Search

Purpose: Artificial intelligence (AI)-tools hold great potential to compensate for missing resources in health-care systems but often fail to be implemented in clinical routine. Intriguingly, no-code and low-code technologies allow clinicians to develop Artificial intelligence (AI)-tools without requiring in-depth programming knowledge. Clinician-driven projects allow to adequately identify and address real clinical needs and, therefore, hold superior potential for clinical implementation. In this light, this study aimed for the clinician-driven development of a tool capable of measuring corneal lesions relative to total corneal surface area and eliminating inaccuracies in two-dimensional measurements by three-dimensional fitting of the corneal surface. Methods: Standard slit-lamp photographs using a blue-light filter after fluorescein instillation taken during clinical routine were used to train a fully convolutional network to automatically detect the corneal white-to-white distance, the total fluorescent area and the total erosive area. Based on these values, the algorithm calculates the affected area relative to total corneal surface area and fits the area on a three-dimensional representation of the corneal surface. Results: The developed algorithm reached dice scores >0.9 for an automated measurement of the relative lesion size. Furthermore, only 25% of conventional manual measurements were within a ± 10% range of the ground truth. Conclusions: The developed algorithm is capable of reliably providing exact values for corneal lesion sizes. Additionally, three-dimensional modeling of the corneal surface is essential for an accurate measurement of lesion sizes. Besides telemedicine applications, this approach harbors great potential for clinical trials where exact quantitative and observer-independent measurements are essential. [ABSTRACT FROM AUTHOR]

Microscopy of stained blood smears is still a ubiquitous technique in pathology. It is often used in addition to automated electronic counters or flow cytometers to evaluate leukocytes and their morphologies in a rather simple manner and has low requirements for resources and equipment. However, despite the constant advances in microscopy, computer science, and pathology, it still usually follows the traditional approach of manual assessment by humans. We aimed to extend this technique using AI-based automated cell recognition methods while maintaining its technical simplicity. Using the web platform IKOSA, we developed an AI-based workflow to segment and identify all blood cells in DAPI-Giemsa co-stained blood smears. Thereby, we could automatically detect and classify neutrophils (young and segmented), lymphocytes, eosinophils, and monocytes, in addition to erythrocytes and platelets, in contrast to previously published algorithms, which usually focus on only one type of blood cell. Furthermore, our method delivers quantitative measurements, unattainable by the classical method or formerly published AI techniques, and it provides more sophisticated analyses based on entropy or gray-level co-occurrence matrices (GLCMs), which have the potential to monitor changes in internal cellular structures associated with disease states or responses to treatment. We conclude that AI-based automated blood cell evaluation has the potential to facilitate and improve routine diagnostics by adding quantitative shape and structure parameters to simple leukocyte counts of classical analysis. [ABSTRACT FROM AUTHOR]

Grading intraepithelial neoplasia is crucial to derive an accurate estimate of pre-cancerous stages and is currently performed by pathologists assessing histopathological images. Inter- and intra-observer variability can significantly be reduced, when reliable, quantitative image analysis is introduced into diagnostic processes. On a challenging dataset, we evaluated the potential of learning a classifier to grade anal intraepitelial neoplasia. Support vector machines were trained on images represented by fractal and statistical features. We show that pursuing a learning-based grading strategy yields highly reliable results. Compared to existing methods, the proposed method outperformed them by a significant margin.

Fractal analysis (FA) should be able to yield reliable and fast results for high-resolution digital images to be applicable in fields that require immediate outcomes. Triggered by an efficient implementation of FA for binary images, we present three new approaches for fractal dimension (D) estimation of images that utilize image pyramids, namely, the pyramid triangular prism, the pyramid gradient, and the pyramid differences method (PTPM, PGM, PDM). We evaluated the performance of the three new and five standard techniques when applied to images with sizes up to 8192 × 8192 pixels. By using artificial fractal images created by three different generator models as ground truth, we determined the scale ranges with minimum deviations between estimation and theory. All pyramidal methods (PM) resulted in reasonable D values for images of all generator models. Especially, for images with sizes ≥1024×1024 pixels, the PMs are superior to the investigated standard approaches in terms of accuracy and computation time. A measure for the possibility to differentiate images with different intrinsic D values did show not only that the PMs are well suited for all investigated image sizes, and preferable to standard methods especially for larger images, but also that results of standard D estimation techniques are strongly influenced by the image size. Fastest results were obtained with the PDM and PGM, followed by the PTPM. In terms of absolute D values best performing standard methods were magnitudes slower than the PMs. Concluding, the new PMs yield high quality results in short computation times and are therefore eligible methods for fast FA of high-resolution images.