TEARNET: AI-ENHANCED SMARTPHONE APPLICATION FOR REAL-TIME DRY EYE DIAGNOSIS USING OCULAR IMAGING AND

Abstract

TearNet: Al - Enhanced Smartphone Application for Real-Time Dry Eye Diagnosis Using Ocular Imaging and Blink Detection ABSTRACT OF THE INVENTION: This invention is an Al-driven diagnostic tool for assessing Dry Eye Disease (DED) and Meibomian Gland Dysfunction (MGD) using a smartphone. It employs advanced deep learning techniques, including MobileNetV3 for ocular image classification and LSTM networks for analyzing blink patterns. A handheld infrared camera connects to the smartphone to capture detailed eye images for MGD analysis. The system also detects blink patterns using the Eye Aspect Ratio (EAR) method, providing insights into abnormal blinking related to dry eye symptoms. Additionally, a symptom questionnaire allows users to report subjective symptoms, which, combined with image and blink analysis, helps grade the severity of DED and MGD. This portable and affordable tool enables real-time analysis, enhancing diagnostic accuracy and efficiency, ultimately improving early detection and treatment outcomes for patients and healthcare providers.

Specification

FORM 2
THE PATENT ACT 1970
(39 OF 1970)
&
The Patent Rules, 2003
PROVISIONAL/COMPLETE SPECIFICATION
(See Section 10 and rule 13)

1.TITLE OF THE INVENTION:
TearNet: Al - Enhanced Smartphone Application for Real-Time Dry Eye Diagnosis Using Ocular
Imaging and Blink Detection

2.APPLICANTS (S)
(a) NAME: 1. Easwari Engineering College
(b) NATIONALITY: INDIAN
(c) ADDRESS: 1. Easwari Engineering College, Bharathi Salai, Ramapuram, Chennai - 600089

3.PREAMBLE TO THE DESCRIPTION:
28-Qct-2024/130854/202441082232/Form 2(Title Page)
PROVISIONAL
The-following-specification-describes-the
invention
COMPLETE
The following specification particularly describes
the invention and the manner in which it is to be
performed.

4. DESCRIPTION (Description shall start from next page.)
ATTACHED
5. CLAIMS (not applicable for provisional specification: Claims should start with the preamble
"l/we claim" on separate page)
ATTACHED
6. DATE AND SIGNATURE (to be given at the end of last page of specification)
ATTACHED
7. ABSTRACT OF THE INVENTION (to be given along with complete specification on separate
page)
ATTACHED

Note:
*Repeat boxes in case of more than one entry.
*To be signed by the applicant(s) or by authorized registered patent agent.
*Name of the applicant should be given in full, family name in the beginning.
*Complete address of the applicant should be given stating the postal Index no./code, state
and country.
*Strike out the column which is/are not applicable

DESCRIPTION:
28-Q.pt-2024/130854/20244.1082232/Form 2(Title Page)
[0001] TearNet is a smartphone-based Al-powered application designed to diagnose and assess the severity of Dry Eye Disease (DED) in real time, focusing on Meibomian Gland Dysfunction (MGD). Using MobileNetV3, a deep learning model, TearNet classifies ocular images captured by a handheld infrared camera that can be attached to a smartphone. This enables accurate detection of MGD abnormalities. The system also utilizes LSTM (Long Short-Term Memory) networks to analyze blink patterns through the Eye Aspect Ratio (EAR) method, captured by the smartphone camera. By integrating blink detection with image analysis, TearNet provides a multimodal approach to diagnosing DED. Additionally, TearNet includes a questionnaire-based assessment using a random forest model to evaluate patient-reported symptoms. The combination of these data inputs allows the system to grade the severity of DED (normal, mild, moderate, severe), offering a holistic and cost-effective diagnostic solution. This portable and easy- to-use platform enhances accessibility to early DED diagnosis, empowering clinicians and users alike to manage the disease effectively.

PRIOR ART AND BACKGROUND:
[0002] Dry Eye Disease (DED) and Meibomian Gland Dysfunction (MGD) are prevalent ocular conditions that significantly affect quality of life. Current diagnostic methods rely on clinical tools such as tear film breakup time (TFBUT) and symptom-based questionnaires, but these methods often lack precision, are time-consuming, and require specialized equipment. Existing solutions, such as digital eyewear systems (e.g., US11940626B1) and smartphone applications (e.g., "DryEyeRhythm"), offer partial assessments by focusing on monitoring blink patterns or self-reported symptoms. However, these approaches fail to provide a comprehensive, real-time diagnosis of DED and MGD, often overlooking the integration of structural and functional assessments. Moreover, the reliance on manual monitoring and clinical environments limits accessibility and the ability to promptly respond to fluctuations in patient symptoms. Addressing these gaps requires innovative solutions that leverage Al, image analysis, and blink detection to provide real-time, automated, and comprehensive diagnostic tools for DED and MGD, enhancing accessibility and accuracy in various settings.

[00030] CN.10.4334129B - "Dry Eye Disease Detection System": This patent describes a system for detecting dry eye disease by using infrared imaging to measure tear film breakup time (TFBUT), which assesses tear stability. The system focuses on evaluating tear film characteristics to diagnose dry eye disease. By utilizing infrared technology, it aims to provide a non-invasive approach to detecting tear film abnormalities and diagnosing related conditions.

[0004] US9028065B2 - "Ophthalmologic Apparatus and Image Classification Method": Covers an ophthalmologic device that uses image classification techniques to assess various eye conditions, including dry eye disease, by capturing and analyzing ocular images. The apparatus focuses on diagnosing eye diseases through objective image analysis, providing a tool for clinicians to evaluate ocular health based on visual data. The system emphasizes the use of conventional image classification methods to detect and classify eye abnormalities.

[0005] US11940626B1 - "Digital Eyewear with Procedures Related to Dry Eye Management": It describes digital eyewear designed to monitor eye health by tracking blink frequency and moisture levels, providing real-time feedback for managing dry eye disease. The system focuses on using wearable technology to observe key eye parameters, helping users monitor and adjust their eye care routine based on the data collected. The primary goal is to offer a personalized approach to managing dry eye symptoms through continuous monitoring.

[0006] JP7308144B2 - "System and Method for Detection of Eye Disease": Our invention highlights the importance of advanced technologies such as deep learning models, real- time image analysis, and comprehensive diagnostic tools, which are not typically addressed in traditional methods like the one described in CN104334129B. While the patent focuses on tear film breakup analysis, our invention emphasizes a multimodal approach involving blink detection, meibomian gland analysis, and symptom questionnaires. CN104334129B might rely on conventional tear film evaluation techniques, whereas our invention underscores the need for a broader diagnostic framework and enhanced accessibility through smartphone integration.

[0007] US11166871B2 - "Eyelid Care Appliance": Outlines a device designed for eyelid care, targeting conditions like dry eyes and meibomian gland dysfunction by applying heat and moisture. In contrast, TearNet leverages smartphone technology to diagnose dry eye disease through advanced image analysis and blink pattern detection. While US11166871B2 focuses on treatment via a physical appliance, TearNet emphasizes diagnostic capabilities, offering a multimodal approach that combines imaging, blink detection, and patient questionnaires. Additionally, TearNet's smartphone-based solution enhances portability and accessibility, allowing users to conduct assessments in realtime, unlike the potentially less portable device in US11166871B2

[008] US20210205131A1 - "Methods and Apparatuses for Treatment of Meibomian Gland Dysfunction": Details methods and devices intended to treat meibomian gland dysfunction (MGD) by enhancing gland function and restoring tear film stability. In contrast, TearNet employs smartphone technology for diagnostic purposes, analyzing eye images and detecting blink patterns to assess dry eye disease severity. While US20210205131A1 concentrates on the treatment of MGD, TearNet offers a holistic diagnostic approach that encompasses symptom assessment through questionnaires. Furthermore, US20210205131A1 may require specialized equipment, limiting portability, whereas TearNet's smartphone-based application ensures greater accessibility and convenience for users.

OBJECTIVE:
[010] The objective of TearNet is to develop a portable, Al-driven diagnostic tool for realtime detection and monitoring of Dry Eye Disease (DED) using a smartphone application. TearNet aims to provide a highly accessible, user-friendly solution that eliminates the need for expensive clinical equipment, making it ideal for use in remote, rural, and resource-constrained environments.

[011] By analyzing ocular images captured with a handheld infrared camera, TearNet utilizes MobileNetV3 CNN architecture to classify Meibomian Gland Dysfunction (MGD), a primary cause of DED. In addition, it performs automated blink pattern analysis using the smartphone's front camera and the Eye Aspect Ratio (EAR) method. The application integrates LSTM networks to detect abnormal blink patterns commonly associated with DED, providing a comprehensive assessment of blink frequency, duration, and irregularity.

[012] TearNet delivers real-time grading of DED severity by combining MGD classification and blink analysis, offering instant feedback on eye health categorized into normal, mild, moderate, or severe levels. This makes it possible for users to monitor the progression of their condition effectively.

[0013] In addition to image and blink analysis, TearNet incorporates a questionnairebased symptom assessment, which, using Random Forest models, predicts the severity of DED based on self-reported symptoms. The combination of Al-driven diagnosis and user input offers a holistic view of eye health, empowering users to take proactive steps in managing their condition.

[014] TearNet's cost-effective approach to diagnosing and monitoring DED, coupled with its ability to facilitate remote monitoring and follow-up care, makes it an ideal solution for both individual users and healthcare providers seeking continuous, accurate, and accessible eye health diagnostics.

SUMMARY:
[0015] TearNet uses a convolutional neural network (CNN) architecture based on MobileNetV3 to analyze ocular images and classify the severity of Meibomian Gland Dysfunction (MGD). MGD is a major contributor to DED, and accurate classification of gland dysfunction is essential for diagnosing and managing dry eye conditions. In addition to MGD analysis, the application also tracks blink patterns using the Eye Aspect Ratio (EAR) method through the smartphone's camera. To detect abnormal blink patterns, a Long Short-Term Memory (LSTM) neural network is employed, which tracks blink frequency, duration, and irregularity over time.

[0016] The results of both the MGD classification and blink pattern analysis are combined to grade the severity of Dry Eye Disease into categories such as normal, mild, moderate, and severe. This grading system allows the user to understand the progression of their condition in real-time.

DETAILED TECHNICAL DESCRIPTION:
[0017] TearNet is a cutting-edge smartphone application designed to provide accurate and efficient diagnostics for dry eye disease (DED), specifically focusing on the assessment of meibornian, gland dysfunction (MGD). The application integrates advanced image processing techniques and machine learning algorithms to enhance the evaluation of ocular health using widely available smartphone technology.

[0018] At the core of TearNet is MobileNetV3, a powerful convolutional neural network optimized for mobile devices. This network is utilized for the classification of ocular images captured through a handheld infrared camera that can be easily attached to a smartphone. The infrared imaging technology allows for a non- invasive examination of the meibomian glands, which are crucial for maintaining tear film stability. By analyzing the quality and quantity of meibomian gland output, the application can provide insights into the severity of MGD and its contribution to dry eye symptoms.

[0019] The workflow of TearNet begins with image acquisition, where users can capture high-resolution infrared images of their eyelids and the surrounding ocular surface. These images are then processed through the MobileNetV3 model, which classifies the condition of the meibomian glands based on learned features from extensive training datasets. The model is fine-tuned to detect various indicators of dysfunction, such as gland obstruction or atrophy, allowing for a nuanced assessment of ocular health.

[0020] In addition to image analysis, TearNet employs Long Short-Term Memory (LSTM) networks to evaluate blink patterns, another critical factor in diagnosing dry eye disease. The smartphone's camera continuously tracks blink frequency, duration, and the intervals between blinks, providing a dynamic view of the user's blink behavior over time. Abnormalities in blink patterns, such as reduced blink rate or incomplete blinks, can be indicative of dry eye conditions and are integrated into the overall diagnostic framework of TearNet.

[0021] To further enhance the diagnostic process, TearNet includes a symptom questionnaire that gathers subjective patient-reported data regarding eye comfort, frequency of symptoms, and overall visual health. This multimodal approach ensures a holistic evaluation, as it combines objective findings from image analysis and blink detection with the patient's own experiences. The integration of these data streams allows TearNet to generate a comprehensive severity grading for dry eye disease, categorizing it into four levels: normal, mild, moderate, and severe.

[0022] User accessibility is a key design principle of TearNet. By leveraging the widespread availability of smartphones, the application offers a portable, cost- effective solution for users to monitor their ocular health from the comfort of their homes. This democratization of technology not only makes diagnostic tools more accessible but also encourages proactive management of dry eye symptoms, leading to better patient outcomes.

[0023] Moreover, TearNet's design incorporates a user-friendly interface that simplifies navigation and enhances user engagement. Users can easily capture images, record blink data, and respond to the symptom questionnaire, all within a cohesive app experience. The application also provides real-time feedback and insights based oh the collected data, enabling users to understand their condition better and make informed decisions regarding their eye care.

[0024] Overall, TearNet represents a significant advancement in the diagnosis and management of dry eye disease. By combining sophisticated image analysis, blink detection, and user-centered design, the application aims to improve the accuracy and efficiency of dry eye diagnostics while empowering users to take control of their ocular health. This innovative approach addresses existing gaps in current diagnostic practices and contributes to the ongoing effort to enhance patient care in the field of ophthalmology.

BRIEF DESCRIPTION OF THE DRAWING:
[0025] This flowchart illustrates how TearNet's system captures and analyzes ocular images and blink patterns to diagnose dry eye disease and meibomian gland dysfunction (MGD). The smartphone-based infrared camera captures eye images, which are processed using MobileNetV3 for meibomian gland analysis. Simultaneously, the smartphone's camera tracks the user's blink patterns, analyzed using LSTM networks for abnormal blink pattern detection. The diagnostic data is combined with the results from a user-completed symptom questionnaire and then processed by the system's backend to grade the severity of dry eye disease.

Fig 1. Represents System Architecture for TearNet
Fig 2. Depicts Smartphone and Infrared Camera Integration
Fig 3. Demonstrates Prototype Implementation for Blink Detection and MGD Analysis

TearNet prototype utilizing smartphone infrared cameras to capture ocular images for MGD analysis. The combination of machine learning models for blink detection and image classification facilitates real-time dry eye disease diagnostics.
[0027] The image shows a smartphone setup with an attached infrared camera capturing an ocular image. This setup is connected to a power source via USB, indicating the smartphone is running diagnostic software. The background contains laboratory tools, suggesting an experimental environment.


CLAIM:
Claim
l/We claim

1. The Al-based diagnostic system for detecting dry eye disease and meibomian gland dysfunction (MGD) as claimed in claim 1, uses smartphone technology integrated with an infrared camera for real-time image capture and analysis of
ocular features.

2. The Al-based diagnostic system as claimed in claim 1, utilizes MobileNetV3 for meibomian gland analysis and LSTM networks for blink pattern detection to provide a comprehensive and automated diagnosis.

3. The Al-based diagnostic system as claimed in claim 1, incorporates a symptom questionnaire for the patient to complement the Al-driven image and blink analysis, offering a multi-faceted diagnostic approach for dry eye disease severity.

4. The Al-based diagnostic system as claimed in claim 1, is portable and costeffective, leveraging smartphone technology to allow easy access to dry eye disease diagnosis without the need for specialized equipment.

5. The Al-based diagnostic system as claimed in claim 1, uses real-time data transmission to a central server or cloud platform, enabling users and healthcare providers to remotely monitor and review diagnostic results via smartphone or web
application.

6. The Al-based diagnostic system as claimed in claim 1, integrates an alert mechanism to notify the user of any detected abnormalities in meibomian gland function or blink patterns, prompting timely medical consultation.

7. The Al-based diagnostic system as claimed in claim 1, is designed to be userfriendly and accessible to non-specialists, making it suitable for widespread use in both clinical and non-clinical settings for early detection and management eye disease.

Documents