MSGID: 1:317/3 6451e36f   
   PID: hpt/lnx 1.9.0-cur 2019-01-08   
   TID: hpt/lnx 1.9.0-cur 2019-01-08   
    Deep neural network provides robust detection of disease biomarkers in   
   real time    
      
     Date:   
         May 2, 2023   
     Source:   
         University of California - Santa Cruz   
     Summary:   
         A lab has developed a deep neural network that improves the accuracy   
         of their unique devices for detecting pathogen biomarkers.   
      
      
         Facebook Twitter Pinterest LinkedIN Email   
      
   ==========================================================================   
   FULL STORY   
   ==========================================================================   
   Sophisticated systems for the detection of biomarkers -- molecules such   
   as DNA or proteins that indicate the presence of a disease -- are crucial   
   for real- time diagnostic and disease-monitoring devices.   
      
   Holger Schmidt, distinguished professor of electrical and computer   
   engineering at UC Santa Cruz, and his group have long been focused on   
   developing unique, highly sensitive devices called optofluidic chips to   
   detect biomarkers.   
      
   Schmidt's graduate student Vahid Ganjalizadeh led an effort to use machine   
   learning to enhance their systems by improving its ability to accurately   
   classify biomarkers. The deep neural network he developed classifies   
   particle signals with 99.8 percent accuracy in real time, on a system   
   that is relatively cheap and portable for point-of-care applications,   
   as shown in a new paper in Nature Scientific Reports.   
      
   When taking biomarker detectors into the field or a point-of-care setting   
   such as a health clinic, the signals received by the sensors may not be   
   as high quality as those in a lab or a controlled environment. This may   
   be due to a variety of factors, such as the need to use cheaper chips to   
   bring down costs, or environmental characteristics such as temperature   
   and humidity.   
      
   To address the challenges of a weak signal, Schmidt and his team developed   
   a deep neural network that can identify the source of that weak signal   
   with high confidence. The researchers trained the neural network with   
   known training signals, teaching it to recognize potential variations   
   it could see, so that it can recognize patterns and identify new signals   
   with very high accuracy.   
      
   First, a parallel cluster wavelet analysis (PCWA) approach designed   
   in Schmidt's lab detects that a signal is present. Then, the neural   
   network processes the potentially weak or noisy signal, identifying its   
   source. This system works in real time, so users are able to receive   
   results in a fraction of a second.   
      
   "It's all about making the most of possibly low quality signals, and   
   doing that really fast and efficiently," Schmidt said.   
      
   A smaller version of the neural network model can run on portable   
   devices. In the paper, the researchers run the system over a Google Coral   
   Dev board, a relatively cheap edge device for accelerated execution of   
   artificial intelligence algorithms. This means the system also requires   
   less power to execute the processing compared to other techniques.   
      
   "Unlike some research that requires running on supercomputers to do high-   
   accuracy detection, we proved that even a compact, portable, relatively   
   cheap device can do the job for us," Ganjalizadeh said. "It makes it   
   available, feasible, and portable for point-of-care applications."   
   The entire system is designed to be used completely locally, meaning   
   the data processing can happen without internet access, unlike other   
   systems that rely on cloud computing. This also provides a data security   
   advantage, because results can be produced without the need to share   
   data with a cloud server provider.   
      
   It is also designed to be able to give results on a mobile device,   
   eliminating the need to bring a laptop into the field.   
      
   "You can build a more robust system that you could take out to   
   under-resourced or less- developed regions, and it still works,"   
   Schmidt said.   
      
   This improved system will work for any other biomarkers Schmidt's lab's   
   systems have been used to detect in the past, such as COVID-19, Ebola,   
   flu, and cancer biomarkers. Although they are currently focused on   
   medical applications, the system could potentially be adapted for the   
   detection of any type of signal.   
      
   To push the technology further, Schmidt and his lab members plan to add   
   even more dynamic signal processing capabilities to their devices. This   
   will simplify the system and combine the processing techniques needed to   
   detect signals at both low and high concentrations of molecules. The team   
   is also working to bring discrete parts of the setup into the integrated   
   design of the optofluidic chip.   
      
       * RELATED_TOPICS   
             o Health_&_Medicine   
                   # Medical_Devices # Diseases_and_Conditions #   
                   Immune_System   
             o Plants_&_Animals   
                   # Developmental_Biology # Extreme_Survival #   
                   Epigenetics_Research   
             o Matter_&_Energy   
                   # Telecommunications # Detectors # Microarrays   
             o Computers_&_Math   
                   # Computers_and_Internet # Neural_Interfaces #   
                   Mobile_Computing   
       * RELATED_TERMS   
             o Neural_network o Artificial_neural_network o Retina   
             o Computer_worm o Robot_calibration o Mobile_phone o   
             Local_area_network o Biophysics   
      
   ==========================================================================   
   Story Source: Materials provided by   
   University_of_California_-_Santa_Cruz. Original written by Emily   
   Cerf. Note: Content may be edited for style and length.   
      
      
   ==========================================================================   
   Journal Reference:   
      1. Vahid Ganjalizadeh, Gopikrishnan G. Meena, Matthew A. Stott,   
      Aaron R.   
      
         Hawkins, Holger Schmidt. Machine learning at the edge for AI-enabled   
         multiplexed pathogen detection. Scientific Reports, 2023; 13 (1)   
         DOI: 10.1038/s41598-023-31694-6   
   ==========================================================================   
      
   Link to news story:   
   https://www.sciencedaily.com/releases/2023/05/230502155410.htm   
      
   --- up 1 year, 9 weeks, 1 day, 10 hours, 50 minutes   
    * Origin: -=> Castle Rock BBS <=- Now Husky HPT Powered! (1:317/3)   
   SEEN-BY: 15/0 106/201 114/705 123/120 153/7715 218/700 226/30 227/114   
   SEEN-BY: 229/110 112 113 307 317 400 426 428 470 664 700 292/854 298/25   
   SEEN-BY: 305/3 317/3 320/219 396/45   
   PATH: 317/3 229/426