Limited Data Set Research Articles

An innovative variant based on generative adversarial network (GAN): Regression GAN combined with hyperspectral imaging to predict pesticide residue content of Hami melon

The rapid and non-destructive detection of pesticide residues in Hami melons plays substantial importance in protecting consumer health. However, the investment of time and resources needed to procure sample data poses a challenge, often resulting in limited data set and consequently leading insufficient accuracy of the established models. In this study, an innovative variant based on generative adversarial network (GAN) was proposed, named regression GAN (RGAN). It was used to synchronically extend the visible near-infrared (VNIR) and short-wave infrared (SWIR) hyperspectral data and corresponding acetamiprid residue content data of Hami melon. The support vector regression (SVR) and partial least squares regression (PLSR) models were trained using the generated data, and subsequently validate them with real data to assess the reliability of the generated data. In addition, the generated data were added to the real data to extend the dataset. The SVR model based on SWIR-HSI data achieved the optimal performance after data augmentation, yielding the values of Rp2, RMSEP and RPD were 0.8781, 0.6962 and 2.7882, respectively. The RGAN extends the range of GAN applications from classification problems to regression problems. It serves as a valuable reference for the quantitative analysis of chemometrics.

A Tourist Destination Recommender System for Nepali Tourism Industry

Tourism is one of the major sources of revenue in developing countries with natural beauty like Nepal. Understanding tourist preferences is essential to optimize this industry. Many places suffer obscurity due to not being recommended to the tourists who would’ve liked the destination. Our goal is to analyse what every tourist likes and recommend them accordingly. In case of Nepal, since there are a lot of under-explored potential tourist destinations and not many data available for the reviews and ratings, the data becomes sparse and limited. To tackle the challenge, our system utilizes a blend of content-based and collaborative filtering, which is known as hybrid filtering. The system leverages the strengths of Cosine Similarity, k-Nearest Neighbors, and Matrix Factorization to personalize recommendations based on user preferences and available destination information even in a sparse as well as limited dataset. This approach not only tackles the under-exploration of hidden gems but also presents a scalable framework applicable to any domain with limited data, potentially impacting personalized recommendations across various industries.

Risk Assessment and Diagnosis Code Prediction from Electronic Health Records (EHR) using deep learning

Knowledge in the medical field is often expressed by distinct and subjective norms. Much research and development efforts have focused on using deep learning algorithms to predict the likelihood of illnesses from EHR in recent years. When it comes to risk prediction, deep learning-based techniques outperform more conventional machine learning models. But nothing in the literature fully accounts for what doctors already know, such as the connections between illnesses and their risk factors. This research examines the use of Multi-layer Perceptron models for the categorisation of diagnoses in electronic health records. The raw data and a modified version of the EHR dataset are used to train two MLPs with distinct topologies. For comparative purposes, a Random Forest is used as a baseline. To phenotype patients using their electronic health records, we provide a deep learning method. Predictive modelling of chronic illnesses is the particular scenario used to verify the suggested model on a real-world EHR data warehouse. Many deep learning applications on EHRs have been effective, and there is still a lot of potential to be realised. It has been discovered that deep learning models can learn from the limited EHR dataset, but not to a level where they outperform the baseline model.

Multifunctional Image Enhancement Tool with Image Processing Methods

The success of image-based machine learning and deep learning algorithms depends on the amount and variety of data sets. The richness and heterogeneity of the dataset has a direct impact on the capacity of the model to generalize to new, previously unknown data. Obtaining a large number and variety of images is a very difficult and laborious process, especially for practical applications to be developed in areas such as health, industry and agriculture. If this difficulty cannot be overcome, the performance of machine learning models may decrease or may not reach the desired level at all. This problem may lead to problems such as poor performance or overfitting of the developed models. To solve this problem, data augmentation techniques are used to increase the amount and variety of data sets. These strategies help to provide the necessary data set for the model to learn more successfully. In this publication, we present an image improvement tool that is specifically developed for academics working with limited data sets. This tool uses a variety of data augmentation techniques, including rotation, zooming, panning, noise addition, and blurring, to expand the scale and diversity of available data. This application offers a realistic method for dealing with the issues of limited data availability in machine learning research.

MRDPDA: A multi-Laplacian regularized deepFM model for predicting piRNA-disease associations.

PIWI-interacting RNAs (piRNAs) are a typical class of small non-coding RNAs, which are essential for gene regulation, genome stability and so on. Accumulating studies have revealed that piRNAs have significant potential as biomarkers and therapeutic targets for a variety of diseases. However current computational methods face the challenge in effectively capturing piRNA-disease associations (PDAs) from limited data. In this study, we propose a novel method, MRDPDA, for predicting PDAs based on limited data from multiple sources. Specifically, MRDPDA integrates a deep factorization machine (deepFM) model with regularizations derived from multiple yet limited datasets, utilizing separate Laplacians instead of a simple average similarity network. Moreover, a unified objective function to combine embedding loss about similarities is proposed to ensure that the embedding is suitable for the prediction task. In addition, a balanced benchmark dataset based on piRPheno is constructed and a deep autoencoder is applied for creating reliable negative set from the unlabeled dataset. Compared with three latest methods, MRDPDA achieves the best performance on the pirpheno dataset in terms of the five-fold cross validation test and independent test set, and case studies further demonstrate the effectiveness of MRDPDA.

HarmonyTM: multi-center data harmonization applied to distributed learning for Parkinson's disease classification.

Distributed learning is widely used to comply with data-sharing regulations and access diverse datasets for training machine learning (ML) models. The traveling model (TM) is a distributed learning approach that sequentially trains with data from one center at a time, which is especially advantageous when dealing with limited local datasets. However, a critical concern emerges when centers utilize different scanners for data acquisition, which could potentially lead models to exploit these differences as shortcuts. Although data harmonization can mitigate this issue, current methods typically rely on large or paired datasets, which can be impractical to obtain in distributed setups. We introduced HarmonyTM, a data harmonization method tailored for the TM. HarmonyTM effectively mitigates bias in the model's feature representation while retaining crucial disease-related information, all without requiring extensive datasets. Specifically, we employed adversarial training to "unlearn" bias from the features used in the model for classifying Parkinson's disease (PD). We evaluated HarmonyTM using multi-center three-dimensional (3D) neuroimaging datasets from 83 centers using 23 different scanners. Our results show that HarmonyTM improved PD classification accuracy from 72% to 76% and reduced (unwanted) scanner classification accuracy from 53% to 30% in the TM setup. HarmonyTM is a method tailored for harmonizing 3D neuroimaging data within the TM approach, aiming to minimize shortcut learning in distributed setups. This prevents the disease classifier from leveraging scanner-specific details to classify patients with or without PD-a key aspect for deploying ML models for clinical applications.

Patient Factors Associated With Reimplantation After Girdlestone Resection Arthroplasty for Treatment of Periprosthetic Joint Infections of the Hip.

For patients unsuitable for prosthesis reimplantation or temporary spacer placement, Girdlestone resection arthroplasty (GRA) is a suitable option to eliminate infection. Using a large-scale database, this study aims to determine factors associated with reimplantation. This study included patients who underwent GRA and subsequent total hip arthroplasty (2012 to 2015 Medicare Limited Data Set with ≥5-year follow-up). A mixed-effects model measured associations between patient characteristics and reimplantation. Odds ratios (OR) with 95% confidence intervals (CI) were reported. Among 2,772 GRA cases, 2,025 (73.1%) were reimplanted (median time to reimplantation 3.0 months). In multivariable analysis, patient factors associated with reduced odds of reimplantation were increased age (OR 0.96; CI, 0.94 to 0.97; P < 0.0001), Black race (OR, 0.58; CI, 0.37 to 0.90; P = 0.0149), obesity (OR, 0.74; CI, 0.58 to 0.94; P = 0.0150), and increased Deyo-Charlson comorbidities (1 comorbidity: OR, 0.78; CI, 0.61 to 0.99; P = 0.0453; two comorbidities: OR, 0.53; CI, 0.39 to 0.71; P < 0.0001; ≥3 comorbidities: OR, 0.69; CI, 0.49 to 0.95; P = 0.0244). Male (versus female) patients, however, had increased odds of reimplantation (OR, 1.64; CI, 1.32 to 2.02; P < 0.0001). Age, race, and comorbidities influence the likelihood of reimplantation after GRA. Owing to variability in patients who undergo additional surgery, additional studies should be conducted to determine the rationale of patient selection.

Generating Contrast-Enhanced Liver MRI Images from Native Sequences

Abstract Generating synthetic contrast-enhanced liver MRI scans from native MRI images can serve to mitigate the issue of sparse contrast-enhanced image datasets while concurrently circumventing the time-consuming and costly process of administering contrast agents during image acquisition. In this study, we conducted three experiments using paired image-to-image translation techniques. Native T1 sequences showing the abdominal liver region served as the input, while contrast-enhanced T1 sequences were the target. The data preprocessing methods and image boundaries were varied for the individual experiments in addition to the implementation of a 5-fold cross-validation for the top-performing approach. Focusing on liver regions achieved the best results with a mean absolute error (MAE) of 0.0837 ± 0.0068, a mean squared error (MSE) of 0.0128 ± 0.0023 and a peak signal-to-noise ratio (PSNR) of 18.99 ± 0.81 dB. Our findings serve as a proof-of-concept, demonstrating the feasibility of generating contrast-enhanced MRI images. However, the current state necessitates further enhancements for effectively addressing the challenge posed by limited dataset sizes with difficult anatomical circumstances as well as MR imaging-related heterogenous tissue contrasts.

Flood inundation monitoring using multi-source satellite imagery: a knowledge transfer strategy for heterogeneous image change detection

Flood emergency mapping is essential for flood management, often requiring near real-time extraction of large-scale flood extents by combining pre- and post-event multi-source remote sensing images. Pre-event optical imagery delineates the normal water extent, providing a benchmark for estimation of post-flood changes. Synthetic aperture radar (SAR) imagery provides rapid and accurate interpretation of flood inundation extent during the rainy period. Post-classification comparison is one of the fundamental methods for flood mapping of multi-source imagery, as the image characteristics of SAR and optical imagery are inherently different. However, the accuracy of flood mapping depends on the accuracy of water body delineation from single temporal imagery, which can lead to error propagation in flood extent determination. Change detection methods can extract flood extent directly from pre- and post-event imagery, but the models need to learn a consistent flood feature description between optical and SAR imagery from a limited flood dataset. Here, we propose an automatic method for flood inundation extent extraction using pre- and post-event multi-source imagery that requires only a small number of pseudo-change labels. The methodology adopts a pseudo-Siamese network framework as a Heterogeneous - Flood Inundation Extraction Network (H-FIENet), to detect flood extent between the pre‐ and post-event heterogeneous images. Spatially inconsistent multi-source pre- and post-event imagery was used to create a pseudo flood dataset, and this dataset was used to transfer water body feature knowledge from the pre-trained model to the flood extraction model using a cross-task knowledge transfer strategy. Pre- and post-event images from different satellite sources and different flood scenarios were used to evaluate the performance of H-FIENet. We found that: (1) The methodology produced an overall accuracy of over 0.94 for flood inundation extraction from multi-source heterogeneous pre- and post-event imagery. (2) H-FIENet can detect both expansion and recession of the flood extent using any dual-temporal imagery. Our work makes it possible to automate time-series flood monitoring from multi-source remote sensing imagery.

Energy absorption performance of Kresling origami tubes under impact loading

Thin-walled tubes with an origami design, particularly the Kresling pattern, exhibit superior mechanical properties compared to traditional straight tubes, including a more constant reaction force and predictable deformation. Despite their potential, research on these patterned structures, especially when made from structural materials like metal and tested under dynamic conditions, remains limited. This study investigates the compressive performance of aluminium Kresling origami tubes (KOTs) under quasi-static and impact scenarios (up to 30 m/s) in the axial direction. Results show that increased impact velocity leads to more localized deformation and improved energy absorption metrics. A validated numerical model was used to analyze the influence of hierarchy rotation, sector angles, and loading velocity on mechanical performance. Comparisons with Miura-ori patterned tubes and hexagonal cross-section straight tubes of the same relative density revealed that KOTs have superior energy absorption performance. An empirical model was developed to effectively predict the mean crushing stress of KOTs. In addition. a generative machine learning model was introduced to synthesize a large dataset from initial simulations, providing an efficient and reliable solution for energy absorption analysis in origami structures, addressing the challenge of limited specimen datasets.

Evaluation of synthetic clear gelatin as an acceptable surrogate for low-velocity penetrating impacts using the depth of penetration calibration standard

Ballistic gelatin has been extensively used in ballistics research for decades, but calibration standards were established on limited datasets, and only few studies have attempted to recreate these experiments with biological tissues. Recent studies have demonstrated better biofidelity with 20% ordnance ballistic gelatin, but researchers have discredited the use of synthetic gelatin claiming different behavior than ordnance gelatin. To investigate the use of synthetic clear gelatin as an acceptable surrogate of biological tissue, depth of penetration was compared between low-velocity impacts of various projectiles into porcine tissue (n = 192), post-mortem human subjects (n = 29), and Clear Ballistics synthetic gelatin (n = 39). The predicted depth of penetration of the 0.177" steel BB (38.1 mm) was consistent with the manufacturer's calibration standard (31.75–44.45 mm) and within calibration bounds of recently proposed empirical equations. Compared to impacts in biological tissue, synthetic gelatin demonstrated the least variability in depth of penetration (R2 = 0.96). Using ANCOVA, velocity was a significant covariate (p < 0.001), and there were no significant differences in normalized depth of penetration over density between porcine tissue, post-mortem human subjects, and 20% synthetic gelatin (p = 0.22). Ultimately, this study confirmed the use of 20% synthetic gelatin as an acceptable tissue simulant using standard calibration methods for use in future ballistic studies.

Beyond pixel: Superpixel-based MRI segmentation through traditional machine learning and graph convolutional network

Background and Objective:Tendon segmentation is crucial for studying tendon-related pathologies like tendinopathy, tendinosis, etc. This step further enables detailed analysis of specific tendon regions using automated or semi-automated methods. This study specifically aims at the segmentation of Achilles tendon, the largest tendon in the human body. Methods:This study proposes a comprehensive end-to-end tendon segmentation module composed of a preliminary superpixel-based coarse segmentation preceding the final segmentation task. The final segmentation results are obtained through two distinct approaches. In the first approach, the coarsely generated superpixels are subjected to classification using Random Forest (RF) and Support Vector Machine (SVM) classifiers to classify whether each superpixel belongs to a tendon class or not (resulting in tendon segmentation). In the second approach, the arrangements of superpixels are converted to graphs instead of being treated as conventional image grids. This classification process uses a graph-based convolutional network (GCN) to determine whether each superpixel corresponds to a tendon class or not. Results:All experiments are conducted on a custom-made ankle MRI dataset. The dataset comprises 76 subjects and is divided into two sets: one for training (Dataset 1, trained and evaluated using leave-one-group-out cross-validation) and the other as unseen test data (Dataset 2). Using our first approach, the final test AUC (Area Under the ROC Curve) scores using RF and SVM classifiers on the test data (Dataset 2) are 0.992 and 0.987, respectively, with sensitivities of 0.904 and 0.966. On the other hand, using our second approach (GCN-based node classification), the AUC score for the test set is 0.933 with a sensitivity of 0.899. Conclusions:Our proposed pipeline demonstrates the efficacy of employing superpixel generation as a coarse segmentation technique for the final tendon segmentation. Whether utilizing RF, SVM-based superpixel classification, or GCN-based classification for tendon segmentation, our system consistently achieves commendable AUC scores, especially the non-graph-based approach. Given the limited dataset, our graph-based method did not perform as well as non-graph-based superpixel classifications; however, the results obtained provide valuable insights into how well the models can distinguish between tendons and non-tendons. This opens up opportunities for further exploration and improvement.

Investigating the missing-wedge problem in small-angle X-ray scattering tensor tomography across real and reciprocal space.

Small-angle-scattering tensor tomography is a technique for studying anisotropic nanostructures of millimetre-sized samples in a volume-resolved manner. It requires the acquisition of data through repeated tomographic rotations about an axis which is subjected to a series of tilts. The tilt that can be achieved with a typical setup is geometrically constrained, which leads to limits in the set of directions from which the different parts of the reciprocal space map can be probed. Here, we characterize the impact of this limitation on reconstructions in terms of the missing wedge problem of tomography, by treating the problem of tensor tomography as the reconstruction of a three-dimensional field of functions on the unit sphere, represented by a grid of Gaussian radial basis functions. We then devise an acquisition scheme to obtain complete data by remounting the sample, which we apply to a sample of human trabecular bone. Performing tensor tomographic reconstructions of limited data sets as well as the complete data set, we further investigate and validate the missing wedge problem by investigating reconstruction errors due to data incompleteness across both real and reciprocal space. Finally, we carry out an analysis of orientations and derived scalar quantities, to quantify the impact of this missing wedge problem on a typical tensor tomographic analysis. We conclude that the effects of data incompleteness are consistent with the predicted impact of the missing wedge problem, and that the impact on tensor tomographic analysis is appreciable but limited, especially if precautions are taken. In particular, there is only limited impact on the means and relative anisotropies of the reconstructed reciprocal space maps.

Constraint based Bayesian optimization of bioink precursor: a machine learning framework.

Current research practice for optimizing bioink involves exhaustive experimentation with multi-material composition for determining the printability, shape fidelity and biocompatibility. Predicting bioink properties can be beneficial to the research community but is a challenging task due to the non-Newtonian behavior in complex composition. Existing models such as Cross model become inadequate for predicting the viscosity for heterogeneous composition of bioinks. In this paper, we utilize a machine learning framework to accurately predict the viscosity of heterogeneous bioink compositions, aiming to enhance extrusion-based bioprinting techniques. Utilizing Bayesian optimization (BO), our strategy leverages a limited dataset to inform our model. This is a technique especially useful of the typically sparse data in this domain. Moreover, we have also developed a mask technique that can handle complex constraints, informed by domain expertise, to define the feasible parameter space for the components of the bioink and their interactions. Our proposed method is focused on predicting the intrinsic factor (e.g. viscosity) of the bioink precursor which is tied to the extrinsic property (e.g. cell viability) through the mask function. Through the optimization of the hyperparameter, we strike a balance between exploration of new possibilities and exploitation of known data, a balance crucial for refining our acquisition function. This function then guides the selection of subsequent sampling points within the defined viable space and the process continues until convergence is achieved, indicating that the model has sufficiently explored the parameter space and identified the optimal or near-optimal solutions. Employing this AI-guided BO framework, we have developed, tested, and validated a surrogate model for determining the viscosity of heterogeneous bioink compositions. This data-driven approach significantly reduces the experimental workload required to identify bioink compositions conducive to functional tissue growth. It not only streamlines the process of finding the optimal bioink compositions from a vast array of heterogeneous options but also offers a promising avenue for accelerating advancements in tissue engineering by minimizing the need for extensive experimental trials.

Smart customer service in unmanned retail store enhanced by large language model

In unmanned retail store, providing smart customer service requires two stages: understanding customer needs, and guiding the customer to the product. In this paper, we propose an end-to-end (Customer-to-Shelf) software service framework for unmanned retail. The framework integrates visual recognition technology to detect retail objects, large language models to analyze customer shopping needs and make proper recommendations. First, deep neural network based image recognition models are studied for implementing effective stock keeping units (SKUs) object recognition on the shelf. Second, a novel method is proposed to fine-tune large language models (LLMs) with limited training dataset. Metaheuristic approaches are used to optimize the mask locations in a low dimensional parameter space, resulting a more efficient parameter updating method for limited downstream data. Third, by facilitating an automatic analysis of customer preferences powered by large language models, we present a smart recommender system based on domain-specific knowledge, which completes the Customer-to-Shelf software service framework. Experimental results show that our proposed fine-tuning method, is more efficient than other state-of-the-art training methods for limited downstream domain dataset. Using fine-tuned large models, we can successfully create a seamless shopping experience for customers by understanding personalized needs and providing shopping advice in the unmanned retail store.

Learning from the few: Fine-grained approach to pediatric wrist pathology recognition on a limited dataset

Wrist pathologies, particularly fractures common among children and adolescents, present a critical diagnostic challenge. While X-ray imaging remains a prevalent diagnostic tool, the increasing misinterpretation rates highlight the need for more accurate analysis, especially considering the lack of specialized training among many surgeons and physicians. Recent advancements in deep convolutional neural networks offer promise in automating pathology detection in trauma X-rays. However, distinguishing subtle variations between pediatric wrist pathologies in X-rays remains challenging. Traditional manual annotation, though effective, is laborious, costly, and requires specialized expertise. In this paper, we address the challenge of pediatric wrist pathology recognition with a fine-grained approach, aimed at automatically identifying discriminative regions in X-rays without manual intervention. We refine our fine-grained architecture through ablation analysis and the integration of LION. Leveraging Grad-CAM, an explainable AI technique, we highlight these regions. Despite using limited data, reflective of real-world medical study constraints, our method consistently outperforms state-of-the-art image recognition models on both augmented and original (challenging) test sets. Our proposed refined architecture achieves an increase in accuracy of 1.06% and 1.25% compared to the baseline method, resulting in accuracies of 86% and 84%, respectively. Moreover, our approach demonstrates the highest fracture sensitivity of 97%, highlighting its potential to enhance wrist pathology recognition.

Incidence of Dementia Among Medicare Beneficiaries With and Without Urinary Incontinence.

Dementia and urinary incontinence (UI) have been associated in cross-sectional studies. The temporal relationship between these 2 conditions is not well understood. The aim of the study was to investigate the risk of incident dementia in female adults with and without UI using the Medicare 5% Limited Data Set. This retrospective cohort study matched females with UI to continent controls by index year, age, and dual Medicare/Medicaid eligibility. A 2-year look back period was used to exclude prevalent dementia and neurologic disorders. The 5-year cumulative incidence of dementia was estimated for women with UI and controls. Cox proportional hazard models assessed the association of variables with dementia, adjusting for age, dual Medicare/Medicaid eligibility, race, and comorbidities. A total of 8,651 female beneficiaries with UI (n = 8,651) were more likely than matched controls (n = 8,651) to report White race and several medical comorbidities. The 5-year cumulative incidence of dementia diagnosis was lower in the UI versus controls (8.8% vs 10.6%, P < 0.001). In multivariable analysis with adjustment for covariates, UI diagnosis was associated with a lower hazard of dementia diagnosis (hazard ratio 0.82 [0.74, 0.91], P < 0.001). Among female Medicare beneficiaries without baseline neurologic disorders, having any UI diagnosis was associated with a lower risk of dementia diagnosis. Further studies assessing UI symptoms and dementia diagnosis with rigorous and valid assessment tools are needed to confirm this finding.

Enhancing wind power generation prediction using relevance assessment-based transfer learning

Accurate wind power generation forecasting can help build a reliable grid; however, the limited dataset makes accurate forecasting results a challenging work. This study introduces a relevant assessment-based transfer learning architecture to solve this problem. Linear fuzzy neighborhood mutual information is adopted to assess the relevance of the source domain selection. A convolutional structure with long-term memory architecture is designed as the deep learning model. The pre-trained model is transferred to the other wind turbines by calculating the linear fuzzy neighborhood mutual information. The proposed model avoids the lack of a dataset from analogous turbines. The simulation results indicate the proposed model surpasses the popular models in forecasting accuracy and exhibits superior time efficiency compared to popular deep-learning approaches.

Optimization of optical spectroscopy classification algorithms for limited data scenarios in the food industry: Tomato sauce samples case

- This study addresses the problem of training deep learning models with limited datasets, a significant challenge in sectors like medical imaging and food quality analysis. To tackle this issue, generative adversarial networks (GANs) will be employed to augment the available data and improve model performance. An innovative approach is introduced here, integrating semi-supervised learning and generative modeling to maximize the use of small datasets in developing robust models. The method involves reversing the conventional distribution of training and testing data to focus on model evaluation and generalization from limited samples. Wasserstein GANs (WGANs) and Semi-Supervised GANs (SGANs), are utilized to supplement datasets with synthetic but realistic examples, enhancing the training process in scenarios of data scarcity. These techniques are applied in the context of visible reflectance spectroscopy to analyze tomato sauces, demonstrating the method's effectiveness in non-invasively assessing key quality parameters such as oil content, °Brix, and pH. The results show significant improvements in model performance metrics: for %Oil content, overall accuracy increased from 0.47 to 0.66; for °Bx, it rose from 0.65 to 0.71; and for pH measurement, accuracy improved from 0.43 to 0.62. These outcomes highlight the model's improved capability to generalize and maintain accuracy with limited data.

Thy-DAMP: deep artificial neural network model for prediction of thyroid cancer mortality.

Despite the rising incidence of differentiated thyroid cancer (DTC), mortality rates have remained relatively low yet crucial for effective patient management. This study aims to develop a deep neural network capable of predicting mortality in patients with differentiated thyroid cancer. Leveraging data from the Surveillance, Epidemiology, and End Results (SEER) database, we developed Thy-DAMP (Deep Artificial Neural Network Model for Prediction of Thyroid Cancer) to forecast mortality in DTC patients. The dataset comprised demographic, histologic, and staging information. Following data normalization and feature encoding, the dataset was partitioned into training, testing, and validation subsets, with model hyperparameters fine-tuned via cross-validation. Among the 63,513 patients, the mean age was 48.22 years (SD = 14.8), with 77.32% being female. Papillary carcinoma emerged as the predominant subtype, representing 62.94% of cases. The majority presented with stage I disease (73.96%). Thy-DAMP demonstrated acceptable performance metrics on both the test and validation sets. Sensitivity was 83.24% (95% CI 76.95-88.40%), specificity was 93.53% (95% CI 93.01-94.02%), and accuracy stood at 93.33% (95% CI 92.82-93.83%). The model exhibited a positive predictive value of 19.76% (95% CI 18.20-21.42%) and a negative predictive value of 99.66% (95% CI 99.53-99.75%). Additionally, Thy-DAMP demonstrated a positive likelihood ratio of 12.86 (95% CI 11.62-14.23), a negative likelihood ratio of 0.18 (95% CI 0.13-0.25), and an area under the receiver operating characteristic curve (AUROC) of 0.95. The model was externally validated on a separate dataset with nearly identical performance. Thy-DAMP showcases considerable promise in accurately predicting mortality in DTC patients, leveraging limited set of patient data.