Biomedicine is becoming “AI-dependent” as algorithms outperform human diagnostics

AI is not limited to diagnostics or imaging. It also plays a transformative role in biomedical research, computational modeling, genomics and drug development. The authors highlight how machine-learning models trained on simulated data can capture complex biophysical behaviors that are difficult to study experimentally. For example, AI classifiers analyzing red blood cell elasticity provide new avenues for studying diseases such as sickle cell anemia, hereditary elliptocytosis and other hemolytic disorders.

Artificial intelligence (AI) technologies are moving deeper into frontline healthcare, reshaping diagnostics, drug development, clinical decision-making and digital epidemiology in ways that are beginning to redefine modern medical practice. A new expert editorial warns that although AI’s impact on biomedicine is expanding at unprecedented speed, the transformation demands equally rapid progress in clinical validation, ethical governance and responsible deployment.

The peer-reviewed editorial, “The Use of Artificial Intelligence (AI) Technologies in Biomedicine,” published in Applied Sciences, reviews recent advances in AI applied to medical imaging, infectious disease monitoring, hematology, genetics and public health. The authors analyze a series of new studies demonstrating how multimodal deep learning, radiomics, natural language processing, spatial analysis and machine-learning classifiers are already improving diagnostic accuracy and operational efficiency in multiple healthcare settings.

AI expands its footprint in diagnostics, imaging and predictive medicine

The authors explore how deep learning, neural networks and multimodal computational tools are now central to tasks once entirely dependent on human expertise. Medical imaging, pathology, genomic interpretation and clinical data analysis are among the fields already reshaped by automated pattern recognition and algorithmic prediction.

The review highlights a featured study that uses a U-Net–based deep learning model to perform automated bone marrow segmentation on PET/CT scans of patients with non-Hodgkin lymphoma. The model integrates semantic segmentation with high-resolution radiomic features to classify disease involvement with remarkable precision. According to the editorial, this approach demonstrates how AI can accelerate and standardize image analysis in oncology, reducing inter-observer variability and enabling rapid clinical decision support. The model achieved strong segmentation performance and showed how radiomic signatures extracted from PET/CT could inform prognosis and therapeutic planning.

Another study summarized in the editorial explores the use of machine-learning models, including random forests, support vector machines and gradient boosting algorithms, to classify red blood cell elasticity using simulation-based datasets. The accuracy achieved by these models suggests potential diagnostic applications in hematological disorders, where cell deformability is an important biomarker. The authors note that AI-driven biophysical classification could complement laboratory diagnostics in conditions involving structural or functional abnormalities of red blood cells.

These examples point to a broader trend: AI is enhancing the resolution, reliability and interpretive power of diagnostic imaging and biomedical signal processing. As hospitals digitize clinical workflows and accumulate vast imaging datasets, AI-enabled automation is expected to reduce diagnostic delays, expand access to specialized interpretation and support earlier intervention.

The editorial stresses that across all imaging fields, oncology, radiology, pathology and emergency medicine, AI systems are now being trained to detect abnormalities, segment tissues, quantify lesions and predict outcomes from complex, multimodal clinical data. This trend signals a structural shift toward predictive, personalized and data-driven care.

AI brings new advances in tuberculosis diagnosis and real-time disease surveillance

The authors uncover an AI-driven model for tuberculosis detection that integrates structured clinical data with natural language processing applied to electronic medical records. By combining symptom descriptions, diagnostic histories, laboratory values and clinician notes, the model outperformed conventional smear microscopy, a method still widely used in many low-income regions.

The editorial notes that tuberculosis remains one of the world’s leading infectious killers, and diagnostic delays are a major barrier to treatment. The AI-based system demonstrated higher sensitivity by identifying subtle patterns and correlations across multiple data types. This finding is particularly important for settings where access to advanced imaging, molecular diagnostics or expert clinicians is limited. AI-assisted triage tools may therefore help reduce missed diagnoses and support earlier treatment initiation.

Another study examined in the editorial introduces an innovative approach to epidemic surveillance in Lombardy, Italy. Researchers fused emergency medical dispatch data with geographic information systems and random forest classifiers to build a spatial prediction model capable of identifying municipalities at different epidemic stages. This system processed real-time call volumes, symptom clusters and spatial transmission patterns to classify areas as emerging, peaking or declining in infection activity.

The editorial underscores the significance of this approach: disease surveillance often lags behind real-world transmission dynamics, but AI can transform raw emergency response data into actionable early-warning systems. Such tools could help health authorities deploy resources more efficiently and respond faster to developing outbreaks, especially during viral epidemics like influenza, COVID-19 or respiratory syncytial virus.

Together, these studies reflect AI’s growing role in infectious disease prediction. By integrating clinical text, telemedicine data, geospatial information and emergency response records, AI is shifting public health from reactive intervention to proactive early detection.

AI advances in simulation, bioinformatics and multimodal analysis signal new era of biomedical research

AI is not limited to diagnostics or imaging. It also plays a transformative role in biomedical research, computational modeling, genomics and drug development. The authors highlight how machine-learning models trained on simulated data can capture complex biophysical behaviors that are difficult to study experimentally. For example, AI classifiers analyzing red blood cell elasticity provide new avenues for studying diseases such as sickle cell anemia, hereditary elliptocytosis and other hemolytic disorders.

The editorial also discusses how AI technologies help integrate heterogeneous biomedical information, genomic sequences, medical images, laboratory values, electronic health records and environmental data, into unified predictive models. This multimodal integration is essential for precision medicine, allowing researchers and clinicians to understand patient conditions from multiple perspectives.

In fields like genetics, immunology and cellular biology, machine-learning algorithms are increasingly used to identify disease-associated variants, analyze gene expression patterns, study protein interactions and optimize therapeutic strategies. AI-based models can process complex biological data faster and more thoroughly than human analysts, uncovering patterns that support new hypotheses and accelerate discovery cycles.

The editorial points out that in mental health and behavioral medicine, AI tools are being deployed to categorize symptoms, predict treatment outcomes, and analyze speech or facial cues that correlate with psychological disorders. Although these technologies remain early in development, they suggest future possibilities for more personalized and continuous behavioral health monitoring.

Across biomedical science, the authors emphasize that AI is enabling new research frontiers. From drug screening to molecular pathway modeling, machine-learning systems are opening doors to analyses previously too complex or time-consuming to perform manually.

But the researchers also warn that these advances require rigorous oversight to ensure that algorithms remain transparent, reproducible and clinically interpretable. Without strong methodological standards, the authors argue, AI could introduce bias, magnify data gaps or produce outputs that clinicians cannot easily validate.

AI’s growing power requires ethical governance, clinical validation and responsible use

While the editorial highlights the promise of AI across biomedicine, it also delivers a strong caution: the integration of AI into clinical care must be carefully governed. The authors stress that every breakthrough brings new ethical considerations around transparency, privacy, accountability and regulatory oversight.

Key concerns raised in the editorial include:

• Algorithmic transparency: Many deep learning models operate as “black boxes,” making it difficult to understand how they reach decisions. Clinicians require interpretable outputs to maintain diagnostic responsibility.

• Data quality and bias: AI models can inherit biases from the datasets used to train them. Incomplete or skewed data may lead to unequal performance across populations.

• Privacy and security: Expanding use of AI in biomedicine increases exposure to sensitive patient information. Strict safeguards are essential.

• Clinical validation: AI models must undergo rigorous testing across diverse settings to ensure reliability before deployment.

• Ethical accountability: AI does not replace clinical judgment. Systems must be monitored to ensure they support, rather than override, medical professionals.

• Regulatory frameworks: The rapid pace of AI innovation demands updated regulatory standards to ensure safe integration into healthcare systems.

To sum up, AI tools should enhance, not replace, human expertise. The authors argue that responsible AI deployment requires collaboration between clinicians, researchers, ethicists, regulators and technologists to ensure that machine-learning models serve the best interests of patients.