Problem Statement

One of the key challenges faced by urostomy patients is the lack of physiological feedback when their bag is filling, particularly during sleep, leaving them unaware if the bag is overfilling. The risk of urine spillage presents significant difficulties, including physical discomfort, social anxiety in public settings, and disrupted sleep.

Solution Overview

I designed and developed a smart device consisting of a reusable sleeve equipped with Hall effect sensors that accurately measure the bag's filling. The system offers both discrete and audible alerts to accommodate the diverse needs of patients in their daily lives. With a 95% accurate two-level warning system, this solution ensures timely alerts, preventing spillages and thereby reducing the inequalities associated with the use of stoma bags.

Project Phases

Throughout this project, I was responsible for the entire process, from ideation to the final testing of the prototype. I began by conceptualizing the solution and creating initial designs. I then moved on to prototyping, where I iteratively refined the design and functionality of the sensing sleeve. After extensive testing and optimization, I built the final version of the device and conducted comprehensive testing to ensure its reliability and effectiveness.

Impact and Significance

A smart urostomy bag monitoring system has the potential to significantly improve the quality of life for urostomy patients. By providing real-time sensory feedback and preventing accidental leakage, the system can address the physical discomfort, social anxiety, and sleep disruption associated with urostomy bags. This innovative solution aims to reduce the inequalities faced by urostomy patients and enhance their overall well-being.

Segmentation of Dental Implants and Metal Fillers

A series of image processing techniques are applied to the X-ray images. The process begins by isolating areas corresponding to dental implants and metal fillers. This segmentation is achieved using intensity-based thresholds and morphological operations tailored for dental X-rays.

Nearest Neighbor Pixel Analysis

After the initial segmentation, a nearest neighbor algorithm is employed to refine the boundaries of the regions of interest. This step helps in accurately classifying and segmenting the implants by considering the local pixel neighborhoods.

Volume Calculation and Implant Count

Once the implants are accurately delineated, the system computes the volume of each implant. This is achieved by integrating the pixel area across the segmented regions. The total number of implants is then determined by counting the distinct segmented regions.

Implementation

The project is implemented in Python, utilizing libraries for image processing, NumPy for numerical operations, segmentation and image analysis techniques.

Quality Management System (QMS)

The QMS ensures that the design, development, and commercialization processes of MindSync adhere to the highest quality standards and regulatory requirements.

Regulatory Strategy

A detailed regulatory strategy was developed to navigate the complex landscape of medical device regulations in USA, EU, and the UK. My participation foucused on the development of the regulatory strategy for the UK market. This includes identifying the applicable regulatory pathways, such as the Medical Device Regulation (MDR) or the Software as a Medical Device (SaMD) framework, and ensuring compliance with all relevant requirements.

Data Description

The dataset I used consists of structural connectivity matrices for each baby, along with their age at scan and preterm birth status. The matrices have dimensions of 90x90, representing the connectivity between 90 brain regions. I leveraged this data to develop a predictive model for prematurity.

Machine Learning Approach

I employed a Support Vector Machine (SVM) classifier in combination with Principal Component Analysis (PCA) for dimensionality reduction. I trained the SVM algorithm on the connectivity matrices to learn patterns associated with preterm birth, while using PCA to extract the most informative features from the high-dimensional data.

Model Evaluation

I evaluated the trained model on a separate test set to assess its performance. The key metrics I used for evaluation include accuracy, sensitivity, specificity, and mean recall. My model achieved an impressive accuracy of 92%, indicating its ability to correctly predict preterm birth in a high percentage of cases. The sensitivity obtained was 79% and a specificity of 96%. The obtained recall was 87% further confirming the model's effectiveness in correctly identifying both preterm and full-term cases.

Identification of Predictive Connections

In addition to predicting preterm birth, I aimed to identify the brain connections that are most predictive of prematurity. By analyzing the learned weights of the SVM classifier and the principal components obtained from PCA, I determined the most informative connections. These findings can provide valuable insights into the underlying neurological mechanisms associated with preterm birth and guide further research in this domain.

Technologies Used

• Next.js - A React framework for building server-side rendered and statically generated websites
• Tailwind CSS - A utility-first CSS framework for rapidly building custom user interfaces
• React - A JavaScript library for building user interfaces
• HTML5 and CSS3 - Markup and styling languages for structuring and presenting web content
• Vercel - A platform for hosting and deploying web applications
• Git & Github - Version control
• Matter.js - A 2D physics engine for building interactive web applications

Key Features

• Responsive design that adapts to different screen sizes and devices
• Dark mode toggle for enhanced user experience and accessibility
• Smooth scrolling navigation for easy exploration of different sections
• Project showcase with detailed descriptions
• Contact form for potential clients and collaborators to get in touch

Development Process

The development process involved several stages, including:

1. Planning and wireframing the website structure and layout
2. Setting up the Next.js project and configuring Tailwind CSS
3. Developing reusable components and implementing the design
4. Integrating project data and optimizing performance
5. Deploying the website to a hosting platform