Lung Tumor Detection using Computer Vision: Cheenta Research Program

How AI is Helping Detect Lung Cancer: A Young Researcher's Journey

Artificial intelligence (AI) is making significant advancements in medicine. One young researcher, Rushil Reddy, a 10th grader from Germantown Academy, is contributing to this progress with an AI project designed to improve lung cancer detection. His work earned him second place at the Pennsylvania Junior Academy of Science (PJAS) competition, marking the beginning of an exciting journey.

Why This Project?

Rushil has always been passionate about science, especially medicine, inspired by his father, a doctor. He combined this interest with machine learning to develop a project that enhances lung cancer detection using AI.

Lung cancer remains one of the leading causes of cancer-related deaths in the U.S. Early detection significantly increases survival rates. Doctors typically use CT and PET scans to identify abnormalities. However, determining whether a tumor is cancerous or benign requires time and expertise. Rushil aimed to speed up this process and improve accuracy with AI.

How Does It Work?

Rushil’s project optimizes machine learning models to detect lung tumors. His approach improves the screening process, which involves:

Initial Checkups – Doctors assess risk factors such as smoking history, age, and family background.
Medical Imaging – CT or PET scans help detect abnormalities.
Diagnosis – If a tumor is found, doctors determine if it is cancerous or benign through further tests like biopsies.

Rushil’s AI model automates the final step. Instead of manually analyzing scans, the AI extracts key features and predicts the probability of cancer. This allows doctors to make faster and better-informed decisions.

The AI Model Behind the Project

A key tool in this project is the Brock Model, a widely used predictive model for lung cancer. This model evaluates several factors, including:

Age and sex
Family history of cancer
Nodule size and type
Nodule location in the lung
Irregular growth patterns (speculation)

Rushil’s AI model extracts these features from CT and PET scans. It then applies this data to the Brock Model to determine the likelihood of cancer. As a result, doctors can decide whether additional tests, such as a biopsy, are necessary.

Challenges and Future Work

Although the project has already shown promising results, Rushil plans to make further improvements. His next steps include:

Enhancing tumor segmentation using AI models like Segment Anything Foundation Model.
Testing different machine learning models to improve accuracy.
Extracting additional features like speculation and emphysema for more precise predictions.

Beyond building an AI model, Rushil’s work focuses on making AI decisions more transparent for doctors. Many medical professionals hesitate to rely on AI-based diagnoses because they do not fully understand how models reach conclusions. By aligning AI with well-established medical methods, Rushil ensures greater trust in its results.

Final Thoughts

This project highlights how young innovators contribute to medical advancements. Rushil’s work bridges the gap between AI and healthcare, making cancer detection faster and more accessible. As research continues, AI is set to revolutionize early cancer detection and diagnosis.

Shonku - Math Messenger and Micro Blogging App

Shonku math messenger has the following features:

Chat using math equations and real-time preview. Write in Latex.
Do micro-blogging with math. Follow other users.
Make groups with your connections. Make selected users invisible or muted.

Screenshots

Sign up for the test - version.

Optimizing Urban Accessibility: Building a 15-Minute City with Steiner Tree Approximation

A Research Paper by Prishaa Shrimali (USA, Grade 10)

Introduction

Urban planning is increasingly focused on creating sustainable, accessible cities where essential services are within easy reach. The 15-minute city (15-MC) model is an innovative approach aimed at structuring urban spaces so that residents can access key services, like healthcare, shopping, and recreational facilities, within a short walking or biking distance. In the study Optimizing Urban Accessibility: Constructing a 15-Minute City Using Steiner Tree Approximation, researchers introduce a method of applying graph theory—particularly the Steiner tree problem—to efficiently design 15-minute cities.

Methodology

The study employs the Steiner tree problem, which seeks to find the minimum-weight network that connects selected key points, called terminals (e.g., service locations). Using this graph-based approach, the model minimizes travel time between key amenities by optimizing the pathways that connect them. Unlike models that place a focus on residential areas, this approach prioritizes service locations, making it computationally efficient.

The model is applied to Manhattan, using the city's pedestrian network to highlight service accessibility. Here, amenities such as pharmacies, post offices, and supermarkets serve as the Steiner tree's focal points. The OSMnx Python library is used to pull data from Open Street Maps, allowing for a practical analysis of service accessibility within a 15-minute walking radius.

Watch the Video

Key Highlights

Efficient Service Connectivity: By focusing on connecting service points, this model minimizes computational complexity and offers a feasible layout for urban planners to improve walkability.
Dense Network Coverage in Manhattan: The analysis reveals that central and southern Manhattan already supports a high level of walkability, with the Steiner tree model indicating most residents in these areas can reach essential services within a short walk.
Areas for Improvement: The study highlights gaps in the northern parts of Manhattan, suggesting areas where pedestrian access to amenities could be enhanced.
Digital City Models: The study's approach yields detailed digital models that serve as practical tools for urban planners to optimize mobility, service placement, and sustainable design.

Inference

The Steiner tree-based method for designing a 15-minute city provides urban planners with an actionable framework to improve urban accessibility. While central areas of Manhattan demonstrate a high density of accessible services, regions like northern Manhattan could benefit from increased service points or better connectivity. This graph-based approach also shows promise for future expansions, such as multi-criteria optimization considering factors like environmental impact and cost.

In sum, the paper underscores the effectiveness of leveraging graph theory in urban planning and establishes a solid foundation for implementing sustainable, accessible city models that can adapt to the unique needs of various urban landscapes.

Research in School: Epidemiological Modelling and Outbreak Prediction using Hyperbolic Geometry | by Raghav Pai and Shreyas Vivek

Schedule: 26th October 2024 (Saturday)

Time: 5:00PM IST

About the Presenters:

Raghav Pai: He is a Grade 11 student based in Maharastra, Mumbai. He is part of Cheenta since 8 months

Shreyas Vivek: He is a Grade 11 student based in Dubai, UAE.

Abstract:

This paper introduces a novel approach to modeling disease transmission using hyperbolic geometry, specifically the Poincaré disk model. Traditional models like Susceptible-Infected-Recovered (SIR) assume homogeneous populations, which oversimplifies real-world interactions. By incorporating hyperbolic distance, the Poincaré disk model captures spatial clustering and irregular social interactions, offering a more realistic framework for studying epidemics. Simulations of the first wave of COVID-19 in India were performed using both the Poincaré disk and SIR models. Results show that the Poincaré disk model better captures localized transmission patterns and spatial dynamics, providing deeper insights into how diseases spread through structured populations. This approach highlights the importance of accounting for social network structures in epidemic modeling, offering valuable guidance for targeted public health interventions such as localized lockdowns and vaccination strategies.Our findings demonstrate the advantages of hyperbolic geometry in epidemiological modeling, with potential applications for improving future outbreak predictions and interventions.

This session highlighted how mathematical concepts can be applied to understand and predict the spread of infectious diseases more accurately.

Watch the session here

Modeling Epidemics in Hyperbolic Space

Traditional epidemiological models use Euclidean geometry, which may not capture the complex structure of real-world social networks. Hyperbolic geometry, with its curved spaces, better represents these networks by accounting for high clustering and varying levels of social interactions. The model presented maps social interactions onto a hyperbolic plane, visualizing the spread of disease as expanding waves through a network.

Enhancing Prediction Accuracy

The hyperbolic approach allows for the identification of critical clusters where outbreaks may intensify. Compared to traditional methods, these predictions can be more precise, helping to target interventions like vaccinations or lockdowns in specific high-risk zones.

Applications Beyond Epidemiology

While the primary focus was on epidemic modeling, the use of hyperbolic geometry extends to other areas, such as analyzing information spread in social media, enhancing cybersecurity, and understanding financial network risks.