The study, “A Mini-review on CNN-based Biomedical Image Processing: Modality, Applications, and Future Trends,” was carried out by Jie Li and Li-Xin Wang in Current Healthcare Research.
A new review from researchers at Taylor’s University in Malaysia and Ningxia Medical University in China maps how a powerful class of artificial neural networks is being applied across the full breadth of medical imaging, identifies where the technology is delivering real clinical value, and makes the case that closing the gap between promising research results and routine hospital use will require urgent progress on three fronts: making these systems more transparent, more efficient, and more trustworthy.
If you want to publish your article please visit : https://bit.ly/4de0DRi
From X-rays to Brain Scans: How CNNs Read Medical Images
Convolutional Neural Networks (CNNs) are a type of deep learning model originally designed to analyse visual data by automatically identifying patterns at progressively complex levels, from simple edges and textures up to recognisable structures such as tumours, lesions, and organ boundaries. In medical imaging, this capability makes them particularly well-suited to tasks that have traditionally depended on expert human interpretation, which is time-consuming, subject to variability between clinicians, and difficult to scale. This review systematically examines how CNNs are being applied not just to conventional two-dimensional images such as X-rays and histopathology slides, but also to three-dimensional volumetric data from MRI and CT scans, spectral signals such as audio features used to detect voice disorders, time-series data such as EEG recordings for seizure detection, and multimodal combinations of all of the above. The breadth of data types that CNNs can handle, and the consistent pattern of performance improvement they have delivered across all of them, has made them the dominant computational tool in medical image processing over the past decade. The review covers three core tasks that together define most of what medical image analysis involves: disease detection and classification, image segmentation, and image registration.
Cancer, Cardiovascular Disease, and the Eye: Where CNNs Are Already Working
The range of clinical applications the review documents is substantial. In cancer detection, CNN-based models have demonstrated strong performance across breast, lung, colon, brain, and cervical cancer, using both tissue slides and radiological scans. In cardiovascular medicine, CNNs have been applied to detect arterial blockages and identify vascular abnormalities using gradient-based optimisation approaches. In neurology, multiple CNN architectures have shown strong results in detecting and classifying Alzheimer’s disease from MRI data, as well as identifying acute ischaemic stroke. In ophthalmology, CNN models have been built to detect glaucoma using fundus images and optical coherence tomography, and to classify diabetic retinopathy in at-risk patients. For infectious disease, CNNs have been applied to pneumonia detection using chest X-rays, malaria detection from blood cell images, and COVID-19 screening. Alongside disease detection, the review examines three categories of image segmentation, the process of partitioning an image into clinically meaningful regions. Volumetric segmentation, which operates on three-dimensional scan data, enables precise delineation of tumours in MRI. Semantic segmentation assigns a label to every pixel in an image, identifying lesions in X-rays by class. Multi-class segmentation differentiates multiple structures simultaneously, such as distinguishing different tissue types in CT brain scans. On image registration, which involves aligning images taken from different angles, time points, or imaging modalities so they can be meaningfully compared, the review covers rigid, deformable, affine, and multi-modal approaches, noting that combining CNN models with spatial transformer networks has shown particular promise for improving alignment accuracy across different imaging systems.
The Gap That Still Needs Closing
Despite the scale of progress documented in this review, the authors are frank about the obstacles that continue to limit how widely and reliably these tools are used in actual clinical practice. The most persistent is the heterogeneity of medical data: images vary in resolution, format, and noise levels across hospitals, scanners, and patient populations, and models trained in one setting frequently perform less well when deployed in another. The second major obstacle is interpretability. CNNs make their decisions through complex internal calculations that cannot easily be traced back to a human-readable explanation, and clinicians are understandably reluctant to act on a recommendation from a system whose reasoning they cannot follow. Tools such as heat maps and saliency maps, which highlight the regions of an image that most influenced a CNN’s output, are a step toward addressing this, but are not yet standard practice. The review identifies three directions for future development: the integration of CNN-based tools into Clinical Decision Support Systems in ways that fit naturally into existing clinical workflows and electronic health record systems; the adoption of cloud-based infrastructure to make computationally demanding models accessible in resource-limited settings; and engagement with the regulatory and ethical landscape, including the FDA’s framework for AI-based medical software, which requires evidence of good machine learning practices, real-world performance monitoring, and data management standards. The research was led by corresponding author Dr. Jie Li of the Digital Health and Medical Advancement Impact Lab, School of Computer Science, Taylor’s University, 47500, Selangor, Malaysia, alongside co-author Li-Xin Wang of the Medical Laboratory Center of Ningxia Medical University General Hospital, Ningxia, China.
Read the published article here: https://bit.ly/4yMr4Hp
Article title: A Mini-review on CNN-based Biomedical Image Processing: Modality, Applications, and Future Trends
DOI: 10.2174/0130507731432304260421212210
Current Healthcare Research
10.2174/0130507731432304260421212210
Literature review
A Mini-review on CNN-based Biomedical Image Processing: Modality, Applications, and Future Trends