{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,30]],"date-time":"2026-09-30T17:54:12Z","timestamp":1790790852563,"version":"4.1.0"},"reference-count":55,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,7,17]],"date-time":"2026-07-17T00:00:00Z","timestamp":1784246400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>This study presents a Field-Programmable Gate Array (FPGA)-based convolutional neural network (CNN) accelerator for preliminary Parkinson's disease (PD) handwriting classification using hand-drawn circle images, with emphasis on arithmetic-level optimization through efficient multiplier architectures. Although optimized multipliers have been extensively studied for machine learning acceleration, their application-specific effects on inference consistency and hardware efficiency in healthcare-oriented FPGA implementations remain underexplored.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>To address this issue, a lightweight binary CNN classifier, trained on Google Colab, is deployed on FPGA hardware and evaluated with three multiplier architectures: standard multipliers, approximate logarithmic multipliers, and Karatsuba multipliers. The desktop CNN model was evaluated using both non-augmentation validation and standard augmentation strategies. The primary evaluation methodology used a non-augmentation validation approach, in which augmentation was applied exclusively to the training set, resulting in a software validation accuracy of 92.86%. The standard augmentation strategy achieved a validation accuracy of 97.83% and was used to compare the effects of augmentation before splitting. The trained model was quantized to Q4.12 fixed-point precision and implemented on FPGA hardware, where dense-layer computations were performed using different multiplier architectures. Hardware inference was validated on the NewHandPD hand-drawn circle dataset, and FPGA outputs were compared with software inference results via graphical analysis and Mean Absolute Deviation (MAD) to assess numerical consistency.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Experimental results indicate that the FPGA-based implementation achieved classification behavior closely aligned with software inference while improving hardware efficiency. Under the non-augmentation validation approach, the FPGA implementation achieved 89.73% accuracy compared to 92.86% in software, whereas the standard augmentation strategy achieved 95.40% accuracy compared to 97.83% in software using the Approximate Logarithmic Multiplier.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Discussion<\/jats:title>\n                    <jats:p>The results indicate the potential feasibility of lightweight CNN deployment with optimized multipliers for resource-efficient edge healthcare applications. However, due to the limited dataset size, the presented findings should be interpreted as a preliminary proof-of-concept study rather than definitive clinical validation.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/frai.2026.1813130","type":"journal-article","created":{"date-parts":[[2026,7,17]],"date-time":"2026-07-17T05:45:39Z","timestamp":1784267139000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["An efficient multiplier-based FPGA CNN accelerator for Parkinson's disease detection using hand-drawn circle images"],"prefix":"10.3389","volume":"9","author":[{"given":"VedanthSrivatson","family":"A.","sequence":"first","affiliation":[{"name":"School of Electronics Engineering, Vellore Institute of Technology","place":["Vellore, Tamil Nadu, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sivanantham","family":"Sathasivam","sequence":"additional","affiliation":[{"name":"School of Electronics Engineering, Vellore Institute of Technology","place":["Vellore, Tamil Nadu, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Prakash","family":"Ramachandran","sequence":"additional","affiliation":[{"name":"School of Electronics Engineering, Vellore Institute of Technology","place":["Vellore, Tamil Nadu, India"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,7,17]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"151897","DOI":"10.1109\/ACCESS.2021.3126838","article-title":"Power efficient design of high-performance convolutional neural networks hardware accelerator on fpga: a case study with googlenet","volume":"9","author":"Abd El-Maksoud","year":"2021","journal-title":"IEEE Access"},{"key":"B2","doi-asserted-by":"publisher","first-page":"871","DOI":"10.1007\/s41870-023-01601-3","article-title":"Image-based Parkinson's disease detection using deep transfer learning and optimization algorithm","volume":"16","author":"Agrawal","year":"2024","journal-title":"Int. J. Inform. Technol"},{"key":"B3","doi-asserted-by":"publisher","first-page":"126","DOI":"10.1109\/LES.2023.3304678","article-title":"Approximate row-merging-based multipliers for neural network acceleration on fpgas","volume":"16","author":"Aizaz","year":"2023","journal-title":"IEEE Embed. Syst. Lett"},{"key":"B4","doi-asserted-by":"publisher","first-page":"6850","DOI":"10.3934\/math.2024334","article-title":"Modeling and diagnosis Parkinson disease by using hand drawing: deep learning model","volume":"9","author":"Aldhyani","year":"2024","journal-title":"AIMS Math"},{"key":"B5","doi-asserted-by":"publisher","first-page":"2429","DOI":"10.32604\/cmes.2025.063938","article-title":"Hybrid techniques of multi-cnn and ensemble learning to analyze handwritten spiral and wave drawing for diagnosing Parkinson's disease","volume":"143","author":"Al-Jabbar","year":"2025","journal-title":"Comput. Model. Eng. Sci"},{"key":"B6","doi-asserted-by":"publisher","first-page":"267","DOI":"10.11591\/ijeecs.v30.i1.pp267-275","article-title":"Convolutional neural network for the detection of Parkinson's disease based on hand-draw spiral images","volume":"30","author":"Alniemi","year":"2023","journal-title":"Indonesian J. Electr. Eng. Comput. Sci"},{"key":"B7","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3564243","article-title":"Toward optimal softcore carry-aware approximate multipliers on xilinx fpgas","volume":"22","author":"Awais","year":"2023","journal-title":"ACM Trans. Embedd. Comput. Syst"},{"key":"B8","doi-asserted-by":"publisher","first-page":"13422","DOI":"10.1109\/TPAMI.2023.3290544","article-title":"Data-efficient learning via minimizing hyperspherical energy","volume":"45","author":"Cao","year":"2023","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell"},{"key":"B9","doi-asserted-by":"publisher","first-page":"2794","DOI":"10.1109\/TC.2015.2498606","article-title":"Optimised multiplication architectures for accelerating fully homomorphic encryption","volume":"65","author":"Cao","year":"2015","journal-title":"IEEE Trans. Comput"},{"key":"B10","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3773075","article-title":"Analytical survey of learning with low-resource data: from analysis to investigation","volume":"58","author":"Cao","year":"2025","journal-title":"ACM Comput. Surve"},{"key":"B11","doi-asserted-by":"publisher","first-page":"3662","DOI":"10.1002\/cta.4289","article-title":"Fpga-based implementation of real-time cardiologist-level arrhythmia detection and classification in electrocardiograms using novel deep learning","volume":"53","author":"Chandrasekaran","year":"2025","journal-title":"Int. J. Circuit Theory Applic"},{"key":"B12","doi-asserted-by":"publisher","first-page":"104451","DOI":"10.1016\/j.micpro.2022.104451","article-title":"Coha-ntt: a configurable hardware accelerator for ntt-based polynomial multiplication","volume":"89","author":"Derya","year":"2022","journal-title":"Microprocess. Microsyst"},{"key":"B13","doi-asserted-by":"publisher","first-page":"155011","DOI":"10.1016\/j.aeue.2023.155011","article-title":"An optimized fpga architecture for machine learning applications","volume":"174","author":"Elsaid","year":"2024","journal-title":"AEU-Int. J. Electr. Communic"},{"key":"B14","doi-asserted-by":"publisher","first-page":"1453743","DOI":"10.3389\/fmed.2024.1453743","article-title":"Utilizing deep learning models in an intelligent spiral drawing classification system for Parkinson's disease classification","volume":"11","author":"Farhah","year":"2024","journal-title":"Front. Med"},{"key":"B15","doi-asserted-by":"publisher","first-page":"76","DOI":"10.1109\/OJCAS.2020.3007334","article-title":"Power efficient tiny yolo cnn using reduced hardware resources based on booth multiplier and wallace tree adders","volume":"1","author":"Farrukh","year":"2020","journal-title":"IEEE Open J. Circuits Syst"},{"key":"B16","doi-asserted-by":"publisher","first-page":"211235","DOI":"10.1109\/ACCESS.2025.3642938","article-title":"High-efficiency fpga-based cnn accelerator with optimized data handling for convolution and fully connected layers","volume":"13","author":"Foroumandi","year":"2025","journal-title":"IEEE Access"},{"key":"B17","doi-asserted-by":"publisher","first-page":"215","DOI":"10.23919\/MIXDES.2019.8787132","article-title":"\u201cFlexible hls-based implementation of the Karatsuba multiplier targeting homomorphic encryption schemes,\u201d","author":"Foster","year":"2019","journal-title":"2019 MIXDES-26th international conference mixed design of integrated circuits and systems"},{"key":"B18","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3233300","article-title":"Low complexity multiply-accumulate units for convolutional neural networks with weight-sharing","volume":"15","author":"Garland","year":"2018","journal-title":"ACM Trans. Archit. Code Optimiz"},{"key":"B19","doi-asserted-by":"publisher","first-page":"3317","DOI":"10.3390\/electronics14163317","article-title":"Hardware accelerator design by using rt-level power optimization techniques on fpga for future ai mobile applications","volume":"14","author":"Gundrapally","year":"2025","journal-title":"Electronics"},{"key":"B20","doi-asserted-by":"publisher","first-page":"5919","DOI":"10.1109\/TCSI.2024.3416204","article-title":"Hardware-efficient multipliers with fpga-based approximation for error-resilient applications","volume":"71","author":"Guo","year":"2024","journal-title":"IEEE Trans. Circuits Syst. I: Regular Papers"},{"key":"B21","doi-asserted-by":"publisher","first-page":"118899","DOI":"10.1109\/ACCESS.2020.3005286","article-title":"Align: a highly accurate adaptive layerwise log_2_lead quantization of pre-trained neural networks","volume":"8","author":"Gupta","year":"2020","journal-title":"IEEE Access"},{"key":"B22","doi-asserted-by":"publisher","first-page":"5251","DOI":"10.1109\/JSYST.2023.3308333","article-title":"Metaparkinson: a cyber-physical deep meta-learning framework for n-shot diagnosis and monitoring of Parkinson's patients","volume":"17","author":"Hossain","year":"2023","journal-title":"IEEE Syst. J"},{"key":"B23","doi-asserted-by":"publisher","first-page":"5433740","DOI":"10.1049\/cds2\/5433740","article-title":"A high-efficiency cnn accelerator with mixed low-precision quantization","volume":"2025","author":"Hu","year":"2025","journal-title":"IET Circuits Dev. Syst"},{"key":"B24","doi-asserted-by":"publisher","first-page":"4069","DOI":"10.1109\/TNNLS.2021.3055814","article-title":"Fpga-based high-throughput cnn hardware accelerator with high computing resource utilization ratio","volume":"33","author":"Huang","year":"2021","journal-title":"IEEE Trans. Neural Netw. Learn. Syst"},{"key":"B25","doi-asserted-by":"publisher","first-page":"220","DOI":"10.3390\/info15040220","article-title":"Early Parkinson's disease diagnosis through hand-drawn spiral and wave analysis using deep learning techniques","volume":"15","author":"Huang","year":"2024","journal-title":"Information"},{"key":"B26","doi-asserted-by":"publisher","first-page":"2615","DOI":"10.3390\/diagnostics14232615","article-title":"Early detection of Parkinson's disease using ai techniques and image analysis","volume":"14","author":"Ianculescu","year":"2024","journal-title":"Diagnostics"},{"key":"B27","doi-asserted-by":"publisher","first-page":"455","DOI":"10.1109\/TVLSI.2024.3354726","article-title":"Area-efficient iterative logarithmic approximate multipliers for ieee 754 and posit numbers","volume":"32","author":"Kim","year":"2024","journal-title":"IEEE Trans. Very Large Scale Integr. Syst"},{"key":"B28","doi-asserted-by":"publisher","first-page":"59438","DOI":"10.1109\/ACCESS.2023.3285279","article-title":"A reconfigurable cnn-based accelerator design for fast and energy-efficient object detection system on mobile fpga","volume":"11","author":"Kim","year":"2023","journal-title":"IEEE Access"},{"key":"B29","doi-asserted-by":"publisher","first-page":"105455","DOI":"10.1109\/ACCESS.2020.3000009","article-title":"A novel fpga accelerator design for real-time and ultra-low power deep convolutional neural networks compared with titan x gpu","volume":"8","author":"Li","year":"2020","journal-title":"IEEE Access"},{"key":"B30","doi-asserted-by":"publisher","first-page":"8383","DOI":"10.1007\/s00521-024-09516-1","article-title":"Quantum mayfly optimization based feature subset selection with hybrid cnn for biomedical Parkinson's disease diagnosis","volume":"36","author":"Mansour","year":"2024","journal-title":"Neural Comput. Applic"},{"key":"B31","doi-asserted-by":"publisher","first-page":"104998","DOI":"10.1016\/j.micpro.2023.104998","article-title":"A real-time svm-based hardware accelerator for hyperspectral images classification in fpga","volume":"104","author":"Martins","year":"2024","journal-title":"Microprocess. Microsyst"},{"key":"B32","doi-asserted-by":"publisher","first-page":"102654","DOI":"10.1016\/j.vlsi.2026.102654","article-title":"A resource-constrained cnn accelerator for real-time license plate character recognition on fpga platforms","volume":"108","author":"Nardes","year":"2026","journal-title":"Integration"},{"key":"B33","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/WorldSUAS66815.2025.11199127","article-title":"\u201cEarly Parkinson's disease diagnosis using hand-drawn image patterns using lightweight and deep learning models with explainable ai,\u201d","volume-title":"2025 World skills conference on universal data analytics and sciences (WorldSUAS)","author":"Nithya","year":"2025"},{"key":"B34","doi-asserted-by":"publisher","first-page":"107198","DOI":"10.1016\/j.bspc.2024.107198","article-title":"Elasticnet-based vision transformers for early detection of Parkinson's disease","volume":"101","author":"\u00d6zdemir","year":"2025","journal-title":"Biomed. Signal Process. Control"},{"key":"B35","doi-asserted-by":"publisher","first-page":"32759","DOI":"10.1109\/ACCESS.2023.3263392","article-title":"Fpg-ai: a technology-independent framework for the automation of cnn deployment on fpgas","volume":"11","author":"Pacini","year":"2023","journal-title":"IEEE Access"},{"key":"B36","doi-asserted-by":"publisher","first-page":"1377","DOI":"10.1109\/TC.2025.3525606","article-title":"Karatsuba matrix multiplication and its efficient custom hardware implementations","volume":"74","author":"Pogue","year":"2025","journal-title":"IEEE Trans. Comput"},{"key":"B37","doi-asserted-by":"publisher","first-page":"106220","DOI":"10.1016\/j.bspc.2024.106220","article-title":"Cosine deep convolutional neural network for Parkinson's disease detection and severity level classification using hand drawing spiral image in iot platform","volume":"94","author":"Pragadeeswaran","year":"2024","journal-title":"Biomed. Signal Process. Control"},{"key":"B38","doi-asserted-by":"publisher","first-page":"7120","DOI":"10.1109\/TED.2022.3214167","article-title":"A nonvolatile all-spin nonbinary matrix multiplier: an efficient hardware accelerator for machine learning","volume":"69","author":"Rahman","year":"2022","journal-title":"IEEE Trans. Electron Devices"},{"key":"B39","doi-asserted-by":"publisher","first-page":"102508","DOI":"10.1016\/j.vlsi.2025.102508","article-title":"Post-training quantization for efficient fpga-based neural network acceleration","volume":"105","author":"Salah","year":"2025","journal-title":"Integration"},{"key":"B40","doi-asserted-by":"publisher","first-page":"4735","DOI":"10.1109\/TCAD.2023.3296375","article-title":"Cryptensor: a resource-shared co-processor to accelerate convolutional neural network and polynomial convolution","volume":"42","author":"See","year":"2023","journal-title":"IEEE Trans. Comput"},{"key":"B41","doi-asserted-by":"publisher","first-page":"206","DOI":"10.1007\/s11596-025-00017-3","article-title":"Deep learning-based diagnostic model for Parkinson's disease using handwritten spiral and wave images","volume":"45","author":"Shastry","year":"2025","journal-title":"Curr. Med. Sci"},{"key":"B42","doi-asserted-by":"publisher","first-page":"2137","DOI":"10.1109\/TCSI.2019.2892588","article-title":"Design methodology to explore hybrid approximate adders for energy-efficient image and video processing accelerators","volume":"66","author":"Soares","year":"2019","journal-title":"IEEE Trans. Circuits Syst. I: Regular Papers"},{"key":"B43","doi-asserted-by":"publisher","first-page":"211","DOI":"10.1109\/TCAD.2021.3056337","article-title":"High-performance accurate and approximate multipliers for fpga-based hardware accelerators","volume":"41","author":"Ullah","year":"2021","journal-title":"IEEE Trans. Comput.-Aided Design Integr. Circuits Syst"},{"key":"B44","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3663480","article-title":"Dyrecmul: fast and low-cost approximate multiplier for fpgas using dynamic reconfiguration","volume":"18","author":"Vakili","year":"2024","journal-title":"ACM Trans. Reconfigurable Technol. Syst"},{"key":"B45","doi-asserted-by":"publisher","first-page":"4088","DOI":"10.1109\/TCAD.2022.3197489","article-title":"Via: a novel vision-transformer accelerator based on fpga","volume":"41","author":"Wang","year":"2022","journal-title":"IEEE Trans. Comput.-Aided Design Integr. Circuits Syst"},{"key":"B46","doi-asserted-by":"publisher","first-page":"1566","DOI":"10.1109\/TCSII.2021.3065333","article-title":"Axbms: approximate radix-8 booth multipliers for high-performance fpga-based accelerators","volume":"68","author":"Waris","year":"2021","journal-title":"IEEE Trans. Circuits Syst. II: Express Briefs"},{"key":"B47","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3474597","article-title":"Low-precision floating-point arithmetic for high-performance fpga-based cnn acceleration","volume":"15","author":"Wu","year":"2021","journal-title":"ACM Trans. Reconfig. Technol. Syst"},{"key":"B48","doi-asserted-by":"publisher","first-page":"163","DOI":"10.1016\/j.neucom.2021.11.032","article-title":"Resource efficient activation functions for neural network accelerators","volume":"482","author":"Wuraola","year":"2022","journal-title":"Neurocomputing"},{"key":"B49","doi-asserted-by":"publisher","first-page":"101991","DOI":"10.1016\/j.sysarc.2021.101991","article-title":"Sparknoc: an energy-efficiency fpga-based accelerator using optimized lightweight cnn for edge computing","volume":"115","author":"Xia","year":"2021","journal-title":"J. Syst. Architect"},{"key":"B50","doi-asserted-by":"publisher","first-page":"158","DOI":"10.1109\/LES.2023.3296507","article-title":"An efficient hardware accelerator for block sparse convolutional neural networks on fpga","volume":"16","author":"Yin","year":"2023","journal-title":"IEEE Embed. Syst. Lett"},{"key":"B51","doi-asserted-by":"publisher","first-page":"106570","DOI":"10.1016\/j.mejo.2025.106570","article-title":"Design of lightweight on-chip one-dimensional convolutional neural network accelerators for edge-end chips","volume":"157","author":"Yu","year":"2025","journal-title":"Microelectronics J"},{"key":"B52","doi-asserted-by":"publisher","first-page":"101153","DOI":"10.1016\/j.jestch.2022.101153","article-title":"Minimum signed digit approximation for faster and more efficient convolutional neural network computation on embedded devices","volume":"36","author":"Zaman","year":"2022","journal-title":"Eng. Sci. Technol"},{"key":"B53","doi-asserted-by":"publisher","first-page":"9921809","DOI":"10.1155\/2023\/9921809","article-title":"A spatio-temporal siamese neural network for multimodal handwriting abnormality screening of Parkinson's disease","volume":"2023","author":"Zhao","year":"2023","journal-title":"Int. J. Intell. Syst"},{"key":"B54","doi-asserted-by":"publisher","first-page":"115830","DOI":"10.1016\/j.knosys.2026.115830","article-title":"Haq-vit: a hardware-aware post-training quantization for efficient vision transformer inference","volume":"342","author":"Zhao","year":"2026","journal-title":"Knowl"},{"key":"B55","doi-asserted-by":"publisher","first-page":"1485","DOI":"10.1109\/TVLSI.2024.3409649","article-title":"A design framework for generating energy-efficient accelerator on fpga toward low-level vision","volume":"32","author":"Zhou","year":"2024","journal-title":"IEEE Trans. Very Large Scale Integr. Syst"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1813130\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,17]],"date-time":"2026-07-17T05:45:39Z","timestamp":1784267139000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1813130\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,17]]},"references-count":55,"alternative-id":["10.3389\/frai.2026.1813130"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1813130","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,17]]},"article-number":"1813130"}}