The review examines publicly available mammography datasets and their influence on the development of machine-learning and deep-learning systems for classification, lesion detection and segmentation. It covers conventional feature-based methods, convolutional neural networks, transfer learning, object detectors, segmentation networks, Vision Transformers, multimodal systems, explainable AI and privacy-preserving approaches. A structured narrative search covered studies published from 2016 through 2025, with preference given to research using public mammography collections. The authors also assessed each dataset qualitatively for scale, modality, annotation availability and label richness; those suitability ratings were not experimental performance scores.

Public collections have expanded markedly over three decades, but their capabilities remain uneven. MIAS, released in 1994, contains 322 film mammograms from 161 patients with image-level labels and has become a simple benchmark for conventional algorithms and lightweight convolutional networks. DDSM and its curated version CBIS-DDSM provide larger collections with pathology information, lesion regions and, in the curated dataset, segmentation masks. INbreast contains 410 high-quality full-field digital mammograms from 115 patients with pixel-level masks, making it valuable for lesion detection and segmentation research but too small for training large models from scratch. These older resources also rely heavily on digitized film or geographically limited populations.

More recent repositories offer scale, richer labels and, in some cases, clinical context. VinDr-Mammo contains 5,000 Vietnamese full-field digital examinations and 20,000 images with BI-RADS assessments and bounding boxes. CMMD includes 3,712 images from 1,775 Chinese patients, histopathology labels and molecular-subtype information, although it lacks region-of-interest annotations. OMI-DB provides a very large UK collection with lesion annotations, clinical metadata and longitudinal screening information, but access is restricted. DMID, derived from an Indian population, includes 510 full-field digital images with lesion and clinical metadata and remains comparatively limited and evolving. Together, these resources have enabled more ambitious research, but no single public collection combines large scale, broad demographics, standardized annotation, comprehensive clinical metadata and open access.

Annotation depth determines which questions an AI system can learn to answer. Whole-image labels support image classification but provide no clue about where a suspicious finding lies. Bounding boxes identify approximate lesion locations for detectors such as YOLO and Faster R-CNN, while pixel-level masks delineate boundaries for segmentation models such as U-Net and Mask R-CNN. Multi-level annotations can support systems that jointly classify, locate and outline lesions. Standardized BI-RADS descriptors, covering findings such as lesion shape, margins, calcifications and breast density, can add consistent clinical semantics. Yet annotation formats and labeling practices differ across collections, and inter-observer variability introduces label noise. Expert-drawn pixel masks are also labor-intensive, limiting the scale of precisely annotated data.

The technical trajectory has moved from manually engineered texture, shape and statistical features to end-to-end deep networks. Convolutional neural networks learn hierarchical image representations and remain the dominant approach, while transfer learning reuses pretrained weights to cope with smaller labeled collections. YOLO detectors favor speed and real-time inference, whereas Faster R-CNN generally provides stronger localization at greater computational cost. U-Net variants produce detailed lesion boundaries but depend on scarce pixel-level annotations. Vision Transformers use self-attention to capture long-range spatial relationships, and hybrid CNN–Transformer models combine local texture with global context. The review notes, however, that transformers usually require more training data than CNNs. Ensembles and multi-stage systems can integrate several tasks, but their computational complexity may complicate routine clinical use.

High benchmark results do not resolve the central generalization problem. The reviewed literature contains many reports of accuracy above 90%, and some studies describe values above 98% on benchmark collections. Such figures concern the datasets, preprocessing and evaluation protocols used in those studies and do not establish equivalent performance in hospitals with different equipment, acquisition protocols, patient populations or annotation practices. Models evaluated on an independent dataset can perform less well, particularly when breast density, disease distribution, image characteristics or preprocessing differ. Transfer learning, domain adaptation and ensemble methods may reduce these shifts, but the review says they do not eliminate dataset bias, class imbalance and limited interpretability.

Geographic concentration compounds the problem. Much of the research literature relies on North American and European collections, particularly DDSM, CBIS-DDSM, MIAS and INbreast. Public data from Africa, South America and parts of Asia remain scarce, including in India despite its substantial breast-cancer burden. DMID is an important step toward regional representation, but its institutional and geographic coverage is not yet sufficient. Private hospital datasets may contain richer clinical information and greater patient diversity, yet restricted access prevents reproduction and standardized comparison. This imbalance makes it difficult to determine whether a model trained in one population will retain useful sensitivity and specificity in another, or whether domain differences could increase false-positive or false-negative predictions.

The review therefore calls for a comprehensive, geographically diverse mammography benchmark assembled across institutions and healthcare settings. Such a resource would need consistent BI-RADS labels, lesion localization, segmentation where feasible, pathology and follow-up information, and clearly documented data splits and evaluation metrics. Privacy-preserving collaboration through federated learning could allow institutions to contribute to model development without centralizing sensitive patient data. Prospective, multicenter validation would also be needed to test whether benchmark performance survives contact with routine clinical workflows. The central message is that AI-assisted mammography may become more capable through transformers, multimodal learning and explainable systems, but trustworthy deployment depends on data that represent the people and conditions the software is intended to serve.

Subject of Research: How dataset quality, demographic bias and annotation standards affect AI-based mammography for breast cancer detection

Article Title: Mammography datasets and artificial intelligence for breast cancer detection: challenges, bias, and future directions

Article References: Mohideen, A. K., Prakash, P. N. S., & Priyadharshini, M. K. (2026). Mammography datasets and artificial intelligence for breast cancer detection: challenges, bias, and future directions. Frontiers in Artificial Intelligence, 9, Article 1933770. https://doi.org/10.3389/frai.2026.1933770

Image Credits: AI Generated

DOI: 10.3389/frai.2026.1933770

Keywords: mammography, breast cancer, artificial intelligence, medical imaging, dataset bias, deep learning, Vision Transformers, annotation standards, external validation, federated learning, datasets, artificial

The review, led by Thanuja M and Soumyashree M. Panchal of the Manipal Institute of Technology Bengaluru together with Prasanna Lakshmi G S of Sai Vidya Institute of Technology, followed the PRISMA 2020 guidelines that govern rigorous evidence synthesis. The team searched PubMed, IEEE Xplore, Scopus, Web of Science, and other major databases, screening 1,247 records before settling on the 83 studies that met their inclusion criteria. Each study was then scored with a customized quality-assessment tool covering data sources, preprocessing clarity, model development rigor, evaluation practices, external validation, reproducibility, and interpretability. The verdict was sobering: only 32 studies, or 38.6 percent, showed a low risk of bias. Another 49.4 percent carried moderate risk, and 12 percent were judged high risk. In other words, nearly two-thirds of the published evidence base for AI-based ECG diagnosis rests on studies with methodological weaknesses that could inflate or distort their reported performance.

The electrocardiogram itself is a deceptively simple signal. Electrodes on the skin capture the heart’s electrical activity, producing the familiar P wave, QRS complex, and T wave that correspond to atrial depolarization, ventricular depolarization, and ventricular repolarization. Clinicians read these waveforms for subtle clues: a slightly elevated ST segment may signal an evolving myocardial infarction, an irregularly irregular rhythm points to atrial fibrillation, and a prolonged QT interval suggests vulnerability to dangerous arrhythmias. These are pattern-recognition tasks that demand years of training and remain subject to variability between observers. Automating them has been a dream almost as old as the ECG itself, but early rule-based systems, which measured predefined features and compared them against fixed criteria, plateaued well below expert human performance because they could not capture the holistic, nuanced pattern recognition that experienced interpreters develop.

Machine learning changed the equation by learning features directly from data rather than relying on hand-crafted measurements. The review found that traditional machine learning methods, used in 49.4 percent of the studies surveyed, still hold real advantages in specific settings. Ensemble classifiers such as Random Forest and Gradient Boosting consistently outperform single classifiers by 4 to 7 percent in accuracy, and expert-derived features grounded in clinical knowledge of ECG morphology add another 5 to 10 percent compared with generic statistical features. Traditional approaches also demand far less training data, making them the sensible choice when fewer than 1,000 recordings are available, and they offer inherent interpretability through feature importance scores and coefficient magnitudes. But on raw or minimally preprocessed signals, they trail deep learning by 8 to 15 percent, a gap that explains why neural networks have come to dominate the field, appearing in 78 percent of studies from 2023 to 2025.

Among deep architectures, convolutional neural networks reign supreme, employed in 67 percent of the reviewed studies. CNNs apply learnable filters across the one-dimensional ECG signal, automatically detecting local morphological patterns such as QRS complexes and ST-segment changes, with pooling layers providing translation invariance so that clinically relevant features are recognized regardless of their exact timing. Their weakness is temporal: they struggle to capture dependencies spanning multiple heartbeats. That role falls to Long Short-Term Memory networks, a form of recurrent architecture whose internal memory gates preserve information across hundreds of time steps, making them well suited to rhythm analysis and to catching paroxysmal atrial fibrillation hidden in long recordings. Hybrid CNN-LSTM models, which pair convolutional feature extraction with temporal modeling, appeared in 34 percent of recent studies and consistently outperform standalone architectures by 3 to 8 percent on benchmarks, particularly for conditions involving both morphological abnormalities and rhythm disturbances. Emerging tools, including transformer self-attention models in 15 percent of 2024 and 2025 studies and graph neural networks that model relationships between ECG leads, are pushing the frontier further, while transfer learning from large repositories such as PTB-XL, which contains 21,799 recordings, cuts training data requirements by 60 to 80 percent.

Before any model sees data, however, the signal must be cleaned, and the review devotes detailed attention to this unglamorous but critical stage. Raw ECG recordings are contaminated by baseline wander, powerline interference at 50 or 60 hertz, muscle activity, respiratory artifacts, and motion noise from patient movement or electrode displacement. Traditional frequency-selective filters, including bandpass filters with cutoffs between 0.5 and 150 hertz and Butterworth designs prized for their maximally flat response, remain workhorses, but newer techniques are gaining ground. Wavelet transforms decompose the signal in time and frequency simultaneously, isolating noise while preserving transient diagnostic features. Empirical mode decomposition adaptively breaks the signal into intrinsic components for selective noise removal. Adaptive filters such as Least Mean Squares algorithms adjust their parameters dynamically, and Kalman filters support real-time estimation. Most strikingly, neural networks themselves, including denoising autoencoders and generative adversarial networks, now learn noise patterns directly from data. Segmentation, normalization, and R-peak alignment complete the pipeline, and the review notes that even deep learning models, which supposedly bypass manual feature engineering, remain heavily dependent on high-quality preprocessing.

The most alarming findings concern the data itself. Ninety-five percent of the reviewed studies relied on publicly available ECG repositories, with only 3 percent using private institutional data and 2 percent collecting their own recordings. That reliance on open datasets aids reproducibility but concentrates risk: 78 percent of public ECG data originates from North America or Europe, ethnicity is reported in fewer than 10 percent of datasets, pediatric populations account for less than 3 percent of recordings, and average female representation stands at 41.5 percent. Annotation quality is equally uneven, with only 35 percent of datasets specifying annotator qualifications and just 22 percent reporting inter-annotator agreement. The consequences are measurable. Studies performing cross-dataset validation report average performance degradation of 8 to 15 percent when models are tested on data from different sources, and only 28 percent of studies include any form of external validation. A model trained primarily on middle-aged European patients may simply fail when deployed in a demographically different hospital, a phenomenon known as dataset shift that raises serious fairness and equity concerns.

Transparency poses another barrier. Deep networks function as black boxes, producing accurate predictions with little insight into their reasoning, and clinicians justifiably hesitate to act on alerts they cannot interrogate. Explainable AI methods aim to open that box: SHAP values, rooted in game theory, quantify each feature’s marginal contribution to a prediction, while LIME approximates a complex model locally with an interpretable surrogate. Visual techniques such as Class Activation Maps and saliency maps highlight which portions of the waveform drove a classification, letting observers see exactly which ECG segments influenced the model’s decision. Yet only 52 percent of recent studies incorporate any explainability method, demographic fairness assessments remain rare, and the review argues that current explanations often satisfy technical curiosity without answering the questions clinicians actually ask, such as whether a model’s reasoning aligns with established pathophysiology and whether its recommendation can be trusted for a specific patient.

The reproducibility picture is equally troubling. Only 29 percent of reviewed studies made their code publicly available, preprocessing details were frequently too sparse to permit replication, and the field lacks standardized benchmarks, making cross-study comparison nearly impossible. Few papers include comprehensive ablation analyses quantifying which preprocessing steps or architectural components actually drive performance. Evaluation practices compound the problem: most studies emphasize overall accuracy while neglecting class-specific performance, calibration, and uncertainty quantification, all of which matter enormously in clinical decision-making where a false negative on a lethal arrhythmia carries very different stakes than a false alarm.

The authors close with a roadmap that reads as both a critique and a call to action. In the near term, they urge standardized evaluation benchmarks, mandatory demographic reporting, and routine external validation; in the medium term, multi-center data consortia and explainability integrated with clinical workflows; and in the long term, prospective validation in diverse practice settings and continuous learning systems that adapt as populations change. They also point to federated learning, which trains models across institutions without centralizing sensitive patient data, as a promising route around privacy barriers. The convergence of abundant ECG data, powerful architectures, and pressing clinical need creates a genuine opportunity, the review concludes, but seizing it will require engineers to work alongside clinicians from project inception, researchers to prize generalizability over benchmark scores, and the whole community to treat transparency and fairness not as optional extras but as the foundation of AI-assisted cardiovascular care.

Subject of Research: Artificial intelligence methods for ECG-based cardiovascular disease prediction

Article Title: A systematic survey of artificial intelligence methods for ECG-based cardiovascular disease prediction

Article References: M, T., Panchal, S. M., & S, P. L. G. (2026). A systematic survey of artificial intelligence methods for ECG-based cardiovascular disease prediction. Discover Artificial Intelligence, 6(1), Article 1259. https://doi.org/10.1007/s44163-026-02170-7

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02170-7

Keywords: artificial intelligence, electrocardiogram, cardiovascular disease, deep learning, machine learning, explainable AI, systematic review, clinical validation, dataset bias, CNN, LSTM, signal preprocessing

The problem, Addaquay contends, is not simply that the models fail to imitate surface sounds. They often succeed at that. The failure lies deeper, in what he calls sonic data coloniality: the processes through which musical cultures are collected, classified, abstracted, trained upon, generated and monetized within technological frameworks that remain overwhelmingly Western in design and control. The term extends the idea of data colonialism, developed by scholars Nick Couldry and Ulises Mejias, into the sonic domain. Just as human experience has been harvested as raw data for capitalist extraction, African music is being scraped, compressed and recombined into a homogenized continental sound, one that serves global markets while the communities who created the traditions remain largely absent from the pipelines, the training data decisions and the profits.

The scale of the asymmetry is striking. Recent studies have documented that generative music models almost entirely omit the Global South. A 2024 preprint titled Missing Melodies found what its authors described as a nearly complete omission of non-Western musical traditions from major text-to-music systems, and follow-up work presented at NAACL in 2025 confirmed representational bias and poor cross-cultural adaptability in leading models. A bibliometric analysis of the first twenty-five years of the International Society for Music Information Retrieval, the field’s main research community, showed that its authorship remains heavily concentrated at a Western center. Meanwhile, the commercial footprint of generative music keeps expanding: Deezer reported in April 2026 that AI-generated tracks already represent 44 percent of all newly uploaded music on its platform, meaning the biases baked into these models are now shaping the global soundscape at industrial scale.

Why do the models fail even when they try? The technical answer begins with training data. Large-scale music generation systems such as Meta’s MusicGen, Google DeepMind’s Lyria, Stability AI’s Stable Audio, Suno and ElevenLabs are trained predominantly on Western popular and classical recordings, a corpus that reflects existing licensing structures and internet availability rather than global musical diversity. African recordings, where they exist at all, are often locked in archives such as the International Library of African Music at Rhodes University, whose vast collections of field recordings have only recently been digitized. What little African music does enter training pipelines is frequently labeled under the catch-all category of world music, a classification that flattens thousands of distinct traditions, from Ewe agbadza drumming to isicathamiya choral singing to Yoruba talking-drum language, into a single undifferentiated label that the model then averages into stereotype.

The deeper technical difficulty is that many African musical systems encode knowledge that Western-derived representations struggle to capture. In many African tone languages, the pitch contour of speech carries lexical meaning, and songs frequently map melodic lines directly onto tonal patterns, so that singing the wrong melody changes the words themselves. Research on Northern Ewe music by musicologist Kofi Agawu, studies of Akan tone encoding across musical modalities, and work on the Yoruba gángan talking drum, which reproduces speech tones to communicate proverbs and names, all demonstrate that pitch, language and meaning are inseparable in these traditions. A generative model trained on spectrograms and tokenized audio has no representation of linguistic tone, no notion that a melodic interval might be a grammatical error. Similarly, West African polyrhythm and the standard bell pattern are not merely rhythmic sequences but networks of interlocking parts, microtiming relationships and dance implications that only make sense within a performance system.

Addaquay’s framework proposes a shift in how AI-generated African music should be judged. Rather than asking whether an output sounds authentic or pure, a criterion he rejects as essentialist, he argues for cultural competence: whether the system can produce music that is comprehensible within the cultural, linguistic and social systems that give the tradition meaning. Evaluation, he suggests, should examine vocality, tone-language relationships, rhythm, groove, microtiming, instrumental idiom, call-and-response structures, dance implications, social function, and the questions of consent and benefit-sharing that surround the data itself. Call-and-response in Ewe agbadza songs, for instance, is not a decorative texture but one element in a network of musical and social factors; a model that renders it as a static loop has reproduced the sound without the practice.

To make such evaluation practical, the article outlines a prompt-based and source-conditioned testing methodology. Generic prompts, such as a request for African music, reveal the model’s default stereotypes. Culturally specific prompts, naming particular traditions, instruments or social contexts, test whether the model can differentiate at all. Expert prompts, written with musicological precision, probe the limits of controllability. Source-conditioned workflows, in which users condition generation on rights-cleared African recordings, test whether the system can learn from a specific tradition rather than a continental average. Listening tests can then be structured using established subjective assessment protocols such as the ITU-R BS.1534 MUSHRA standard and web-based frameworks like WebMUSHRA, with evaluators drawn from the relevant cultural communities rather than from generic crowdsourcing pools. Live evaluation arenas for text-to-music models, now emerging in the research literature, could be extended with culturally stratified judge panels.

The governance dimension is as important as the technical one. The African Union has moved to assert continental control over the data and AI agenda, adopting a data policy framework in 2022 and a continental artificial intelligence strategy in 2024, and signing a partnership with the Music In Africa Foundation in 2025. International instruments are also converging on the issue: UNESCO’s 2021 Recommendation on the Ethics of Artificial Intelligence and its 2025 expert report on AI and culture, alongside World Intellectual Property Organization guidance on documenting traditional knowledge and traditional cultural expressions, all point toward frameworks in which communities hold rights over their cultural data. Legal pressure is mounting elsewhere too, with musicians’ unions suing record labels over AI licensing and the Swedish collecting society STIM launching the world’s first AI license for music. Whether African musicians and institutions will hold equivalent leverage over the sonic data extracted from their traditions remains an open and urgent question.

Encouragingly, African-led initiatives are building the missing infrastructure. Large-scale multilingual speech corpora such as WAXAL, developed with Google Research, demonstrate that rights-cleared, community-aware African audio datasets are feasible at scale. Organizations including Masakhane, the African Languages Lab and Khaya AI are building African language technology from within the continent, and the CompMusic project has long argued for music information research organized around specific musical cultures rather than Western defaults. Scholars of decolonial AI, including work on epistemic sovereignty in African AI governance, argue that the goal is not to bolt African content onto Western systems but to reshape the taxonomies, evaluation criteria and ownership structures themselves. Addaquay’s own prior research on sub-Saharan singing styles, traditional dance accompaniment and music preservation in Ghana supplies the musicological grounding for that reshape.

The stakes extend far beyond academic debate. If AI-generated music continues its explosive growth, the sonic image of an entire continent will increasingly be produced by models that hear Africa through Western ears, trained on data extracted without consent, evaluated by metrics that reward surface resemblance and monetized by companies far from the source communities. Addaquay’s argument is that African music should enter AI systems not as a homogenized continental sound but through its distinct traditions, hybrid practices, performance expertise and cultural authorities. Cultural competence, consent and benefit-sharing, he suggests, are not constraints on innovation but the conditions under which musical intelligence can claim to be intelligent at all. The alternative is a future in which anyone can prompt for Africa, but almost no one, including the machines, actually hears it.

Subject of Research: Cultural bias and sonic data coloniality in generative AI music systems for African musical traditions

Article Title: Prompting Africa, hearing the West: generative AI music, sonic data coloniality and the future of culturally accountable musical intelligence

Article References: Addaquay, A. P. (2026). Prompting Africa, hearing the West: generative AI music, sonic data coloniality and the future of culturally accountable musical intelligence. AI & SOCIETY. https://doi.org/10.1007/s00146-026-03345-7

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03345-7

Keywords: generative AI music, African music, sonic data coloniality, music information retrieval, cultural competence, AI ethics, dataset bias, text-to-music models, decolonial AI, tone language, call-and-response, benefit-sharing