School of Mechanical Engineering Seminar
Wednesday, July 10, 2024 at 14:00
Wolfson Building of Mechanical Engineering, Room 206

Classifying High Fluoride  Concentrations in Water Based on Data From India's Central Ground Water Board.

 Arseni Riabkov

This work was carried out under the supervision of

Dr. Hadas Maman
&
Mr. Asaf Pras

Fluoride is a naturally occurring element that is found in soil, water, and air. While small concentrations of fluoride are beneficial for dental health, excessive exposure to fluoride in the form of ingestion or inhalation can lead to a variety of health problems, including skeletal and dental fluorosis. Click or tap here to enter text. It is estimated that the global total for people affected by Dental fluorosis alone may exceed 70 million. The World Health Organization (WHO) noted that dental fluorosis is associated with fluoride levels in drinking water above 1.5 mg/L  and recommends a fluoride concentration of no more than 1.5 mg/L in drinking water as a level at which dental fluorosis should be minimal. Even so, it is important to note that The 1.5 mg guideline value of WHO is not a “fixed” value but is intended to be adapted according to local conditions.

Traditional methods for monitoring fluoride levels in groundwater typically involve various techniques. These range from electrochemical approaches to colorimetric methodologies, which can include naked-eye detection or spectrophotometric measurements. Some of these analytical techniques require manual sample collection and laboratory analysis, which can be time-consuming, and costly. And on top of that may require specialized equipment and expertise. In contrast, machine learning techniques can leverage available data to develop predictive models to estimate groundwater fluoride levels.

In recent years, there has been an increasing interest in using machine learning (ML) and artificial intelligence (AI) techniques to predict fluoride levels in groundwater. These methods have the potential to provide valuable information about fluoride concentrations in areas where data is limited or difficult to obtain. One of the main benefits of using ML and AI techniques for this purpose is their ability to analyze large amounts of data and identify patterns that may not be immediately obvious to human analysts. In particular, ML algorithms such as neural networks and decision trees are effective at identifying complex relationships between different variables.

The ML models in this study were trained Primarily on The India Central Ground Water Board data set, which covers the years 2000 to 2018. And contains more than 150,000 rows of information from a total of about 18,000 groundwater wells which include information on fluoride concentration, PH, Electric Conductivity, Nitrate, Bicarbonate, and Calcium. Additionally, Calcicol concentrations in the ground and precipitation data were added. By training the ML models on this data set, we developed 3 different models that could predict high fluoride levels in groundwater with similar performance. A Random Forest Model, an ADA-boosted Decision tree, and a  Multi-layer perceptron model which had the best performance with an accuracy score of 0.78 and a recall score of 0.76.

Using these ML models, it is possible to identify where the fluoride concentration exceeds the WHO-recommended levels and take necessary actions to mitigate the effects of fluoride on human health. Additionally, these models can help to identify potential sources of fluoride in the groundwater and assist in the development of strategies to reduce fluoride levels. Overall, the use of ML and AI techniques for predicting fluoride levels in groundwater can provide valuable information to help protect public health and support sustainable water management.

LMI Seminar:

Teaching an old equation new tricks: unlocking an elementary

interaction between light and magnetism using insights from

Spintronics  and Quantum technologies

Dr. Amir Capua

Faculty of Science, Applied Physics Department, The Hebrew University

Wednesday  May  29^th ,  2024

13:00-14:00

Abstract:

The ferromagnetic resonance (FMR) experiment is pivotal for detecting spin currents in Spintronics technology. It occurs on Gigahertz timescales due the relatively slow relaxation time of the spins in ferromagnets. In contrast, optical fields oscillate much faster, at ∼ 400 − 800 THz. Therefore, it seems unlikely that such fast-oscillating fields may interact with magnetic moments. However, by combining principles from quantum optics, we have recently realized that the equations governing the FMR experiment are even relevant for magnetic fields that oscillate much faster, at optical frequencies. Namely, the interaction between optical beams and the magnetization is made possible. We find that the strength of the interaction is determined by an elementary efficiency parameter η=αγH/f_opt, where H is the amplitude of the optical magnetic field, α is the dissipation rate of spin angular momentum to the lattice, and f_opt and γ are the optical frequency and gyromagnetic ratio. Our results shed light on a variety of highly debated experimental observations on the interaction between optical fields and ferromagnets that have been reported in the last 25 years.

Dr. Amir Capua -  Short Bio

Assistant Prof. Amir Capua heads the Spintronics Lab at the Institute of Electrical Engineering and Applied Physics at the Hebrew University of Jerusalem. Amir received his Ph.D. (2013) from the Electrical Engineering Department at the Technion, Israel where he worked in the fields of semiconductor laser physics and quantum optics. In 2013 Amir joined the Spintronics research group at the IBM Almaden Research Labs in California managed by Prof. Stuart Parkin and in 2016 he joined the Max Planck Institute for Microstructure Physics, Germany. Since 2017 Amir heads the Spintronics Lab at the Hebrew University. His team explores spin transport and magnetization dynamics for novel sensing, processing, and memory applications.

Electrical Engineering Systems Seminar

Speaker: Elad Yoshai

M.Sc. student under the supervision of Prof. Natan T. Shaked

Wednesday, 29^th May 2024, at 15:30

Room 011, Kitot Building, Faculty of Engineering

Domain Adaptation and Super Resolution in Histopathology Using Deep Learning

Abstract

Histopathology plays a pivotal role in medical diagnostics. In contrast to preparing permanent sections for histopathology, a time-consuming process, preparing frozen sections is significantly faster and can be performed during surgery, where the sample scanning time should be optimized. Super-resolution techniques allow imaging the sample in lower magnification and sparing scanning time. This thesis research presents a new approach to super resolution for histopathological frozen sections, with focus on achieving better distortion measures, rather than pursuing photorealistic images that may compromise critical diagnostic information. Our deep-learning architecture focuses on learning the error between interpolated images and real images; thereby it generates high-resolution images while preserving critical image details, reducing the risk of diagnostic misinterpretation. This is done by leveraging the loss functions in the frequency domain, assigning higher weights to the reconstruction of complex, high-frequency components. In comparison to existing methods, we obtained significant improvements in terms of Structural Similarity Index (SSIM) and Peak Signal-to-Noise Ratio (PSNR), as well as indicated details that lost in the low-resolution frozen-section images, affecting the pathologist’s clinical decisions. Our approach has a great potential in providing more-rapid frozen-section imaging, with less scanning, while preserving the high resolution in the imaged sample.

However, even though frozen sections are often used for rapid diagnosis during surgeries, as they can be produced within minutes, they suffer from artifacts and are generally missing details for diagnosis, specifically inside the cell nuclei region. On the other hand, permanent sections contain more details with diagnostic value but requires intensive process which takes hours of preparation. Therefore, in addition to the super-resolution we present a generative deep learning approach to convert frozen sections into permanent sections using a unique method that puts more attention on the nuclei regions, which are important for diagnosis. This is done by using a segmented attention network, incorporating nuclei-segmented images in the training process and an additional loss to push the network to improve the essential missing contents of the nuclei region in the generated permanent image. We tested our method on various tissues including kidney, breast, and colon. Our approach can significantly improve the histological efficiency and the accuracy and can be integrated in the existing laboratory workflows to generate permanent sections within seconds from frozen sections acquired rapidly.

השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום שם מלא + מספר ת.ז. בדף הנוכחות שיועבר באולם במהלך הסמינר