Accurate Counting Research Articles

Sensitivity of diagnosis of spontaneous bacterial peritonitis is higher with the automated cell count method

BACKGROUND Spontaneous bacterial peritonitis (SBP) is one of the most important complications of patients with liver cirrhosis entailing high morbidity and mortality. Making an accurate early diagnosis of this infection is key in the outcome of these patients. The current definition of SBP is based on studies performed more than 40 years ago using a manual technique to count the number of polymorphs in ascitic fluid (AF). There is a lack of data comparing the traditional cell count method with a current automated cell counter. Moreover, current international guidelines do not mention the type of cell count method to be employed and around half of the centers still rely on the traditional manual method. AIM To compare the accuracy of polymorph count on AF to diagnose SBP between the traditional manual cell count method and a modern automated cell counter against SBP cases fulfilling gold standard criteria: Positive AF culture and signs/symptoms of peritonitis. METHODS Retrospective analysis including two cohorts: Cross-sectional (cohort 1) and case-control (cohort 2), of patients with decompensated cirrhosis and ascites. Both cell count methods were conducted simultaneously. Positive SBP cases had a pathogenic bacteria isolated on AF and signs/symptoms of peritonitis. RESULTS A total of 137 cases with 5 positive-SBP, and 85 cases with 33 positive-SBP were included in cohort 1 and 2, respectively. Positive-SBP cases had worse liver function in both cohorts. The automated method showed higher sensitivity than the manual cell count: 80% vs 52%, P = 0.02, in cohort 2. Both methods showed very good specificity (> 95%). The best cutoff using the automated cell counter was polymorph ≥ 0.2 cells × 109/L (equivalent to 200 cells/mm3) in AF as it has the higher sensitivity keeping a good specificity. CONCLUSION The automated cell count method should be preferred over the manual method to diagnose SBP because of its higher sensitivity. SBP definition, using the automated method, as polymorph cell count ≥ 0.2 cells × 109/L in AF would need to be considered in patients admitted with decompensated cirrhosis.

Automatically Counting Florida Manatees (Trichechus manatus latirostris) from Drone Images Using Object-Based Image Analysis

Florida manatees (Trichechus manatus latirostris) require frequent and extensive surveys to inform conservation efforts. Crewed aircraft surveys can be costly, dangerous, and logistically complex. Unoccupied aerial systems (UASs) can assist with these issues. While manual review of UAS imagery can be time- and labor-intensive, automated detection of manatees in aerial survey footage can help. We present an object-based image analysis workflow for the automated detection and count of Florida manatees in Google Earth Engine, a free platform for research that allows for scripts and imagery sharing. Training and testing datasets were built from randomly extracted image frames from two stationary, unoccupied aerial system videos over thermal refugia. The workflow captured most manatees (93.98 to 95.62% recall; 4.38 to 6.03% false negative rate), but also counted many objects as manatees incorrectly (4.24 to 14.77% precision; 998.40 to 3,885.54% false positive over the detectable rate). Sun glint, mud plumes, and water close to shore were common causes of false positives. While the automated count was too high, the workflow lays markers over each detection, allowing for quick manual review for more accurate (semi-automated) counts. This study is an early step in automated detection tools for Florida manatees in a cloud-based platform. Future efforts could explore other platforms or may improve this workflow by including new classes for confounding objects.

Accuracy and reliability of accelerometer-based pedometers in step counts during walking, running, and stair climbing in different locations of attachment.

This study aimed to investigate the accuracy and reliability of accelerometer-based pedometers placed on the wrist, waist, and shoe's midsole during walking, running, and stair climbing. Twenty healthy adults were recruited. Steps were recorded by the pedometers and visually assessed from simultaneously recorded video to evaluate the accuracy of each pedometer in different locations of attachment. One week later, steps were recorded again by the pedometers only to evaluate the reliability of each pedometer in different locations of attachment. The wrist-worn pedometer presented significantly greater error scores compared to the midsole-worn pedometer during walking (p < 0.001), running (p = 0.006), and stair climbing (p = 0.003). Additionally, mean absolute precent error and Bland-Altman plots indicated that the pedometer worn in the midsole was most accurate for running and stair climbing, followed by waist-worn and wrist-worn pedometers. Furthermore, the midsole-worn and waist-worn pedometers showed strong reliability during walking and running, but only the midsole-worn pedometer presented acceptable reliability during stair climbing. The pedometer's position impacts the accuracy and reliability of step counts, especially for walking and stair climbing. Embedding the pedometer into the midsole seems an effective approach to improve the accuracy and reliability of step counts.

Detection of a Gossypiboma as a Retrocardiac Mass in a Female Patient with a History of Coronary Artery Bypass Graft

Introduction: "Gossypiboma," also known as "textiloma," refers to retained surgical items left in the body after an operation. These retained items can lead to a range of issues, from asymptomatic cases to life-threatening complications. Diagnostically challenging, gossypibomas can mimic malignant masses, making it essential to consider them in the differential diagnosis of postoperative patients with unexplained symptoms. Imaging is crucial for detection and management. Case Presentation: A 69-year-old female with a history of coronary artery bypass graft (CABG) surgery presented with exertional shortness of breath. Imaging revealed a large mass near the left atrium. Initially suspected to be a tumor, further investigation identified it as retained surgical gauze from her surgery ten years prior. Despite the mass, the patient exhibited only mild symptoms, and given her stable condition, regular follow-ups were recommended. Conclusions: Gossypibomas are rare but serious complications that may surface years after surgery. This case highlights the importance of meticulous surgical practices and careful postoperative monitoring. It underscores the need for accurate gauze counting and effective communication in the operating room to prevent errors that can lead to health risks and potential legal issues.

Towards a More Accurate Count: Identifying State-Level Predictors of the Undercount of Young Hispanic Children in the 2010 and 2020 U.S. Censuses

Towards a More Accurate Count: Identifying State-Level Predictors of the Undercount of Young Hispanic Children in the 2010 and 2020 U.S. Censuses

Forecasting age distribution of life-table death counts via α-transformation

We introduce a compositional power transformation, known as an α-transformation, to model and forecast a time series of life-table death counts, possibly with zero counts observed at older ages. As a generalisation of the isometric log-ratio transformation (i.e. α = 0 ), the α transformation relies on the tuning parameter α, which can be determined in a data-driven manner. Using the Australian age-specific period life-table death counts from 1921 to 2020, the α transformation can produce more accurate short-term point and interval forecasts than the log-ratio transformation. The improved forecast accuracy of life-table death counts is of great importance to demographers and government planners for estimating survival probabilities and life expectancy and actuaries for determining annuity prices and reserves for various initial ages and maturity terms.

Vehicular Traffic Flow Detection and Monitoring for Implementation of Smart Traffic Light: A Case Study for Road Intersection in Limeira, Brazil

This paper proposes the development of a smart traffic light prototype based on vehicular traffic flow measurement in the stretch between two avenues in the city of Limeira, SP, Brazil, focusing on the stretch towards UNICAMP’s School of Technology. To this end, we initially developed a Python code using the OpenCV library in order to detect and count vehicles. With the counting in operation, programming logic was inserted, aiming at preparing traffic light timers based on vehicular traffic. Finally, the traffic lights were added to display video via a code change to show the ongoing color changes, also obtaining a code for identifying vehicles and flow, in addition to the virtual traffic light system itself in the system. Vehicle counting accuracy was 75% for large vehicles, 90% for passenger cars, and 100% for motorcycles. The simulation of a smart traffic light implementation worked satisfactorily according to the post-processing of the video recorded for validation.

Performance Characteristics of a New Generation 148-cm Axial Field-of-View uMI Panorama GS PET/CT System with Extended NEMA NU 2-2018 and EARL Standards.

The uMI Panorama GS PET/CT system is a new long-axial-field-of-view scanner featuring high sensitivity, time-of-flight (TOF) resolution, spatial resolution, and count rate performance. The aim of this study is to assess the PET system on the basis of the National Electrical Manufacturers Association (NEMA) NU 2-2018 and European Association of Nuclear Medicine Research Limited (EARL) standards. Methods: Spatial resolution, count rate performance, sensitivity, accuracy, image quality, TOF resolution, and coregistration accuracy were evaluated following the NEMA NU 2-2018 standard. Additional experiments included energy resolution, 200-cm-long line sources for sensitivity, a 175-cm-long scatter phantom for count rate and TOF resolution, as well as the compliance with the EARL guideline. Moreover, an 18F-FDG PET patient study was reconstructed with various frame durations. Results: The PET system achieved sub-3-mm transaxial and axial spatial resolutions at a 1-cm radial offset. The sensitivities with the 70-cm-long and 200-cm-long line sources were observed to be 176.3 and 90.8 kcps/MBq, respectively, at the center of the field of view. The noise-equivalent count rates (NECRs) of the 70-cm-long and 175-cm-long scatter phantoms were measured to be 3.35 Mcps at 57.57 kBq/mL and 2.24 Mcps at 33.27 kBq/mL, respectively. The TOF resolutions for both phantoms were approximately 189 ps at 5.3 kBq/mL and lower than 200 ps below the NECR peaks. The absolute count rate errors of all 34 acquisitions were less than 3% below the NECR peak for the 70-cm-long scatter phantom. With the standard NEMA image quality phantom experiment, the contrast recovery coefficient varied from 68.17% to 94.20% and the background variabilities were all below 2%. The maximum PET/CT coregistration error was 1.33 mm. Regarding EARL compliance, the gaussian filter of 5-mm full width at half maximum could produce acceptable images. The patient data demonstrate visually satisfactory image quality with short frames (less than 1 min). Conclusion: The uMI Panorama GS exhibits spatial resolution and TOF resolution similar to those of the uMI Panorama system (35-cm axial field of view), despite the extended axial field of view. The 148-cm axial coverage, sub-200-ps TOF resolution, high sensitivity, and count rate performances are expected to yield superior image quality and offer new opportunities for various clinical applications.

LarvaeCountAI: a robust convolutional neural network-based tool for accurately counting the larvae of Culex annulirostris mosquitoes.

Accurate counting of mosquito larval populations is essential for maintaining optimal conditions and population control within rearing facilities, assessing disease transmission risks, and implementing effective vector control measures. While existing methods for counting mosquito larvae have faced challenges such as the impact on larval mortality rate, multiple parameters adjustment and limitations in availability and affordability, recent advancements in artificial intelligence, particularly in AI-driven visual analysis, hold promise for addressing these issues. Here, we introduce LarvaeCountAI, an open-source convolutional neural network (CNN)-based tool designed to automatically count Culex annulirostris mosquito larvae from videos captured in laboratory environments. LarvaeCountAI does not require videos to be recorded using an advanced setup; it can count larvae with high accuracy from videos captured using a simple setup mainly consisting of a camera and commonly used plastic trays. Using the videos enables LarvaeCountAI to capitalise on the continuous movement of larvae, enhancing the likelihood of accurately counting a greater number of larvae. LarvaeCountAI adopts a non-invasive approach, where larvae are simply placed in trays and imaged, minimising any potential impact on larval mortality. This approach addresses the limitations associated with previous methods involving mechanical machines, which often increase the risk of larval mortality as larvae pass through multiple sections for counting purposes. The performance of LarvaeCountAI was tested using 10 video samples. Validation results demonstrated the excellent performance of LarvaeCountAI, with an accuracy ranging from 96.25% to 99.13% across 10 test videos and an average accuracy of 97.88%. LarvaeCountAI represents a remarkable advancement in mosquito surveillance technology, offering a robust and efficient solution for monitoring larval populations. LarvaeCountAI can contribute to developing effective strategies for reducing disease transmission and safeguarding public health by providing timely and accurate data on mosquito larvae abundance.

Optical detection of single and multiple seeding using an innovative shape-recognition algorithm

The ability to accurately and continuously monitor seeding quality allows farmers to improve their seeding practices, maintain appropriate seed spacing, and maximize crop productivity. The precise detection and counting of individual seeds, especially those that pass through and overlap in pairs or multiples, is a major challenge in developing seed monitoring systems. In this research, a novel shape-based algorithm is presented that utilizes seven pairs of 3-mm infrared LEDs to efficiently recognize and count both single seeds and overlapping seeds. The outputs of the receivers were converted to binary signals by comparing them to samples taken when no seed flow was present. Each infrared receiver acted as a pixel, and to recognize and count multiple seeds, their outputs were temporarily stored in a matrix with seven columns. When the seeds passed the sensor, their outputs were analyzed together using an innovative and systematic clustering algorithm. This enables shape-based processing and recognition of passing seeds regardless of type and size, effectively reducing counting errors. The developed system was subjected to testing using seven different seed types on a conveyor belt, while one seed type underwent testing on a precision corn seeder in the laboratory. The results indicated a high level of agreement between the sensor data and the actual seed rates. The average counting accuracy for popcorn, hybrid corn, chickpea, pinto beans, coated tomato, mung beans, and soybean seeds on the conveyor belt platform was 0.99, 0.99, 0.98, 0.96, 0.8, 0.99, and 0.99, respectively. The results showed that the accuracy of the algorithm increases with increasing sphericity and size of the seeds relative to the distance between the adjacent LEDs. Moreover, the counting accuracy exceeded 97% for all rates on the precision corn seeder platform, with an average accuracy of 0.986.

MGFNet: Cross-scene Crowd Counting via Multistage Gated Fusion Network

Existing crowd counting methods are mainly trained and tested in similar scenarios. When the testing and training scenarios of the model are different, the counting accuracy of these methods will sharply decrease, which seriously limits their practical application. To address this problem, we propose a multistage gated fusion network (MGFNet) for cross-scene crowd counting. MGFNet is primarily composed of dynamic gated convolution units (DGCU) and multilevel scale attention blocks (MSAB) modules. Specifically, DGCU uses a dynamic gating path to supplement detailed information to reduce the loss of crowd information and overestimation of background in different scenarios. MSAB calibrates crowd information at different scales and perspectives in different scenes by generating attention maps with discriminative information. In addition, we used a new global local consistency loss to optimize the model to adapt to changes in crowd density and distribution. Extensive experiments on four different types of scene counting benchmarks show that the proposed MGFNet achieves superior cross-scene counting performance.

Automatic Apple Detection and Counting with AD-YOLO and MR-SORT.

In the production management of agriculture, accurate fruit counting plays a vital role in the orchard yield estimation and appropriate production decisions. Although recent tracking-by-detection algorithms have emerged as a promising fruit-counting method, they still cannot completely avoid fruit occlusion and light variations in complex orchard environments, and it is difficult to realize automatic and accurate apple counting. In this paper, a video-based multiple-object tracking method, MR-SORT (Multiple Rematching SORT), is proposed based on the improved YOLOv8 and BoT-SORT. First, we propose the AD-YOLO model, which aims to reduce the number of incorrect detections during object tracking. In the YOLOv8s backbone network, an Omni-dimensional Dynamic Convolution (ODConv) module is used to extract local feature information and enhance the model's ability better; a Global Attention Mechanism (GAM) is introduced to improve the detection ability of a foreground object (apple) in the whole image; a Soft Spatial Pyramid Pooling Layer (SSPPL) is designed to reduce the feature information dispersion and increase the sensory field of the network. Then, the improved BoT-SORT algorithm is proposed by fusing the verification mechanism, SURF feature descriptors, and the Vector of Local Aggregate Descriptors (VLAD) algorithm, which can match apples more accurately in adjacent video frames and reduce the probability of ID switching in the tracking process. The results show that the mAP metrics of the proposed AD-YOLO model are 3.1% higher than those of the YOLOv8 model, reaching 96.4%. The improved tracking algorithm has 297 fewer ID switches, which is 35.6% less than the original algorithm. The multiple-object tracking accuracy of the improved algorithm reached 85.6%, and the average counting error was reduced to 0.07. The coefficient of determination R2 between the ground truth and the predicted value reached 0.98. The above metrics show that our method can give more accurate counting results for apples and even other types of fruit.

Log Volume Measurement and Counting Based on Improved Cascade Mask R-CNN and Deep SORT

Logs require multiple verifications to ensure accurate volume and quantity measurements. Log end detection is a crucial step in measuring log volume and counting logs. Currently, this task primarily relies on the Mask R-CNN instance segmentation model. However, the Feature Pyramid Network (FPN) in Mask R-CNN may compromise accuracy due to feature redundancy during multi-scale fusion, particularly with small objects. Moreover, counting logs in a single image is challenging due to their large size and stacking. To address the above issues, we propose an improved log segmentation model based on Cascade Mask R-CNN. This method uses ResNet for multi-scale feature extraction and integrates a hierarchical Convolutional Block Attention Module (CBAM) to refine feature weights and enhance object emphasis. Then, a Region Proposal Network (RPN) is employed to generate log segmentation proposals. Finally, combined with Deep SORT, the model tracks log ends in video streams and counts the number of logs in the stack. Experiments demonstrate the effectiveness of our method, achieving an average precision (AP) of 82.3, APs of 75.3 for small, APm of 70.9 for medium, and APl of 86.2 for large objects. These results represent improvements of 1.8%, 3.7%, 2.6%, and 1.4% over Mask R-CNN, respectively. The detection rate reached 98.6%, with a counting accuracy of 95%. Compared to manually measured volumes, our method shows a low error rate of 4.07%.

On-Tree Mango Fruit Count Using Live Video- Split Image Dataset to Predict Better Yield at Pre-Harvesting Stage

This study introduces a method for fruit counting in agricultural settings using video capture and the YOLOv7 object detection model. By splitting captured videos into frames and strategically selecting representative frames, the approach aims to accurately estimate fruit counts while minimizing the risk of double counting. YOLOv7, known for its efficiency and accuracy in object detection, is employed to analyze selected frames and detect fruits on trees. Demonstrated the method's effectiveness through its ability to provide farmers with precise yield estimations, optimize resource management, and facilitate early detection of orchard issues such as pest infestations or nutrient deficiencies. This technological integration reduces labor costs and supports sustainable agricultural practices by improving productivity and decision-making capabilities. The scalability of the approach makes it suitable for diverse orchard sizes and types, offering a promising tool for enhancing agricultural efficiency and profitability. The researcher compared YOLOv5n, YOLOv5s, YOLOv7, and YOLOv7-tiny with eight-sided imaging techniques around the tree. The experimental results of YOLOv7 with the eight-sided technique performed best and achieved a count accuracy of 97.7% on a single tree in just 17.112 ms of average inference time. On multiple trees, it is 95.48% in just 17 ms of average inference time, with the help of an eight- sided method on tree images.

A generalized model for accurate wheat spike detection and counting in complex scenarios

Wheat is a crucial crop worldwide, and accurate detection and counting of wheat spikes are vital for yield estimation and breeding. However, these tasks are daunting in complex field environments. To tackle this, we introduce RIA-SpikeNet, a model designed to detect and count wheat spikes in such conditions. First, we introduce an Implicit Decoupling Detection Head to incorporate more implicit knowledge, enabling the model to better distinguish visually similar wheat spikes. Second, Asymmetric Loss is employed as the confidence loss function, enhancing the learning weights of positive and hard samples, thus improving performance in complex scenes. Lastly, the backbone network is modified through reparameterization and the use of larger convolutional kernels, expanding the effective receptive field and improving shape information extraction. These enhancements significantly improve the model’s ability to detect and count wheat spikes accurately. RIA-SpikeNet outperforms the state-of-the-art YOLOv8 detection model, achieving a competitive 81.54% mAP and 90.29% R2. The model demonstrates superior performance in challenging scenarios, providing an effective tool for wheat spike yield estimation in field environments and valuable support for wheat production and breeding efforts.

A Crowded Object Counting System with Self-Attention Mechanism

Traditional object counting systems use object detection methods to count objects. However, when objects are small, crowded, and dense, object detection may fail, leading to inaccuracies in counting. To address this issue, we propose a crowded object counting system based on density map estimation. While most density map estimation models employ encoder–decoder or multi-branch approaches to generate feature maps at different scales for obtaining an accurate density map, improving the accuracy of crowded object counting remains a challenge. In this paper, we propose a novel model that can generate more accurate density maps, utilizing the context-aware network as the primary structure and integrating the self-attention mechanism. There are three main contributions in this paper. Firstly, the self-attention mechanism is employed to improve the accuracy of density map estimation. Secondly, the missing vehicle labels in the TRANCOS database are relabeled, ensuring that the ground truth data are more complete than the original TRANCOS database, thus enabling the proposed novel model to have higher crowded object counting accuracy. Thirdly, the parameters of the self-attention mechanism are analyzed to obtain the optimum parameter combination. The experimental results demonstrate that the accuracy of crowded object counting can reach 85.9%, 90.0%, 83.4%, and 92.6% for the TRANCOS, relabeled TRANCOS, ShanghaiTech Part A, and Part B datasets, respectively. Furthermore, the ablation study for the context-aware network with self-attention mechanism analyzes the optimum parameter combination.

Drone Insights: Unveiling Beach Usage through AI-Powered People Counting

Ocean beaches are a major recreational attraction in many coastal cities, requiring accurate visitor counts for infrastructure planning and value estimation. We developed a novel method to assess beach usage on the Gold Coast, Australia, using 507 drone surveys across 24 beaches. The surveys covered 30 km of coastline, accounting for different seasons, times of day, and environmental conditions. Two AI models were employed: one for counting people on land and in water (91–95% accuracy), and another for identifying usage types (85–92% accuracy). Using drone data, we estimated annual beach usage at 34 million people in 2022/23, with 55% on land and 45% in water—approximately double the most recent estimate from lifeguard counts, which are spatially limited and prone to human error. When applying similar restrictions as lifeguard surveys, drone data estimated 15 million visits, aligning closely with lifeguard counts (within 9%). Temporal (time of day, day of the week, season) and spatial (beach location) factors were the strongest predictors of beach usage, with additional patterns explained by weather variables. Our method, combining drones with AI, enhances the coverage, accuracy, and granularity of beach monitoring, offering a scalable, cost-effective solution for long-term usage assessment.

A Coffee Plant Counting Method Based on Dual-Channel NMS and YOLOv9 Leveraging UAV Multispectral Imaging

Accurate coffee plant counting is a crucial metric for yield estimation and a key component of precision agriculture. While multispectral UAV technology provides more accurate crop growth data, the varying spectral characteristics of coffee plants across different phenological stages complicate automatic plant counting. This study compared the performance of mainstream YOLO models for coffee detection and segmentation, identifying YOLOv9 as the best-performing model, with it achieving high precision in both detection (P = 89.3%, mAP50 = 94.6%) and segmentation performance (P = 88.9%, mAP50 = 94.8%). Furthermore, we studied various spectral combinations from UAV data and found that RGB was most effective during the flowering stage, while RGN (Red, Green, Near-infrared) was more suitable for non-flowering periods. Based on these findings, we proposed an innovative dual-channel non-maximum suppression method (dual-channel NMS), which merges YOLOv9 detection results from both RGB and RGN data, leveraging the strengths of each spectral combination to enhance detection accuracy and achieving a final counting accuracy of 98.4%. This study highlights the importance of integrating UAV multispectral technology with deep learning for coffee detection and offers new insights for the implementation of precision agriculture.

Sorghum segmentation and leaf counting using in silico trained deep neural model

AbstractThis paper introduces a novel deep neural model for segmenting and tracking the number of leaves in sorghum plants in phenotyping facilities. Our algorithm inputs a sequence of images of a sorghum plant and outputs the segmented images and the number of leaves. The key novelty of our approach is in training the deep neural model. Manual annotations are tedious, and we have developed a procedural three‐dimensional (3D) sorghum model that provides detailed geometry and texture to generate photorealistic 3D models. The overall shape of the sorghum leaf geometry is determined by its skeleton, and it is detailed by a procedural model that varies its curvature, width, length, and overall shape. The color is determined by using a Monte Carlo path tracer. We mimic the illumination of the phenotyping facility and use reflectance and transmittance on sorghum surfaces to determine the color of the leaves. The 3D procedural model allows us to generate photorealistic and segmented images that we use to train a deep neural model. Our segmentation provides a mean intersection over union score of 0.51, resulting in leaf counting accuracy within the 90% confidence interval for the human counts.

SCKansformer: Fine-Grained Classification of Bone Marrow Cells via Kansformer Backbone and Hierarchical Attention Mechanisms.

The incidence and mortality rates of malignant tumors, such as acute leukemia, have risen significantly. Clinically, hospitals rely on cytological examination of peripheral blood and bone marrow smears to diagnose malignant tumors, with accurate blood cell counting being crucial. Existing automated methods face challenges such as low feature expression capability, poor interpretability, and redundant feature extraction when processing highdimensional microimage data. We propose a novel finegrained classification model, SCKansformer, for bone marrow blood cells, which addresses these challenges and enhances classification accuracy and efficiency. The model integrates the Kansformer Encoder, SCConv Encoder, and Global-Local Attention Encoder. The Kansformer Encoder replaces the traditional MLP layer with the KAN, improving nonlinear feature representation and interpretability. The SCConv Encoder, with its Spatial and Channel Reconstruction Units, enhances feature representation and reduces redundancy. The Global-Local Attention Encoder combines Multi-head Self-Attention with a Local Part module to capture both global and local features. We validated our model using the Bone Marrow Blood Cell FineGrained Classification Dataset (BMCD-FGCD), comprising over 10,000 samples and nearly 40 classifications, developed with a partner hospital. Comparative experiments on our private dataset, as well as the publicly available PBC and ALL-IDB datasets, demonstrate that SCKansformer outperforms both typical and advanced microcell classification methods across all datasets.