical educationists in Pakistan. BMC Medical Education. https://doi.org/10.1186/s12909-026-10078-0

Image Credits: AI Generated

DOI: 10.1186/s12909-026-10078-0

Keywords: Mapping, landscape, current, status, challenges, perspectives, scholarship, medical, educationists, Pakistan, scientific research

The work began with a survey of 112 people with blindness or moderate-to-severe visual impairment. Their responses identified three features as central to an effective outdoor navigation aid: turn-by-turn directions, guidance that helps users remain on a walkable path and the ability to detect and avoid obstacles. Existing technologies commonly provide only part of this package. A white cane offers immediate information about nearby ground-level hazards but does not provide satellite-based route instructions. Guide dogs can help users avoid obstacles and follow routes, but they require extensive training and can be costly or unavailable. Conventional electronic travel aids and smartphone mapping applications may deliver location information while offering limited support for the physical details of the environment directly ahead.

Mobilio was developed to address these gaps through a smartphone-based system that uses machine-learning models, multiple onboard sensors and personalized audio feedback. Rather than treating navigation as a single problem, the application divides it into several connected tasks. Satellite positioning and digital maps can indicate where a person is located and which direction a route should take. Motion sensors, including accelerometers and gyroscopes, can help estimate how the phone and user are moving. The smartphone camera can provide visual information about the surroundings, while other sensors contribute data about orientation and movement. Sensor-fusion algorithms combine these streams, helping the system produce a more stable estimate of the user’s position, heading and immediate navigation context than any one sensor could provide alone.

The distinction between route planning and real-time guidance is crucial. A map can determine that a user should turn left at a particular intersection, but it cannot by itself guarantee that the person is aligned with the correct sidewalk or has avoided a tree, signpost or other object. Mobilio therefore combines broader route instructions with localized path guidance and obstacle detection. Machine-learning models interpret sensor and camera data to identify elements of the walking environment, while the application translates the results into spoken or otherwise audible cues. The researchers designed those cues to be personalized, adjusting how information is delivered so that users receive useful guidance without being overwhelmed by a constant stream of alerts.

Personalization is particularly important for audio-based navigation. A system that speaks too often can mask environmental sounds, increase mental workload or make it difficult to distinguish urgent warnings from routine directions. A system that speaks too rarely may leave users uncertain about whether they are still on course. Mobilio’s audio feedback was evaluated with participants who had blindness or visual impairment, allowing the researchers to examine whether the instructions were understandable and intuitive in practice rather than merely accurate in laboratory tests. The application was intended to communicate the direction and timing of navigation actions while also warning users about obstacles and changes in the path ahead.

Before testing the application with users, the researchers conducted engineering experiments in representative navigation scenarios. These tests assessed the reliability of the smartphone sensors and the machine-learning models that Mobilio depends on. Such testing is important because outdoor navigation systems operate in conditions that can disrupt positioning and perception. Satellite signals may become less precise near buildings, sensor readings can drift as a phone moves, and streets contain an enormous variety of obstacles, surfaces and layouts. Reliable navigation therefore requires algorithms that can combine imperfect measurements and respond quickly when the environment differs from what a digital map predicts.

The user experiments involved 14 participants with blindness or visual impairment. In one assessment, participants used Mobilio together with a white cane to navigate an outdoor community path. Their performance was compared with a conventional arrangement involving Google Maps and a white cane. When using Mobilio, participants completed the route in 13 ± 3 percent less time and made 41 ± 5 percent fewer contacts with environmental objects. These contacts can include brushing against or striking objects in the walking environment, and reducing them may indicate that the system helped participants anticipate hazards more effectively. The results suggest that adding real-time, AI-supported guidance to an established mobility tool can improve both efficiency and physical interaction with the environment.

The researchers also evaluated Mobilio on an obstacle course and compared its performance with navigation supported by a human guide. In the reported experiments, the application achieved outdoor navigation reliability similar to that of the guide. That finding does not mean a smartphone can replace every form of human assistance, nor does it remove the need for established mobility skills. Instead, it indicates that the system was able to provide consistent guidance across the tested situations while allowing users to control their own movement. Participant surveys further reported that Mobilio was easy to use, imposed a low perceived workload and delivered audio feedback that felt intuitive.

Despite the promising findings, the study represents an early evaluation rather than a complete demonstration of universal reliability. The participant group was relatively small, and the experiments took place in specific outdoor settings and controlled obstacle-course conditions. Real-world travel can involve severe weather, crowded sidewalks, construction zones, unusual road layouts, inconsistent pavement, bicycles and vehicles, as well as temporary obstacles that are not represented on maps. Smartphone-based systems also depend on battery life, sensor availability, data processing and the quality of the user’s device. Future research will need to test Mobilio with larger and more diverse populations, across cities and environmental conditions, while examining how performance changes during long journeys and unexpected events.

Even with those limitations, the study highlights a shift in assistive technology: smartphones are increasingly becoming platforms that can interpret the physical world rather than simply display information about it. By combining machine learning with sensor fusion and adaptive audio, Mobilio attempts to bridge the gap between a digital map and the immediate realities of walking outdoors. For people with blindness or visual impairment, the potential benefit is not only faster travel but greater confidence and independence. The researchers’ results suggest that a carefully designed AI navigation assistant, used alongside tools such as a white cane, could make outdoor journeys safer, more efficient and less demanding, turning a widely available consumer device into a sophisticated mobility aid.

Subject of Research: AI-driven smartphone navigation for people with blindness and moderate-to-severe visual impairment

Article Title: Improving outdoor navigation for people with blindness using an AI-driven smartphone application and personalized audio guidance

Article References: Liu, R., Slade, P. Improving outdoor navigation for people with blindness using an AI-driven smartphone application and personalized audio guidance. Nature Biomedical Engineering (2026). https://doi.org/10.1038/s41551-026-01772-x

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41551-026-01772-x

Keywords: blindness, visual impairment, outdoor navigation, artificial intelligence, machine learning, sensor fusion, smartphone application, assistive technology, personalized audio guidance, obstacle detection

The result is particularly striking because semiconductor manufacturing generally depends on extremely pure starting materials. Even small concentrations of unwanted elements, structural defects or irregularities in crystal growth can alter how a semiconductor transports electrical charge or responds to light. These concerns have traditionally made low-grade mineral resources appear incompatible with sophisticated electronic applications. Minnesota, however, possesses one of the largest iron resources in the United States, producing approximately 75 percent of the nation’s iron ore and generating more than $4 billion in annual revenue. The state’s Mesabi Iron Range has supported iron production for decades, while processing innovations such as taconite beneficiation transformed previously uneconomical rock into a major industrial resource.

The University of Minnesota researchers began with a question that challenged a common assumption in pyrite research: whether the material used in electronic experiments truly needed to be highly purified before it could exhibit useful semiconductor behavior. “We realized that pyrite’s really not like a typical semiconductor—it is surprisingly immune to impurities,” said Chris Leighton, a Distinguished McKnight University Professor in the Department of Chemical Engineering and Materials Science and senior author of the study. That unusual tolerance encouraged the team to test whether iron ore taken directly from Minnesota’s Iron Range could be converted into pyrite without the extensive purification steps normally associated with semiconductor materials. The experiments produced a result that initially surprised the researchers: the impurities present in the ore did not prevent the formation of high-quality semiconducting iron sulfide.

The team tested three different grades of iron ore and found that Direct Reduced Grade Taconite produced the most promising results. This grade is among the commonly available iron-ore materials in Minnesota and is typically processed for iron production rather than for electronic applications. In the researchers’ approach, the ore was converted into pyrite through relatively simple processing. The work showed that the low-purity starting material could yield crystals with semiconductor-relevant properties without an additional purification stage. While the study does not claim that the ore can immediately be used to manufacture commercial devices, it demonstrates that the mineral resource itself does not impose the fundamental barrier that scientists had expected.

Pyrite is chemically composed of iron and sulfur, written as FeS₂, two elements that are widely available compared with many materials used in advanced electronics. Its ability to absorb light intensely is one of its most attractive features. A thin layer of a strongly absorbing semiconductor can, in principle, interact with incoming sunlight or other forms of electromagnetic radiation without requiring a large quantity of material. Pyrite also has an electronic band structure that allows it to participate in charge-generation and charge-transport processes, although controlling defects, interfaces and electrical contacts remains essential for practical devices. The material’s abundance and low production cost could make it appealing for applications where expensive or scarce semiconductor compounds limit large-scale deployment.

The researchers’ discovery also offers a new perspective on the role of impurities in functional materials. In many semiconductors, impurities introduce energy states inside the electronic band gap, acting as traps that capture charge carriers or encourage unwanted recombination. These effects can reduce conductivity, shorten carrier lifetimes and lower the efficiency of devices such as solar cells. Pyrite appears to behave differently under the conditions examined by the Minnesota team. The researchers found that the chemical and structural characteristics of pyrite allow it to remain semiconducting even when it is produced from an ore containing components that would normally be regarded as problematic. Understanding why this occurs could help scientists deliberately control pyrite’s composition rather than treating every impurity as an obstacle.

The implications extend beyond the laboratory because the starting material is linked to an established regional mining and processing industry. If future studies confirm that additional grades of Iron Range ore can be converted into useful pyrite, the approach could create new value from resources already extracted at large scale. It could also reduce the energy, chemicals and cost associated with purifying iron before using it in an electronic material. Such an outcome would not automatically make pyrite-based technologies sustainable; mining, sulfur processing, waste management and device manufacturing would still require careful environmental assessment. Nevertheless, using abundant local feedstocks and avoiding extra purification could improve the economic and environmental profile of future material-production pathways.

Potential uses remain at an early research stage, but the researchers identify several areas where semiconductor-quality pyrite could be investigated. In solar panels, its strong optical absorption could be useful in thin-film architectures, provided scientists can improve charge collection and long-term performance. In batteries, iron sulfide chemistry may offer pathways to electrodes based on widely available elements. Pyrite could also be examined in electronic components, sensors and systems designed to remove contaminants from water, where its surface and electronic properties may support chemical or photo-assisted reactions. The next step is to move beyond bulk pyrite crystals and fabricate thin films, which are more directly relevant to real devices. Thin films will allow the team to study interfaces, thickness effects, defects and electrical contacts under conditions closer to those found in working technologies.

The study was conducted by Leighton, graduate student Yeon Lee and Caitlyn Komar of the University of Minnesota’s Department of Chemical Engineering and Materials Science, together with Jennifer T. Mitchell of the University Characterization Facility and Department of Earth and Environmental Sciences, and Matt Mlinar, Jestos Taguta and George Hudak of the University of Minnesota Natural Resources Research Institute. The work was funded by Minnesota’s Environment and Natural Resources Trust Fund, following recommendations from the Legislative Citizen Commission on Minnesota Resources. It was completed in collaboration with the University of Minnesota Characterization Facility and the Minnesota Nano Center. The researchers plan to test more iron-ore grades and clarify how their mineral compositions influence pyrite formation. For now, the central message is both simple and provocative: a material once dismissed as too impure for advanced electronics may contain precisely the chemistry needed to make abundant, low-cost semiconductor research more practical.

Subject of Research: Converting low-purity Minnesota iron ore into semiconductor-quality pyrite (iron sulfide, FeS₂) for potential applications in electronics, solar panels, batteries and water purification.

Article Title: Semiconductor-quality pyrite FeS2 from iron ore

News Publication Date: 14-Aug-2026

Web References: Physical Review Applied article; DOI: 10.1103/6twd-lvvg

References: University of Minnesota Twin Cities; Physical Review Applied, “Semiconductor-quality pyrite FeS2 from iron ore,” published 13-Aug-2026.

Image Credits: Kalie Pluchel, University of Minnesota

Keywords

Pyrite, iron sulfide, FeS₂, iron ore, Minnesota Iron Range, taconite, semiconductors, electronics, thin-film materials, solar panels, batteries, sustainable materials, mineralogy, materials science, University of Minnesota