Cancer Res. 2024 Apr 3. doi: 10.1158/0008-5472.CAN-23-2040. Online ahead of print.

ABSTRACT

Standard-of-care treatment regimens have long been designed for maximal cell killing, yet these strategies often fail when applied to metastatic cancers due to the emergence of drug resistance. Adaptive treatment strategies have been developed as an alternative approach, dynamically adjusting treatment to suppress the growth of treatment-resistant populations and thereby delay, or even prevent, tumor progression. Promising clinical results in prostate cancer indicate the potential to optimize adaptive treatment protocols. Here, we applied deep reinforcement learning (DRL) to guide adaptive drug scheduling and demonstrated that these treatment schedules can outperform the current adaptive protocols in a mathematical model calibrated to prostate cancer dynamics, more than doubling the time to progression. The DRL strategies were robust to patient variability, including both tumor dynamics and clinical monitoring schedules. The DRL framework could produce interpretable, adaptive strategies based on a single tumor burden threshold, replicating and informing optimal treatment strategies. The DRL framework had no knowledge of the underlying mathematical tumor model, demonstrating the capability of DRL to help develop treatment strategies in novel or complex settings. Finally, a proposed five-step pathway, which combined mechanistic modeling with the DRL framework and integrated conventional tools to improve interpretability compared to traditional "black-box" DRL models, could allow translation of this approach to the clinic. Overall, the proposed framework generated personalized treatment schedules that consistently outperformed clinical standard-of-care protocols.

PMID:38569183 | DOI:10.1158/0008-5472.CAN-23-2040

Med Phys. 2024 Apr 3. doi: 10.1002/mp.17047. Online ahead of print.

ABSTRACT

BACKGROUND: Patient head motion is a common source of image artifacts in computed tomography (CT) of the head, leading to degraded image quality and potentially incorrect diagnoses. The partial angle reconstruction (PAR) means dividing the CT projection into several consecutive angular segments and reconstructing each segment individually. Although motion estimation and compensation using PAR has been developed and investigated in cardiac CT scans, its potential for reducing motion artifacts in head CT scans remains unexplored.

PURPOSE: To develop a deep learning (DL) model capable of directly estimating head motion from PAR images of head CT scans and to integrate the estimated motion into an iterative reconstruction process to compensate for the motion.

METHODS: Head motion is considered as a rigid transformation described by six time-variant variables, including the three variables for translation and three variables for rotation. Each motion variable is modeled using a B-spline defined by five control points (CP) along time. We split the full projections from 360° into 25 consecutive PARs and subsequently input them into a convolutional neural network (CNN) that outputs the estimated CPs for each motion variable. The estimated CPs are used to calculate the object motion in each projection, which are incorporated into the forward and backprojection of an iterative reconstruction algorithm to reconstruct the motion-compensated image. The performance of our DL model is evaluated through both simulation and phantom studies.

RESULTS: The DL model achieved high accuracy in estimating head motion, as demonstrated in both the simulation study (mean absolute error (MAE) ranging from 0.28 to 0.45 mm or degree across different motion variables) and the phantom study (MAE ranging from 0.40 to 0.48 mm or degree). The resulting motion-corrected image, I D L , P A R ${I}_{DL,\ PAR}$ , exhibited a significant reduction in motion artifacts when compared to the traditional filtered back-projection reconstructions, which is evidenced both in the simulation study (image MAE drops from 178 ± $ \pm $ 33HU to 37 ± $ \pm $ 9HU, structural similarity index (SSIM) increases from 0.60 ± $ \pm $ 0.06 to 0.98 ± $ \pm $ 0.01) and the phantom study (image MAE drops from 117 ± $ \pm $ 17HU to 42 ± $ \pm $ 19HU, SSIM increases from 0.83 ± $ \pm $ 0.04 to 0.98 ± $ \pm $ 0.02).

CONCLUSIONS: We demonstrate that using PAR and our proposed deep learning model enables accurate estimation of patient head motion and effectively reduces motion artifacts in the resulting head CT images.

PMID:38569143 | DOI:10.1002/mp.17047

Med Phys. 2024 Apr 3. doi: 10.1002/mp.17057. Online ahead of print.

ABSTRACT

BACKGROUND: Proton therapy is a form of radiotherapy commonly used to treat various cancers. Due to its high conformality, minor variations in patient anatomy can lead to significant alterations in dose distribution, making adaptation crucial. While cone-beam computed tomography (CBCT) is a well-established technique for adaptive radiation therapy (ART), it cannot be directly used for adaptive proton therapy (APT) treatments because the stopping power ratio (SPR) cannot be estimated from CBCT images.

PURPOSE: To address this limitation, Deep Learning methods have been suggested for converting pseudo-CT (pCT) images from CBCT images. In spite of convolutional neural networks (CNNs) have shown consistent improvement in pCT literature, there is still a need for further enhancements to make them suitable for clinical applications.

METHODS: The authors introduce the 3D vision transformer (ViT) block, studying its performance at various stages of the proposed architectures. Additionally, they conduct a retrospective analysis of a dataset that includes 259 image pairs from 59 patients who underwent treatment for head and neck cancer. The dataset is partitioned into 80% for training, 10% for validation, and 10% for testing purposes.

RESULTS: The SPR maps obtained from the pCT using the proposed method present an absolute relative error of less than 5% from those computed from the planning CT, thus improving the results of CBCT.

CONCLUSIONS: We introduce an enhanced ViT3D architecture for pCT image generation from CBCT images, reducing SPR error within clinical margins for APT workflows. The new method minimizes bias compared to CT-based SPR estimation and dose calculation, signaling a promising direction for future research in this field. However, further research is needed to assess the robustness and generalizability across different medical imaging applications.

PMID:38569141 | DOI:10.1002/mp.17057

Med Phys. 2024 Apr 3. doi: 10.1002/mp.17044. Online ahead of print.

ABSTRACT

BACKGROUND: With the continuous development of deep learning algorithms in the field of medical images, models for medical image processing based on convolutional neural networks have made great progress. Since medical images of rectal tumors are characterized by specific morphological features and complex edges that differ from natural images, achieving good segmentation results often requires a higher level of enrichment through the utilization of semantic features.

PURPOSE: The efficiency of feature extraction and utilization has been improved to some extent through enhanced hardware arithmetic and deeper networks in most models. However, problems still exist with detail loss and difficulty in feature extraction, arising from the extraction of high-level semantic features in deep networks.

METHODS: In this work, a novel medical image segmentation model has been proposed for Magnetic Resonance Imaging (MRI) image segmentation of rectal tumors. The model constructs a backbone architecture based on the idea of jump-connected feature fusion and solves the problems of detail feature loss and low segmentation accuracy using three novel modules: Multi-scale Feature Retention (MFR), Multi-branch Cross-channel Attention (MCA), and Coordinate Attention (CA).

RESULTS: Compared with existing methods, our proposed model is able to segment the tumor region more effectively, achieving 97.4% and 94.9% in Dice and mIoU metrics, respectively, exhibiting excellent segmentation performance and computational speed.

CONCLUSIONS: Our proposed model has improved the accuracy of both lesion region and tumor edge segmentation. In particular, the determination of the lesion region can help doctors identify the tumor location in clinical diagnosis, and the accurate segmentation of the tumor edge can assist doctors in judging the necessity and feasibility of surgery.

PMID:38569054 | DOI:10.1002/mp.17044

PLoS One. 2024 Apr 3;19(4):e0295474. doi: 10.1371/journal.pone.0295474. eCollection 2024.

ABSTRACT

Insect monitoring is essential to design effective conservation strategies, which are indispensable to mitigate worldwide declines and biodiversity loss. For this purpose, traditional monitoring methods are widely established and can provide data with a high taxonomic resolution. However, processing of captured insect samples is often time-consuming and expensive, which limits the number of potential replicates. Automated monitoring methods can facilitate data collection at a higher spatiotemporal resolution with a comparatively lower effort and cost. Here, we present the Insect Detect DIY (do-it-yourself) camera trap for non-invasive automated monitoring of flower-visiting insects, which is based on low-cost off-the-shelf hardware components combined with open-source software. Custom trained deep learning models detect and track insects landing on an artificial flower platform in real time on-device and subsequently classify the cropped detections on a local computer. Field deployment of the solar-powered camera trap confirmed its resistance to high temperatures and humidity, which enables autonomous deployment during a whole season. On-device detection and tracking can estimate insect activity/abundance after metadata post-processing. Our insect classification model achieved a high top-1 accuracy on the test dataset and generalized well on a real-world dataset with captured insect images. The camera trap design and open-source software are highly customizable and can be adapted to different use cases. With custom trained detection and classification models, as well as accessible software programming, many possible applications surpassing our proposed deployment method can be realized.

PMID:38568922 | DOI:10.1371/journal.pone.0295474

PLoS One. 2024 Apr 3;19(4):e0288121. doi: 10.1371/journal.pone.0288121. eCollection 2024.

ABSTRACT

Deep learning shows promise for automating detection and classification of wildlife from digital aerial imagery to support cost-efficient remote sensing solutions for wildlife population monitoring. To support in-flight orthorectification and machine learning processing to detect and classify wildlife from imagery in near real-time, we evaluated deep learning methods that address hardware limitations and the need for processing efficiencies to support the envisioned in-flight workflow. We developed an annotated dataset for a suite of marine birds from high-resolution digital aerial imagery collected over open water environments to train the models. The proposed 3-stage workflow for automated, in-flight data processing includes: 1) image filtering based on the probability of any bird occurrence, 2) bird instance detection, and 3) bird instance classification. For image filtering, we compared the performance of a binary classifier with Mask Region-based Convolutional Neural Network (Mask R-CNN) as a means of sub-setting large volumes of imagery based on the probability of at least one bird occurrence in an image. On both the validation and test datasets, the binary classifier achieved higher performance than Mask R-CNN for predicting bird occurrence at the image-level. We recommend the binary classifier over Mask R-CNN for workflow first-stage filtering. For bird instance detection, we leveraged Mask R-CNN as our detection framework and proposed an iterative refinement method to bootstrap our predicted detections from loose ground-truth annotations. We also discuss future work to address the taxonomic classification phase of the envisioned workflow.

PMID:38568890 | DOI:10.1371/journal.pone.0288121

IEEE J Biomed Health Inform. 2024 Apr 3;PP. doi: 10.1109/JBHI.2024.3384251. Online ahead of print.

ABSTRACT

As the most common complication of diabetes, diabetic retinopathy (DR) is one of the main causes of irreversible blindness. Automatic DR grading plays a crucial role in early diagnosis and intervention, reducing the risk of vision loss in people with diabetes. In these years, various deep-learning approaches for DR grading have been proposed. Most previous DR grading models are trained using the dataset of single-field fundus images, but the entire retina cannot be fully visualized in a single field of view. There are also problems of scattered location and great differences in the appearance of lesions in fundus images. To address the limitations caused by incomplete fundus features, and the difficulty in obtaining lesion information. This work introduces a novel multi-view DR grading framework, which solves the problem of incomplete fundus features by jointly learning fundus images from multiple fields of view. Furthermore, the proposed model combines multi-view inputs such as fundus images and lesion snapshots. It utilizes heterogeneous convolution blocks (HCB) and scalable self-attention classes (SSAC), which enhance the ability of the model to obtain lesion information. The experimental results show that our proposed method performs better than the benchmark methods on the large-scale dataset.

PMID:38568769 | DOI:10.1109/JBHI.2024.3384251

IEEE Trans Neural Netw Learn Syst. 2024 Apr 3;PP. doi: 10.1109/TNNLS.2024.3379735. Online ahead of print.

ABSTRACT

Recent progress in research on deep graph networks (DGNs) has led to a maturation of the domain of learning on graphs. Despite the growth of this research field, there are still important challenges that are yet unsolved. Specifically, there is an urge of making DGNs suitable for predictive tasks on real-world systems of interconnected entities, which evolve over time. With the aim of fostering research in the domain of dynamic graphs, first, we survey recent advantages in learning both temporal and spatial information, providing a comprehensive overview of the current state-of-the-art in the domain of representation learning for dynamic graphs. Second, we conduct a fair performance comparison among the most popular proposed approaches on node-and edge-level tasks, leveraging rigorous model selection and assessment for all the methods, thus establishing a sound baseline for evaluating new architectures and approaches.

PMID:38568760 | DOI:10.1109/TNNLS.2024.3379735

Appl Opt. 2024 Mar 10;63(8):1982-1990. doi: 10.1364/AO.515383.

ABSTRACT

Currently, the method of establishing the correspondence between the flame light field image and the temperature field by deep learning is widely used. Based on convolutional neural networks (CNNs), the reconstruction accuracy has been improved by increasing the depth of the network. However, as the depth of the network increases, it will lead to gradient explosion and network degradation. To further improve the reconstruction accuracy of the flame temperature field, this paper proposes an ES-ResNet18 model, in which SoftPool is used instead of MaxPool to preserve feature information more completely and efficient channel attention (ECA) is introduced in the residual block to reassign more weights to feature maps of critical channels. The reconstruction results of our method were compared with the CNN model and the original ResNet18 network. The results show that the average relative error and the maximum relative error of the temperature field reconstructed by the ES-ResNet18 model are 0.0203% and 0.1805%, respectively, which are reduced by one order of magnitude compared to the CNN model. Compared to the original ResNet18 network, they have decreased by 17.1% and 43.1%, respectively. Adding Gaussian noise to the flame light field images, when the standard deviation exceeds 0.03, the increase in reconstruction error of the ES-ResNet18 model is lower than that of ResNet18, demonstrating stronger anti-noise performance.

PMID:38568638 | DOI:10.1364/AO.515383

Appl Opt. 2024 Mar 10;63(8):1895-1907. doi: 10.1364/AO.505326.

ABSTRACT

For the deep-learning-based stereo-digital image correlation technique, the initial speckle position is crucial as it influences the accuracy of the generated dataset and deformation fields. To ensure measurement accuracy, an optimized extrinsic parameter estimation algorithm is proposed in this study to determine the rotation and translation matrix of the plane in which the speckle is located between the world coordinate system and the left camera coordinate system. First, the accuracy of different extrinsic parameter estimation algorithms was studied by simulations. Subsequently, the dataset of stereo speckle images was generated using the optimized extrinsic parameters. Finally, the improved dual-branch CNN deconvolution architecture was proposed to output displacements and strains simultaneously. Simulation results indicate that DAS-Net exhibits enhanced expressive capabilities, as evidenced by a reduction in displacement errors compared to previous research. The experimental results reveal that the mean absolute percentage error between the stereo-DIC results and the generated dataset is less than 2%, suggesting that the initial speckle positioning technology effectively minimizes the discrepancy between the images in the dataset and those obtained experimentally. Furthermore, the DAS-Net algorithm accurately measures the displacement and strain fields as well as their morphological characteristics.

PMID:38568627 | DOI:10.1364/AO.505326

Transl Vis Sci Technol. 2024 Apr 2;13(4):6. doi: 10.1167/tvst.13.4.6.

ABSTRACT

PURPOSE: To develop and validate a deep learning system (DLS) for estimation of vertical cup-to-disc ratio (vCDR) in ultra-widefield (UWF) and smartphone-based fundus images.

METHODS: A DLS consisting of two sequential convolutional neural networks (CNNs) to delineate optic disc (OD) and optic cup (OC) boundaries was developed using 800 standard fundus images from the public REFUGE data set. The CNNs were tested on 400 test images from the REFUGE data set and 296 UWF and 300 smartphone-based images from a teleophthalmology clinic. vCDRs derived from the delineated OD/OC boundaries were compared with optometrists' annotations using mean absolute error (MAE). Subgroup analysis was conducted to study the impact of peripapillary atrophy (PPA), and correlation study was performed to investigate potential correlations between sectoral CDR (sCDR) and retinal nerve fiber layer (RNFL) thickness.

RESULTS: The system achieved MAEs of 0.040 (95% CI, 0.037-0.043) in the REFUGE test images, 0.068 (95% CI, 0.061-0.075) in the UWF images, and 0.084 (95% CI, 0.075-0.092) in the smartphone-based images. There was no statistical significance in differences between PPA and non-PPA images. Weak correlation (r = -0.4046, P < 0.05) between sCDR and RNFL thickness was found only in the superior sector.

CONCLUSIONS: We developed a deep learning system that estimates vCDR from standard, UWF, and smartphone-based images. We also described anatomic peripapillary adversarial lesion and its potential impact on OD/OC delineation.

TRANSLATIONAL RELEVANCE: Artificial intelligence can estimate vCDR from different types of fundus images and may be used as a general and interpretable screening tool to improve community reach for diagnosis and management of glaucoma.

PMID:38568608 | DOI:10.1167/tvst.13.4.6

Appl Opt. 2024 Mar 20;63(9):2306-2313. doi: 10.1364/AO.506951.

ABSTRACT

In recent years, significant advancements have been made in the field of computational imaging, particularly due to the application of deep learning methods to imaging problems. However, only a few studies related to deep learning have examined the impact of diffraction distance on image restoration. In this paper, the effect of diffraction distance on image restoration is investigated based on the PhysenNet neural network. A theoretical framework for diffraction images at various diffraction distances is provided along with the applicable propagators. In the experiment, the PhysenNet network is selected to train on diffraction images with different distances and the impact of using different propagators on network performance is studied. Optimal propagators required to recover images at different diffraction distances are determined. Insights obtained through these experiments can expand the scope of neural networks in computational imaging.

PMID:38568586 | DOI:10.1364/AO.506951

Appl Opt. 2024 Mar 20;63(9):2248-2255. doi: 10.1364/AO.514144.

ABSTRACT

Three-dimensional (3D) optical authentication is important for modern information security. Existing 3D optical authentication methods rely on integral imaging devices, necessitating meticulous calibration and incurring high transmission overhead. To streamline the acquisition of 3D information, this paper introduces a novel 3D optical authentication approach, to the best of our knowledge, based on the construction of 3D data from multi-view images. The proposed method simplifies 3D projection by generating fixed-viewpoint elemental images, eliminating the need for additional viewpoint information during transmission and authentication. Compressed sensing is used for compression during transmission, and a deep learning network is designed for 3D reconstruction, enhancing the recovery. Experimental outcomes confirm the efficiency of our proposed approach for 3D authentication across diverse datasets.

PMID:38568579 | DOI:10.1364/AO.514144

Interdiscip Sci. 2024 Apr 3. doi: 10.1007/s12539-024-00618-x. Online ahead of print.

ABSTRACT

With the rapid development of NGS technology, the number of protein sequences has increased exponentially. Computational methods have been introduced in protein functional studies because the analysis of large numbers of proteins through biological experiments is costly and time-consuming. In recent years, new approaches based on deep learning have been proposed to overcome the limitations of conventional methods. Although deep learning-based methods effectively utilize features of protein function, they are limited to sequences of fixed-length and consider information from adjacent amino acids. Therefore, new protein analysis tools that extract functional features from proteins of flexible length and train models are required. We introduce DeepPI, a deep learning-based tool for analyzing proteins in large-scale database. The proposed model that utilizes Global Average Pooling is applied to proteins of flexible length and leads to reduced information loss compared to existing algorithms that use fixed sizes. The image generator converts a one-dimensional sequence into a distinct two-dimensional structure, which can extract common parts of various shapes. Finally, filtering techniques automatically detect representative data from the entire database and ensure coverage of large protein databases. We demonstrate that DeepPI has been successfully applied to large databases such as the Pfam-A database. Comparative experiments on four types of image generators illustrated the impact of structure on feature extraction. The filtering performance was verified by varying the parameter values and proved to be applicable to large databases. Compared to existing methods, DeepPI outperforms in family classification accuracy for protein function inference.

PMID:38568406 | DOI:10.1007/s12539-024-00618-x

Radiol Artif Intell. 2024 Apr 3:e230318. doi: 10.1148/ryai.230318. Online ahead of print.

ABSTRACT

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop an artificial intelligence (AI) for diagnosis of breast cancer in digital breast tomosynthesis (DBT) and investigate whether it could improve diagnostic accuracy and reduce reading time of radiologists. Materials and methods A deep learning AI algorithm was developed and validated for DBT with retrospectively collected examinations (January 2010 to December 2021) from 14 institutions in the United States and South Korea. A multicenter, reader study was performed to compare the performance of 15 radiologists (7 breast specialists, 8 general radiologists) in interpreting DBT examinations from 258 women (mean, 56 years ± 13.41 [SD]), including 65 cancer cases, with and without the use of AI. Area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and reading time were evaluated. Results The AUC for standalone AI performance was 0.93 (95% CI: 0.92,0.94). With AI, radiologists' AUC improved from 0.90 (0.86, 0.93) to 0.92 (0.88, 0.96; P = .003) in the reader study. AI showed higher specificity (89.64% (85.34, 93.94)) than radiologists (77.34% (75.82, 78.87; P < .001)). When reading with AI, radiologists' sensitivity increased from 85.44% (83.22, 87.65) to 87.69% (85.63, 89.75; P = .04), with no evidence of a difference in specificity. Reading time decreased from 54.41 seconds (52.56, 56.27) without AI to 48.52 seconds (46.79, 50.25) with AI (P < .001). Interreader agreement measured by Fleiss kappa increased from 0.59 to 0.62, respectively. Conclusion The AI model showed better diagnostic accuracy than radiologists in breast cancer detection and reduced reading times. The concurrent use of AI in DBT interpretation could improve both accuracy and efficiency. ©RSNA, 2024.

PMID:38568095 | DOI:10.1148/ryai.230318

Radiol Artif Intell. 2024 Apr 3:e230138. doi: 10.1148/ryai.230138. Online ahead of print.

ABSTRACT

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To investigate the accuracy and robustness of prostate segmentation using deep learning across various training data sizes, MRI vendors, prostate zones, and testing methods relative to fellowship-trained diagnostic radiologists. Materials and methods In this systematic review, EMBASE, PubMed, Scopus and Web of Science databases were queried for English-language articles using keywords and related terms for prostate MRI segmentation and deep learning algorithms dated to July 31, 2022. A total of 691 articles from the search query were collected, and subsequently filtered to 48 based on predefined inclusion and exclusion criteria. Multiple characteristics were extracted from selected studies, such as deep learning algorithm performance, MRI vendor, and training dataset features. The primary outcome was comparison of mean Dice similarity coefficient (DSC) for prostate segmentation for deep learning algorithms versus diagnostic radiologists. Results Forty-eight studies were included. The vast majority of published deep learning algorithms for whole prostate gland segmentation (39/42 or 93%) had a DSC at or above expert level (DSC ≥ 0.86). The mean DSC was 0.79 ± 0.06 for peripheral zone, 0.87 ± 0.05 for transition zone, and 0.90 ± 0.04 for whole prostate gland segmentation. For selected studies using one major MRI vendor, the mean DSCs of each were as follows: GE (3/48 studies) 0.92 ± 0.03, Philips (4/48 studies) 0.92 ± 0.02, and Siemens (6/48 studies) 0.91 ± 0.03. Conclusion Deep learning algorithms for prostate MRI segmentation demonstrated comparable accuracy to expert radiologists despite varying parameters, therefore future research should shift toward evaluating segmentation robustness and patient outcomes across diverse clinical settings. ©RSNA, 2024.

PMID:38568094 | DOI:10.1148/ryai.230138

J Clin Ultrasound. 2024 Apr 3. doi: 10.1002/jcu.23679. Online ahead of print.

ABSTRACT

Deep learning techniques have become crucial in the detection of brain tumors but classifying numerous images is time-consuming and error-prone, impacting timely diagnosis. This can hinder the effectiveness of these techniques in detecting brain tumors in a timely manner. To address this limitation, this study introduces a novel brain tumor detection system. The main objective is to overcome the challenges associated with acquiring a large and well-classified dataset. The proposed approach involves generating synthetic Magnetic Resonance Imaging (MRI) images that mimic the patterns commonly found in brain MRI images. The system utilizes a dataset consisting of small images that are unbalanced in terms of class distribution. To enhance the accuracy of tumor detection, two deep learning models are employed. Using a hybrid ResNet+SE model, we capture feature distributions within unbalanced classes, creating a more balanced dataset. The second model, a tailored classifier identifies brain tumors in MRI images. The proposed method has shown promising results, achieving a high detection accuracy of 98.79%. This highlights the potential of the model as an efficient and cost-effective system for brain tumor detection.

PMID:38567722 | DOI:10.1002/jcu.23679

Equine Vet J. 2024 Apr 3. doi: 10.1111/evj.14087. Online ahead of print.

ABSTRACT

BACKGROUND/OBJECTIVES: The aim was to compare ophthalmic diagnoses made by veterinarians to a deep learning (artificial intelligence) software tool which was developed to aid in the diagnosis of equine ophthalmic diseases. As equine ophthalmology is a very specialised field in equine medicine, the tool may be able to help in diagnosing equine ophthalmic emergencies such as uveitis.

STUDY DESIGN: In silico tool development and assessment of diagnostic performance.

METHODS: A deep learning tool which was developed and trained for classification of equine ophthalmic diseases was tested with 40 photographs displaying various equine ophthalmic diseases. The same data set was shown to different groups of veterinarians (equine, small animal, mixed practice, other) using an opinion poll to compare the results and evaluate the performance of the programme. Convolutional Neural Networks (CNN) were trained on 2346 photographs of equine eyes, which were augmented to 9384 images. Two hundred and sixty-one separate unmodified images were used to evaluate the trained network. The trained deep learning tool was used on 40 photographs of equine eyes (10 healthy, 12 uveitis, 18 other diseases). An opinion poll was used to evaluate the diagnostic performance of 148 veterinarians in comparison to the software tool.

RESULTS: The probability for the correct answer was 93% for the AI programme. Equine veterinarians answered correctly in 76%, whereas other veterinarians reached 67% probability for the correct diagnosis.

MAIN LIMITATIONS: Diagnosis was solely based on images of equine eyes without the possibility to evaluate the inner eye.

CONCLUSIONS: The deep learning tool proved to be at least equivalent to veterinarians in assessing ophthalmic diseases in photographs. We therefore conclude that the software tool may be useful in detecting potential emergency cases. In this context, blindness in horses may be prevented as the horse can receive accurate treatment or can be sent to an equine hospital. Furthermore, the tool gives less experienced veterinarians the opportunity to differentiate between uveitis and other ocular anterior segment disease and to support them in their decision-making regarding treatment.

PMID:38567426 | DOI:10.1111/evj.14087

Sci Prog. 2024 Apr-Jun;107(2):368504241232537. doi: 10.1177/00368504241232537.

ABSTRACT

Nasopharyngeal carcinoma is a malignant tumor that occurs in the epithelium and mucosal glands of the nasopharynx, and its pathological type is mostly poorly differentiated squamous cell carcinoma. Since the nasopharynx is located deep in the head and neck, early diagnosis and timely treatment are critical to patient survival. However, nasopharyngeal carcinoma tumors are small in size and vary widely in shape, and it is also a challenge for experienced doctors to delineate tumor contours. In addition, due to the special location of nasopharyngeal carcinoma, complex treatments such as radiotherapy or surgical resection are often required, so accurate pathological diagnosis is also very important for the selection of treatment options. However, the current deep learning segmentation model faces the problems of inaccurate segmentation and unstable segmentation process, which are mainly limited by the accuracy of data sets, fuzzy boundaries, and complex lines. In order to solve these two challenges, this article proposes a hybrid model WET-UNet based on the UNet network as a powerful alternative for nasopharyngeal cancer image segmentation. On the one hand, wavelet transform is integrated into UNet to enhance the lesion boundary information by using low-frequency components to adjust the encoder at low frequencies and optimize the subsequent computational process of the Transformer to improve the accuracy and robustness of image segmentation. On the other hand, the attention mechanism retains the most valuable pixels in the image for us, captures the remote dependencies, and enables the network to learn more representative features to improve the recognition ability of the model. Comparative experiments show that our network structure outperforms other models for nasopharyngeal cancer image segmentation, and we demonstrate the effectiveness of adding two modules to help tumor segmentation. The total data set of this article is 5000, and the ratio of training and verification is 8:2. In the experiment, accuracy = 85.2% and precision = 84.9% can show that our proposed model has good performance in nasopharyngeal cancer image segmentation.

PMID:38567422 | DOI:10.1177/00368504241232537

J Bioinform Comput Biol. 2024 Feb;22(1):2350030. doi: 10.1142/S0219720023500300.

ABSTRACT

The accurate identification of drug-target affinity (DTA) is crucial for advancements in drug discovery and development. Many deep learning-based approaches have been devised to predict drug-target binding affinity accurately, exhibiting notable improvements in performance. However, the existing prediction methods often fall short of capturing the global features of proteins. In this study, we proposed a novel model called ETransDTA, specifically designed for predicting drug-target binding affinity. ETransDTA combines convolutional layers and transformer, allowing for the simultaneous extraction of both global and local features of target proteins. Additionally, we have integrated a new graph pooling mechanism into the topology adaptive graph convolutional network (TAGCN) to enhance its capacity for learning feature representations of chemical compounds. The proposed ETransDTA model has been evaluated using the Davis and Kinase Inhibitor BioActivity (KIBA) datasets, consistently outperforming other baseline methods. The evaluation results on the KIBA dataset reveal that our model achieves the lowest mean square error (MSE) of 0.125, representing a 0.6% reduction compared to the lowest-performing baseline method. Furthermore, the incorporation of queries, keys and values produced by the stacked convolutional neural network (CNN) enables our model to better integrate the local and global context of protein representation, leading to further improvements in the accuracy of DTA prediction.

PMID:38567388 | DOI:10.1142/S0219720023500300