Browsing by Author "Komgrit Jaksukam"

Now showing 1 - 2 of 2
  • Default Image
    Item
    Development of Self-Supervised Learning with Dinov2-Distilled Models for Parasite Classification in Screening
    (Institute of Electrical and Electronics Engineers Inc., 2023) Natchapon Pinetsuksai; Veerayuth Kittichai; Rangsan Jomtarak; Komgrit Jaksukam; Teerawat Tongloy; Siridech Boonsang; Santhad Chuwongin
    At present, parasitic infections in humans, such as intestinal parasitic infections and soil-transmitted helminth (STH) infection, remain a public health concern, with screening methods that are simple but time-consuming and require parasitology experts. Microscopy images are increasingly being used to aid diagnosis but creating labels for supervised learning (SL) is a time-consuming, labor-intensive, and costly process. Self-supervised learning (SSL) is a deep learning approach that aims to train models to represent features in unlabeled datasets using automatically generated labels or annotations from the data itself, rather than explicitly labeled human-labeled labels. It is an appropriate method to address the challenges associated with the difficulty of labeling large datasets. A pretrained model that has learned useful data representations from an SSL task is fine-tuned using labeled data to perform well on a specific downstream task. DINOv2 is an SSL model based on the Vision Transformer (ViT) architecture. In this study, we aim to create a model for screening for helminth egg infection using a fine-tuned Dinov2 with a classification layer head to demonstrate that dataset sizes of 1% and 10% are sufficient when compared to SL model. Rather than SL, which requires a significant amount of human data labeling and is generally impractical, the model developed in this study is expected to be used in active surveillance in the future. © 2023 IEEE.
  • Default Image
    Item
    Superior Automatic Screening for Human Helminthic Ova by Using Self-supervised Learning Approach-Based Object Classification
    (Springer Science and Business Media Deutschland GmbH, 2023) Natchapon Pinetsuksai; Veerayuth Kittichai; Rangsan Jomtarak; Komgrit Jaksukam; Teerawat Tongloy; Siridech Boonsang; Santhad Chuwongin; S. Chuwongin; College of Advanced Manufacturing Innovation, King MongkutÕs Institute of Technology Ladkrabang, Bangkok, Thailand; email: santhad.ch@kmitl.ac.th
    Human parasitic infections remain one of public health concerns for 1.5 billion people worldwide including Thailand. Conventional microscopic examination is a gold standard method and often used to identify the helminth ova and filariform larvae and also protozoa cyst in stool-dependent simple smear. The benefits of traditional techniques are diminished by time-consuming, complicated procedures, massive labor, and skilled and trained parasitologists. An automatically rapid screening of the most in need of treatment is considered to replace the conventional technique. Here, we aim to develop a deep convolutional residual network based self-supervised learning model to identify mostly common parasite ova in Thailand. Although small amounts of training data was used to train the proposed model, the result shows superior performance over 95% accuracy. As a result, low values of false positive and false negative based confusion matrix table found revealed the robustness of the proposed models. General accuracy of self-supervised learning based the area under a ROC curve proposed with greater than 94% is also support an outstanding model studied. Therefore, rank of 1% to 10% of fine-tuning data labelled used bring us about a comparable model to that of using a 100% labelled training data. These findings emphasize the transformative potential of the BYOL method for screening of parasitic infection, particularly in resource-limited settings where is a lack of supportive lab equipment and skilled parasitologists to manage a large amount of challenging data in the future. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.