Development and Preliminary Validation of a Low-Cost Portable Tool for Assessing Joint Mobility

To evaluate continuous and non-invasive sweat-inferred blood lactate monitoring for determining the exercise intensity at the second metabolic threshold (MT2) and maximal lactate steady state (MLSS), compared to capillary blood lactate and gas exchange analysis. 17 physically active individuals (11 males and 6 females) and 19 endurance athletes (14 males and 5 females) completed a maximal graded exercise test (GXT) to quantify oxygen uptake (VO2), percentage of peak oxygen uptake (%VO2peak) and power output (watts) at MT2, using sweat-inferred blood lactate and capillary blood lactate for the second lactate threshold (LT2), and gas exchange analysis for the second ventilatory threshold (VT2). Participants also completed a MLSS test to quantify the power output (watts) at MLSS. Between-methods agreement was assessed using linear mixed models with mean differences (MD), mean absolute error (MAE), mean absolute percentage error (MAPE), Lin's concordance correlation coefficient (CCC) and Bland-Altman analysis. Compared with VT2, LT2 derived from sweat-inferred blood lactate showed no significant differences when using the Log-Exp-Mod-Dmax method, yielding the strongest agreement across VO2 (MD = 0.3 mL·min -1·kg -1, p = 0.999; MAE = 2.1 mL·min -1·kg -1, MAPE = 5.2%), %VO2peak (MD = 0.4%, p = 0.999; MAE = 4.1%, MAPE = 5.2%), and power output (MD = 1.9W, p = 0.999; MAE = 9.1W, MAPE = 5.0%). Direct comparison of capillary blood and sweat-inferred blood lactate using the Log-Exp-Mod-Dmax method showed high agreement for VO2 (MD = 0.2 mL·min -1·kg -1, p = 0.999; MAE = 2.4 mL·min -1·kg -1, MAPE = 6.4%), %VO2peak (MD = -0.1%, p = 0.999; MAE = 4.9%, MAPE = 6.4%), and power output (MD = 0.0W, p = 0.999; MAE = 10.6W, MAPE = 6.1%). Capillary blood and sweat-inferred blood lactate also showed excellent agreement for MLSS-associated power output (MD = -2.0 W, p = 0.089; MAE = 3.1 W, MAPE = 2.5%). Continuous sweat-inferred blood lactate can be used to determine MT2 and MLSS intensities, providing an alternative to capillary blood lactate.

Uniaxial compressive strength (UCS) is one of the most important geomechanical parameters for evaluating rock strength along the wellbore. It is essential for ensuring drilling safety and optimizing drilling operations, particularly in determining the appropriate mud weight window to maintain well stability and improve the rate of penetration. Additionally, the UCS parameter plays a significant role in predicting sand production. Typically, UCS is determined through core sample tests in the laboratory, but this method is costly and time-consuming and provides either scattered data nor continuous log profile. Therefore, many studies around the world have proposed correlation between cored UCS and well log. However, these models often exhibit significant errors when applied to field X, Block Y in the Northern area of Cuu Long basin. In this paper, we propose a correlation between P-wave modulus (M) and Uniaxial compressive strength (UCS) derived from hydraulic flow units. Consequently, the correlation coefficient between cored UCS and log-derived UCS values is very high, with an overall value of 0.82 for all wells. In particular, the mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE) are as follows: for well A-1X — MAE: 426.3 psi, MAPE: 10.4%, RMSE: 582.5 psi; for well A-2X — MAE: 843.17 psi, MAPE: 16%, RMSE: 1115.5 psi; for well A-3X — MAE: 321.9 psi, MAPE: 7%, RMSE: 385.6 psi; and for well A-4X — MAE: 286.46 psi, MAPE: 10%, RMSE: 438.76 psi. The blind-test well B-1X shows acceptable errors, with MAE: 335.5 psi, MAPE: 14%, and RMSE: 383.37 psi. This model enhances the efficiency and safety of drilling and production operations in Field X, Block Y, located in the northeastern part of the Cuu Long Basin.

Globally, wind energy if properly harnessed, could serve as a source of energy generation in Africa. This study compared the performance of two Machine Learning (ML) algorithms (Linear regression and Random Forest) in predicting wind speed in five major cities in Africa (Yaoundé, Pretoria, Nairobi, Cairo and Abuja). Wind data were collected between January 1, 2000, and December 31, 2022, using the Solar Radiation Data Archive. The data preprocessing was carried out with 80% of the data used for training and 20% for validation. The performance of these ML algorithms was evaluated using Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and coefficient of determination (R2). The result shows that Nairobi (3.814795 m/s) closely followed by Cairo (3.606453 m/s) has the highest mean wind speed while Yaoundé (1.090512 m/s) has the lowest. Based on the performance metrics used, the two Machine Learning algorithms were competitive. Still, the Linear Regression (LR) algorithm outperformed the Random Forest Algorithm in predicting wind speed in all the selected major African cities. In Yaoundé (RMSE = 0.3892, MAE= 0.3001, MAPE =0.5030), Pretoria (RMSE=1.2339, MAE=0.9480, MAPE=0.7450) Nairobi (RMSE= 0.4223, MAE =0.6499, MAPE =0.1872), Nairobi (RMSE=0.6499, MAE=0.5171, MAPE =0.1872), Cairo (RMSE =1.0909, MAE =0.8544, MAPE =0.3541) and Abuja (RMSE = 0.70245, MAE =0.5441, MAPE= 0.4515) the Linear regression algorithms was found to outperformed Random Forest Regression. Therefore, the Linear regression algorithm is more reliable in predicting wind speed compared with the Random Forest regression.

BackgroundInfections caused by multidrug-resistant organisms (MDROs) continue to pose serious challenges for hospital infection control, often resulting in longer hospitalizations, increased patient morbidity, and higher healthcare costs. While time series forecasting has gained traction as a tool for anticipating MDROs trends, there remains a lack of real-world studies comparing the effectiveness of different modeling approaches using hospital-based data.ObjectiveThis study aimed to evaluate and compare the predictive performance of four time series models—SARIMA, ETS, Prophet, and NNETAR—using monthly MDROs infection data collected from a tertiary hospital in China between 2014 and 2023, with the goal of forecasting trends for 2024.MethodsMonthly MDROs infection rates from January 2014 to December 2023 were analyzed using R software. Stationarity was assessed through unit root tests, and appropriate differencing was applied as needed. Each model was fitted to the training dataset and used to forecast infection rates for the year 2024. Model accuracy was assessed by comparing forecasted values with actual 2024 data using root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), symmetric mean absolute percentage error (sMAPE), and mean absolute scaled error (MASE).ResultsAmong the models, SARIMA produced the most consistent and reliable forecasts (RMSE = 0.0469, MAE = 0.0424, MAPE = 20.74%, sMAPE = 21.27%, MASE = 0.932), with residuals satisfying tests for independence and normality. Although the ETS model achieved lower numerical point errors (RMSE = 0.0367, MAE = 0.0305, MAPE = 14.46%, sMAPE = 14.81%, MASE = 0.670), its residual diagnostics raised concerns regarding robustness. The Prophet (RMSE = 0.0499, MAE = 0.0439, MAPE = 20.41%, sMAPE = 22.15%, MASE = 0.563) and NNETAR (RMSE = 0.0697, MAPE = 30.60%, sMAPE = 30.60%, MASE = 0.072) models captured certain aspects of the data dynamics but showed lower overall robustness compared with SARIMA.ConclusionBased on its overall robustness and diagnostic consistency, SARIMA is recommended for short- to medium-term forecasting of MDROs infection trends. The other models, while less reliable on their own, may still be valuable for validating trends and conducting sensitivity analyses to support hospital infection control planning.

BackgroundInfluenza in mainland China results in a large number of outpatient and emergency visits related to influenza-like illness (ILI) annually. While deep learning models show promise for improving influenza forecasting, their technical complexity remains a barrier to practical implementation. Large language models, such as ChatGPT, offer the potential to reduce these barriers by supporting automated code generation, debugging, and model optimization.ObjectiveThis study aimed to evaluate the predictive performance of several deep learning models for ILI positive rates in mainland China and to explore the auxiliary role of ChatGPT-assisted development in facilitating model implementation.MethodsILI positivity rate data spanning from 2014 to 2024 were obtained from the Chinese National Influenza Center (CNIC) database. In total, 5 deep learning architectures—long short-term memory (LSTM), neural basis expansion analysis for time series (N-BEATS), transformer, temporal fusion transformer (TFT), and time-series dense encoder (TiDE)—were developed using a ChatGPT-assisted workflow covering code generation, error debugging, and performance optimization. Models were trained on data from 2014 to 2023 and tested on holdout data from 2024 (weeks 1‐39). Performance was evaluated using mean squared error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE).ResultsILI trends exhibited clear seasonal patterns with winter peaks and summer troughs, alongside marked fluctuations during the COVID-19 pandemic period (2020‐2022). All 5 deep learning models were successfully constructed, debugged, and optimized with the assistance of ChatGPT. Among the 5 models, TiDE achieved the best predictive performance nationally (MAE=5.551, MSE=43.976, MAPE=72.413%) and in the southern region (MAE=7.554, MSE=89.708, MAPE=74.475%). In the northern region, where forecasting proved more challenging, TiDE still performed best (MAE=4.131, MSE=28.922), although high percentage errors remained (MAPE>400%). N-BEATS demonstrated the second-best performance nationally (MAE=9.423) and showed greater stability in the north (MAE=6.325). In contrast, transformer and TFT consistently underperformed, with national MAE values of 10.613 and 12.538, respectively. TFT exhibited the highest deviation (national MAPE=169.29%). Extreme regional disparities were observed, particularly in northern China, where LSTM and TFT generated MAPE values exceeding 1918%, despite LSTM’s moderate performance in the south (MAE=9.460).ConclusionsDeep learning models, particularly TiDE, demonstrate strong potential for accurate ILI forecasting across diverse regions of China. Furthermore, large language models like ChatGPT can substantially enhance modeling efficiency and accessibility by assisting nontechnical users in model development. These findings support the integration of AI-assisted workflows into epidemic prediction systems as a scalable approach for improving public health preparedness.

To compare the application of the ARIMA model, the Long Short-Term Memory (LSTM) model and the ARIMA-LSTM model in forecasting foodborne disease incidence. Monthly case data of foodborne diseases in Liaoning Province from January 2015 to December 2023 were used to construct ARIMA, LSTM, and ARIMA-LSTM models. These three models were then applied to forecast the monthly incidence of foodborne diseases in 2024, and their predictions were compared with those of a baseline model. Model performance was evaluated by comparing the predicted and observed values using root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE), allowing identification of the optimal model. The best-performing model was subsequently employed to predict the monthly incidence for 2025. The ARIMA-LSTM model was identified as the optimal model. Specifically, the ARIMA (2,0,0) (0,1,1)12 model produced RMSE = 300.03, MAE = 187.11, and MAPE = 16.38%, while the LSTM model yielded RMSE = 408.71, MAE = 226.03, and MAPE = 17.21%. In contrast, the ARIMA-LSTM model achieved RMSE = 0.44, MAE = 0.44, and MAPE = 0.08%, representing a dramatic improvement over the baseline model (RMSE = 204.17, MAE = 146.75, MAPE = 15.62%), with reductions of 99.5%, 99.7%, and 99.4% in RMSE, MAE, and MAPE, respectively. Based on the ARIMA-LSTM model, the predicted monthly cases of foodborne diseases for 2025 are: 214.62 (Jan), 260.84 (Feb), 462.92 (Mar), 590.92 (Apr), 800.88 (May), 965.11 (Jun), 2410.36 (Jul), 2651.36 (Aug), 1711.15 (Sep), 941.22 (Oct), 628.21 (Nov), and 465.05 (Dec). The ARIMA-LSTM model is considered the optimal model for predicting foodborne disease incidence in Liaoning Province in 2025.

Lung transplantation is the standard treatment for end-stage lung diseases. A crucial factor affecting its success is size matching between the donor's lungs and the recipient's thorax. Computed tomography (CT) scans can accurately determine recipient's lung size, but donor's lung size is often unknown due to the absence of medical images. We aim to predict donor's right/left/total lung volume, thoracic cavity, and heart volume from only subject demographics to improve the accuracy of size matching. A cohort of 4610 subjects with chest CT scans and basic demographics (i.e., age, gender, race, smoking status, smoking history, weight, and height) was used in this study. The right and left lungs, thoracic cavity, and heart depicted on chest CT scans were automatically segmented using U-Net, and their volumes were computed. Eight machine learning models [i.e., random forest, multivariate linear regression, support vector machine, extreme gradient boosting (XGBoost), multilayer perceptron (MLP), decision tree, -nearest neighbors, and Bayesian regression) were developed and used to predict the volume measures from subject demographics. The 10-fold cross-validation method was used to evaluate the performances of the prediction models. -squared ( ), mean absolute error (MAE), and mean absolute percentage error (MAPE) were used as performance metrics. The MLP model demonstrated the best performance for predicting the thoracic cavity volume ( : 0.628, MAE: 0.736L, MAPE: 10.9%), right lung volume ( : 0.501, MAE: 0.383L, MAPE: 13.9%), and left lung volume ( : 0.507, MAE: 0.365L, MAPE: 15.2%), and the XGBoost model demonstrated the best performance for predicting the total lung volume ( : 0.514, MAE: 0.728L, MAPE: 14.0%) and heart volume ( : 0.430, MAE: 0.075L, MAPE: 13.9%). Our results demonstrate the feasibility of predicting lung, heart, and thoracic cavity volumes from subject demographics with superior performance compared with available studies in predicting lung volumes.

IntroductionIn recent years, load-velocity profiles (LVP) have been frequently proposed as a highly reliable and valid alternative to the one-repetition maximum (1RM) for estimating maximal strength and prescribing training loads. However, previous authors commonly report intraclass correlation coefficients (ICC) while neglecting to calculate the measurement error associated with these values. This is important for practitioners, especially in an elite sports setting, to be able to differentiate between small but significant changes in performance and the error rate.Methods49 youth elite athletes (17.71±2.07 years) were recruited and performed a 1RM test followed by a load-velocity profiling test using 30%, 50% and 70% of the 1RM in the bench press and bench pull, respectively. Reliability analysis, ICCs and the coefficient of variability, were calculated and supplemented by an agreement analysis including the mean absolute error (MAE) and mean absolute percentage error (MAPE) to provide the resulting measurement error. Furthermore, validity analyses between the measured 1RM and different calculation models to estimate 1RM were performed.ResultsReliability values were in accordance with current literature (ICC = 0.79–0.99, coefficient of variance [CV] = 1.86–9.32%), however, were accompanied by a random error (mean absolute error [MAE]: 0.05–0.64 m/s, mean absolute percentage error [MAPE]: 2.7–9.5%) arising from test-retest measurement. Strength estimation via the velocity-profile overestimated the bench pull 1RM (limits of agreement [LOA]: -9.73 – -16.72 kg, MAE: 9.80–17.03 kg, MAPE 16.9–29.7%), while the bench press 1RM was underestimated (LOA: 3.34–6.37 kg, MAE: 3.74–7.84 kg, MAPE: 7.5–13.4%); dependent on used calculation model.DiscussionConsidering the observed measurement error associated with LVP-based methods, it can be posited that their utility as a programming strategy is limited. The lack of accuracy required to discriminate between small but significant changes in performance and error, coupled with the potential risks of under- and overestimating 1RM, can result in insufficient stimulus or increased injury risk, respectively. This further diminishes the practicality of these methods, particularly in elite sports settings.

Long-term load forecasting based on adaptive neural fuzzy inference system using real energy data

BackgroundSeveral software programs have specifically been developed to analyse cardiac computed tomography prior to transcatheter aortic valve implantation (TAVI). However, they are not able to perform a complete analysis independently. We report the performance of a fully automated, deep learning-based algorithm for pre-procedural CT analysis as compared to the current clinical standard.MethodsPatients with symptomatic severe aortic stenosis undergoing TAVI were retrospectively enrolled. The pre-procedural dataset was analysed by both a standard TAVI CT-analysis software and by a fully automated CT analysis platform with a deep learning-based algorithm.ResultsNinety-eight patients were included in the analysis. The mean annulus diameter was 24.4 ± 2.4 mm (conventional = 3mensio, Pie Medical Imaging, 3 M) vs. 24.0 ± 2.4 mm (artificial intelligence = AI), mean absolute error (MAE): 0.64 mm, mean absolute percentage error (MAPE): 2.6%. The mean annulus perimeter was measured at 77.7 ± 7.4 mm (3 M) vs. 76.1 ± 7.5 mm (AI), MAE: 2.26 mm, MAPE: 2.9%. The mean annulus area was calculated at 468.9 ± 92.1 mm2 (3 M) vs. 455.6 ± 91.0 mm2 (AI), MAE: 22.4 mm2, MAPE: 4.8%. The intraclass correlation coefficients (ICCs) of all abovementioned parameter were > 0.95 showing an excellent correlation between the two methods. The distance from the annulus to the left coronary artery depicted to 14.0 ± 3.2 mm (3 M) vs. 12.6 ± 2.8 mm (AI), MAE: 2.1 mm, MAPE: 14.3%. The distance to the right coronary artery was 17.1 ± 2.7 mm (3 M) vs. 16.5 ± 3.2 mm (AI), MAE: 1.7 mm, MAPE: 10.1%. The ICCs of the distances to the coronary ostia showed good correlation between both methods.ConclusionIn this retrospective analysis, a deep learning-based analysis of pre-procedural CT datasets showed good to excellent correlation with conventional assessment for the preprocedural TAVI CT measurements. AI-based fully automated CT analysis could emerge to a valuable alternative to conventional CT assessment in the pre-procedural-planning for TAVI.Graphical Supplementary InformationThe online version contains supplementary material available at 10.1007/s00392-025-02790-6.

This article devises the Artificial Intelligence (AI) methods of designing models of short-term forecasting (in 12 h and 24 h horizons) of electricity production in a selected Small Hydropower Plant (SHP). Renewable Energy Sources (RESs) are difficult to predict due to weather variability. Electricity production by a run-of-river SHP is marked by the variability related to the access to instantaneous flow in the river and weather conditions. In order to develop predictive models of an SHP facility (installed capacity 760 kW), which is located in Southern Poland on the Skawa River, hourly data from nearby meteorological stations and a water gauge station were collected as explanatory variables. Data on the water management of the retention reservoir above the SHP were also included. The variable to be explained was the hourly electricity production, which was obtained from the tested SHP over a period of 3 years and 10 months. Obtaining these data to build models required contact with state institutions and private entrepreneurs of the SHP. Four AI methods were chosen to create predictive models: two types of Artificial Neural Networks (ANNs), Multilayer Perceptron (MLP) and Radial Base Functions (RBFs), and two types of decision trees methods, Random Forest (RF) and Gradient-Boosted Decision Trees (GBDTs). Finally, after applying forecast quality measures of Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R2), the most effective model was indicated. The decision trees method proved to be more accurate than ANN models. The best GBDT models’ errors were MAPE 3.17% and MAE 9.97 kWh (for 12 h horizon), and MAPE 3.41% and MAE 10.96 kWh (for 24 h horizon). MLPs had worse results: MAPE from 5.41% to 5.55% and MAE from 18.02 kWh to 18.40 kWh (for 12 h horizon), and MAPE from 7.30% to 7.50% and MAE from 24.12 kWh to 24.83 kWh (for 24 h horizon). Forecasts using RBF were not made due to the very low quality of training and testing (the correlation coefficient was approximately 0.3).

Evaluating a Multitask Artificial Intelligence Model Compared With Humans for Portion-Size Estimation.

Prostate cancer is distinguished by unique histological alterations in glandular architecture, observable in Whole Slide Images (WSIs). The disease is further characterized by pronounced genomic instability, exemplified by biomarkers such as DNA ploidy. While the Gleason Score effectively assesses the histological changes, the manual evaluation of DNA ploidy is subjective and time-consuming. Addressing the need for an objective diagnostic tool, our study developed a deep learning model aimed at accurately predicting DNA ploidy using publicly available WSI from the Cancer Genome Atlas (TCGA). Utilizing a ResNet-18 convolutional neural network, we trained the model on WSIs from 200 TCGA patients to extract features and predict total mRNA expression. The model was further refined using data from 19 TCGA patients, focusing on ploidy number prediction. Comparative analysis against a random reference model revealed significant improvements of ploidy prediction in Mean Absolute Error (MAE) by 43.52%, Mean Absolute Percentage Error (MAPE) by 43.59%, and Mean Squared Error (MSE) by 67.86%. These enhancements were statistically validated (p-values: MAE - 7.07E-06, MAPE - 7.69E-06, MSE - 3.14E-05) and substantiated by large Cohen's d values (MAE - 2.88, MAPE - 2.86, MSE - 2.42), confirming the model's advanced predictive accuracy and practical utility. This approach, validated further on a separate TCGA hold-out test set, signifies a major leap in prostate cancer diagnostics. By integrating mRNA prediction checkpoints, our model not only refines ploidy assessment but also sets the stage for correlating these predictions with patient clinical outcomes. These developments promise to enhance existing diagnostic methods, offering a more objective and efficient tool for DNA ploidy assessment in prostate cancer prognosis and treatment planning. Citation Format: Josselyn Sofia Vergara Cobos, Francisco Carrillo-Perez, Marija Pizurica, Noemi Andor, Olivier Gevaert. Enhancing prostate cancer diagnosis: A deep learning approach for DNA ploidy prediction from whole slide images [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr LB390.

Reference evapotranspiration (ETo), which is used as the basic data in many studies within the scope of hydrology, meteorology, irrigation and soil sciences, can be estimated by using the evaporation (Epan) measured from the class-A pan evaporimeter. However, this method requires reliable pan coefficients (Kp). Many empirical models are used to estimate Kp coefficients. The reliability of these models varies depending on climatic and environmental conditions. Therefore, they need to be tested in the local conditions where they will be used. In this study, conducted in Kahramanmaraş, which has a semi-arid Mediterranean climate in Turkey during the July-October periods of 2020 and 2021, aimed to determine the usability levels of six Kp models in estimating daily and monthly average ETo. The Kp coefficients estimated by the models were multiplied with the daily Epan values, and the daily average ETo values were estimated on the basis of the model. The daily Epan values were measured using an ultrasonic sensor sensitive to the water surface placed on the class-A pan evaporimeter. The ultrasonic sensor was managed by a programmable logic controller (PLC). To enable the sensor to be managed by PLC, a software was prepared using the CODESYS programming language and uploaded to the PLC. The daily average ETo values determined by the FAO-56 Penman-Monteith equation were accepted as actual values. The ETo values estimated by the Kp models were compared with the actual ETo values using the mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE) and determination coefficient (R2) statistical approaches. The Wahed & Snyder outperformed the other models in estimating daily (MAE = 0.78 mm day-1, MAPE = 14.40%, RMSE = 0.97 mm day-1, R2 = 0.82) and monthly (MAE = 0.32 mm day-1, MAPE = 5.88%, RMSE = 0.32 mm day-1, R2 = 0.99) average ETo. FAO-56 showed the nearest performance to Wahed & Snyder. The Snyder model presented the worst performance in estimating daily (MAE = 2.09 mm day-1, MAPE = 37.53%, RMSE = 2.36 mm day-1, R2 = 0.82) and monthly (MAE = 1.83 mm day-1, MAPE = 31.82%, RMSE = 1.87 mm day-1, R2 = 0.99) average ETo. It has been concluded that none of the six Kp models can be used to estimate the daily ETo in Kahramanmaraş located in the Mediterranean-Southeastern Anatolian transitional zone, and only Wahed & Snyder and FAO-56 can be used to estimate the monthly ETo without calibration.

Accurately measured or estimated reference evapotranspiration (ETo) data are needed to properly manage water resources and prioritise their future uses. ETo can be most accurately measured using lysimeter systems. However, high installation and operating costs, as well as difficult and time-consuming measurement processes limit the use of these systems. Therefore, the approach of estimating ETo by empirical models is more preferred and widely used. However, since those models are well in accordance with the climatic and environmental traits of the region in which they were developed, their reliability must be examined if they are utilised in distinctive regions. This study aims to test the usability of mass transfer-based Dalton, Rohwer, Penman, Romanenko, WMO and Mahringer models in Van Lake microclimate conditions and to calibrate them in compatible with local conditions. Firstly, the original equations of these models were tested using 9 years of daily climate data measured between 2012 and 2020. Then, the models were calibrated using the same data and their modified equations were created. The original and modified equations of the models were also tested with the 2021 and 2022 current climate data. Modified equations have been created using the Microsoft Excel program solver add-on, which is based on linear regression. The daily average ETo values estimated using the six mass transfer-based models were compared with the daily average ETo values calculated using the standard FAO-56 PM equation. The statistical approaches of the mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), Nash-Sutcliffe Efficiency (NSE), and determination coefficient (R2) were used as comparison criterion. The best and worst performing models in the original equations were Mahringer (MAE = 0.70 mm day-1, MAPE = 15.86%, RMSE = 0.87 mm day-1, NSE = 0.81, R2 = 0.94) and Penman (MAE = 1.84 mm day-1, MAPE = 33.68%, RMSE = 2.39 mm day-1, NSE = -0.49, R2 = 0.91), respectively, whereas in the modified equations Dalton (MAE = 0.29 mm day-1, MAPE = 7.51%, RMSE = 0.33 mm day-1, NSE = 0.97, R2 = 0.97) and WMO (MAE = 0.36 mm day-1, MAPE = 8.89%, RMSE = 0.43 mm day-1, NSE = 0.95, R2 = 0.97). The RMSE errors of the daily average ETo values estimated using the modified equations were generally below the acceptable error limit (RMSE < 0.50 mm day-1). It has been concluded that the modified equations of the six mass transfer-based models can be used as alternatives to the FAO-56 PM equation under the Van Lake microclimate conditions (NSE > 0.75), while the original equations-except for those of Mahringer (NSE = 0.81), WMO (NSE = 0.79), and Romanenko (NSE = 0.76)-cannot be used.