A portable pneumonia symptom identification auxiliary analysis system integration device

TWI934680BActive Publication Date: 2026-08-01ENERGY RESOURCES INT
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
TW · TW
Patent Type
Patents
Current Assignee / Owner
ENERGY RESOURCES INT
Filing Date
2025-06-26
Publication Date
2026-08-01

Smart Images

  • Figure TWG2TB001904025_001
    Figure TWG2TB001904025_001
  • Figure TWG2TB001904025_002
    Figure TWG2TB001904025_002
  • Figure TWG2TB001904025_003
    Figure TWG2TB001904025_003
Patent Text Reader

Abstract

This invention provides a portable pneumonia symptom identification and analysis system integrated device, designed for rural areas and home healthcare needs, featuring high mobility and real-time diagnostic capabilities. The portable pneumonia symptom identification and analysis system integrated device includes a portable X-ray machine, a box-type X-ray machine carrier, a retractable folding support frame, a digital X-ray image sensor, and a computer device with a built-in pneumonia symptom identification and analysis algorithm. It is battery-powered, suitable for environments without electricity or with unstable power. This system features a multimodal deep learning lung symptom identification algorithm built into the computer module. This algorithm combines a densely connected convolutional neural network (DenseNet121) and a sliding window transformer (Swin). The system employs a Transformer architecture and fuses local and global image features through a cross-attention module to accurately classify chest X-ray images as pneumonia, COVID-19, tuberculosis, or normal. Furthermore, it incorporates data augmentation techniques (such as CutMix, MixUp, and Generative Adversarial Networks (GANs)), loss functions to address class imbalance (such as Focal Loss and Distributed Label Awareness (LDAM) Loss), and the Area Under the Receiver Operating Characteristic (ROC-AUC) curve, along with a threshold adjustment strategy. This enhances the model's stability and generalization ability in identifying rare lesions in real-world environments. The final output includes symptom classification results and Gradient Weighted Class Priming Map (Grad-CAM) visualization of hotspots, which can be embedded in devices, cloud-based diagnostic platforms, or mobile healthcare systems for real-time assisted interpretation or remote clinical decision support.
Need to check novelty before this filing date? Find Prior Art

Claims

1. A portable pneumonia symptom identification and auxiliary analysis system integrated device, characterized in that it comprises: A portable X-ray machine; a box-type X-ray machine carrier; a retractable folding support frame; wherein the portable X-ray machine is placed on the box-type X-ray machine carrier; the box-type X-ray machine carrier is placed on the retractable folding support frame, so that the retractable folding support frame can support the box-type X-ray machine carrier and the portable X-ray machine; a digital X-ray image sensor; placed below the retractable folding support frame, facing the portable X-ray machine; receiving X-rays emitted by the portable X-ray machine, and forming X-ray image data on the digital X-ray image sensor; a computer device with a built-in pneumonia symptom identification auxiliary algorithm; near the digital X-ray image sensor, receiving the X-ray image data transmitted from the digital X-ray image sensor, and performing pneumonia symptom identification, thereby completing the integrated device of a portable pneumonia symptom identification auxiliary analysis system.

2. The portable pneumonia symptom identification and auxiliary analysis system integrated device as described in claim 1, wherein the box-type X-ray machine carrier is characterized in that it comprises: A first box fastener; a second box fastener; wherein the first box fastener and the second box fastener can quickly fix the portable X-ray machine to the two arms of the foldable support frame, or disassemble it; a locking screw can fix the portable X-ray machine to the box fastener X-ray machine carrier to improve operating efficiency and safety.

3. The portable pneumonia symptom identification and analysis system integration device as described in claim 1, wherein the portable X-ray machine, digital X-ray image sensor and computer device with built-in pneumonia symptom identification and analysis algorithm are characterized in that they all have built-in batteries; enabling the portable pneumonia symptom identification and analysis system integration device to perform X-ray image capture and analysis in areas without mains power.

4. The portable pneumonia symptom identification and analysis system integrated device as described in claim 1, wherein the digital X-ray image sensor is characterized in that it adopts a flat-panel CMOS structure, is capable of receiving the image signal formed after X-rays emitted by a portable X-ray machine penetrate the patient, and converts the image into digital data format in real time, and outputs it to a computer device with a built-in pneumonia symptom identification and analysis algorithm for further analysis and classification.

5. A portable pneumonia symptom identification and analysis system integration device as described in claim 1, wherein the computer device with a built-in pneumonia symptom identification and analysis algorithm is characterized in that: the following modules are executed sequentially: image preprocessing module; multimodal feature extraction module; feature fusion module; symptom classification output; and diagnostic output and visualization module; wherein, The image preprocessing module standardizes, enhances contrast, and extracts lung regions (ROIs) from the received chest X-ray images to obtain input images with diagnostic features. The multimodal feature extraction and feature fusion module extracts local and global features in parallel. The local features are obtained through a multilayer densely connected convolutional neural network (DenseNet121), while the global features are obtained through a sliding window transformer (Swin Transformer). Subsequently, the feature fusion module fuses the aforementioned local and global features in a cross-attention block and generates a symptom vector from the symptom classification output. Finally, the diagnostic output and visualization module generates a classification result of pneumonia symptoms based on the symptom vector and uses the gradient-weighted class priming mapping (Grad-CAM) method to generate a visual image of the lesion hot zone, thus completing the overall auxiliary diagnostic process of the portable pneumonia symptom identification and analysis system.

6. A portable pneumonia symptom identification and analysis system integration device as described in claim 5, wherein the diagnostic output and visualization module is characterized in that it can output symptom classification results and generate at least one lesion hot zone visualization image using gradient weighted category initiation mapping (Grad-CAM), superimpose the hot zone result on the original chest X-ray image, and provide image superposition information for physician reference on the display interface.