Register now After registration you will be able to apply for this opportunity online.
The AI Sleep Doctor: AI-based evaluation of medical sleep examinations
The process of evaluating sleep examinations and diagnosing sleep disorders through polysomnographies (PSGs) is labor-intensive as it requires manual analysis from sleep technicians and doctors. In collaboration with Clinic Barmelweid, a leading sleep and rehabilitation clinic in northwestern Switzerland, we plan to automate this process using machine learning models. Clinic Barmelweid conducts approximately 400-450 PSGs annually and has access to a dataset of more than 5,000 recordings.
The project revolves around the implementation of an AI-based solution, termed "The AI Sleep Doctor," aimed at automating the scoring of sleep recordings, specifically polysomnographies (PSGs). Currently, Clinic Barmelweid conducts approximately 400-450 PSGs annually, requiring up to two hours of additional effort by sleep technicians and doctors for scoring.
The project revolves around the implementation of an AI-based solution, termed "The AI Sleep Doctor," aimed at automating the scoring of sleep recordings, specifically polysomnographies (PSGs). Currently, Clinic Barmelweid conducts approximately 400-450 PSGs annually, requiring up to two hours of additional effort by sleep technicians and doctors for scoring.
The project aims to streamline the process of scoring sleep recordings by employing AI models capable of analyzing PSG recordings. Specific subtasks include:
1. **Literature Research:** Review existing literature to gain insights into PSG recordings and machine learning models used for classifying multimodal time series data.
1. **Dataset Gathering and Preprocessing:** Obtain available datasets and preprocess the data to remove noise, artifacts, and standardize formats. Additionally, explore the dataset to understand its distributions and special characteristics.
1. **Development of Binary Classifier:** Collaborate with sleep experts from Barmelweid to develop a binary classifier capable of distinguishing regular from pathological sleep patterns.
1. **Development of Disorder-Specific Classifiers:** Develop classifiers to identify patients with insomnia or sleep apnea syndrome, which are the two most common sleep disorders.
1. **Refinement for Multi-Class Classification:** Extend the classifiers to classify across eight sleep disorder domains, encompassing a broader range of conditions.
1. **Reporting Relevant Metrics:** For all developed models, report relevant metrics such as accuracy, precision, recall, and F1-score.
1. **Optional: Integration of Language Models:** Optionally, explore the integration of language models to automate the generation of diagnostic reports based on the automated analysis of sleep recordings.
The project aims to streamline the process of scoring sleep recordings by employing AI models capable of analyzing PSG recordings. Specific subtasks include:
1. **Literature Research:** Review existing literature to gain insights into PSG recordings and machine learning models used for classifying multimodal time series data. 1. **Dataset Gathering and Preprocessing:** Obtain available datasets and preprocess the data to remove noise, artifacts, and standardize formats. Additionally, explore the dataset to understand its distributions and special characteristics. 1. **Development of Binary Classifier:** Collaborate with sleep experts from Barmelweid to develop a binary classifier capable of distinguishing regular from pathological sleep patterns. 1. **Development of Disorder-Specific Classifiers:** Develop classifiers to identify patients with insomnia or sleep apnea syndrome, which are the two most common sleep disorders. 1. **Refinement for Multi-Class Classification:** Extend the classifiers to classify across eight sleep disorder domains, encompassing a broader range of conditions. 1. **Reporting Relevant Metrics:** For all developed models, report relevant metrics such as accuracy, precision, recall, and F1-score. 1. **Optional: Integration of Language Models:** Optionally, explore the integration of language models to automate the generation of diagnostic reports based on the automated analysis of sleep recordings.
In this interdisciplinary project, you will have access to various public and private databases containing thousands of sleep recordings. Furthermore, you will be able to visit one of the most advanced sleep centers in Switzerland and engage in discussions with experts in sleep medicine, as well as in machine learning. Additionally, you will have access to our LeoMed tenant to efficiently run your models and you will have your own permanent work desk in our new and modern lab in GLC.
In this interdisciplinary project, you will have access to various public and private databases containing thousands of sleep recordings. Furthermore, you will be able to visit one of the most advanced sleep centers in Switzerland and engage in discussions with experts in sleep medicine, as well as in machine learning. Additionally, you will have access to our LeoMed tenant to efficiently run your models and you will have your own permanent work desk in our new and modern lab in GLC.
You are an engineering student with a strong background in machine learning and are interested in applied machine learning.
You are an engineering student with a strong background in machine learning and are interested in applied machine learning.
If interested in this project, please send your CV and latest transcript to Sara Padilla Neira (sara.padillaneira@hest.ethz.ch) and Alexander Breuss (alexander.breuss@hest.ethz.ch).
If interested in this project, please send your CV and latest transcript to Sara Padilla Neira (sara.padillaneira@hest.ethz.ch) and Alexander Breuss (alexander.breuss@hest.ethz.ch).