Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 593 https://internationalpubls.com Image Classification with Hypertensive Angle Disease Detection with Geometric Local Derivative Pre-Processing Dr Anand Rajasekaran1, Rajesh Natarajan2 ,Shankar R3 , Sowmya V L4 , Mr.R vikkram5, Subash Chandra Bose S6 1Associate Professor and Head, Department of MCA, Guru Nanak College (Autonomous), Velachery, Chennai, Tamil Nadu, India. E-mail: anand.r@gurunanakcollege.edu.in 2Lecturer, Information Technology Department, College of Computing and Information Sciences, University of Technology and Applied Sciences-Shinas, AI-Aqr, Shinas-324, Oman. E-mail: rajesh.natarajan@utas.edu 3Assistant Professor, Department of Information Technology, Karpagam College of Engineering, Coimbatore, Tamil Nadu, India. E-mail: vikkram.r@kce.ac.in 4Assistant Professor, Department of Computer Science and Engineering, BMS Institute of Technology and Management, Bengaluru, Karnataka, India. E-mail: shankar@bmsit.in 5Research Scholar, Department of Computer Science and Engineering, BMS Institute of Technology and Management, Bengaluru, Karnataka, India. E-mail: soumyavl@bmsit.in 6Assistant Professor, Department of Computer Science, Islamiah College (Autonomous), Vaniyambadi, Tamil Nadu, India. E-mail: bosesubash@islamiahcollege.edu.in Article History: Received: 08-08-2024 Revised: 17-09-2024 Accepted: 26-09-2024 Abstract: It is recognized as conjunctivitis stands as that of the second-greatest source of disability. Early detection and management of ophthalmology are essential for preventing disease owing to the benign character of loss of vision mostly in early stages of the illness and the permanent condition of vision in later stages. Straight retinal inspection, commercial digital images, laser scanners spectroscopy, scanning infrared additional factor and confocal tomography (OCT) images can all is used it to identify hypertension. These findings suggest using a Vertical Harmonic oscillator Dynamic Hinge Supporting Machine Classifier (Embedded systems) to predict glaucoma. There really are three stages to the Repository system. All those are ophthalmology identification, extraction and classification, and processing. The retina source images first were analyzed using a Geometric Local Derivative Pre - processor architecture to extract the major characteristics required for early diagnosis. The precompiled images then are submitted to Harmonic oscillator Discontinuous Single - input single Probability distribution Extraction Of features to identify significant characteristics with the sensitivity for disease prediction. Lastly, Inter Hinge Gradient Boosting Retinal Identification employs the derived features to detect glaucoma early and effectively. The Retina Mri image datasets was utilized in Simulation experiments to examine the effectiveness of the suggested method, embedded systems. Computational complexity, sensitivities, and correctness performance criteria must be looked at with respect to different Optical image quantities. Keywords: Diagnosis of hypertension, confocal laser spectroscopy, Regional Geometric Sequential Sinusoidal, Continuous Trigonometric functions, Multi-class, Stochastic Destruction, Hinge and Logistic Regression. mailto:anand.r@gurunanakcollege.edu.in mailto:vikkram.r@kce.ac.in mailto:shankar@bmsit.in mailto:soumyavl@bmsit.in mailto:bosesubash@islamiahcollege.edu.in Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 594 https://internationalpubls.com 1. Introduction In recent times, it's been recognized because ophthalmology is a disorder which impacts individuals of all ages on a constant schedule instead of just the aged. The intense pressure of an excess fluid that is accessible inside the eyeball produces this hypertension. This liquid affects the eye's visual cortex, which produces an embarrassing defeat of vision. Ophthalmology is generally not examined unless permanent visual impairment produced by irreversible damage has happened. These objectives justify the need for a precise and timely ophthalmology treatment. By critically investigating the characteristics for retina of the eye through greater image, the Multiple Input Vectors Deep Convolutional neural Network (MFV-DBN) created an approach to identify this Retinal disease at the initial stages altogether. Ophthalmic separation, hemorrhage, and genetic deterioration all were evaluated as parts of a screening process. The resultant method of selecting was subject to an inter analysis mechanism in order to address issues about the handling of similar characteristics. Eventually, discrete wavelet transform and algorithmic methods for machine learning were employed to perform efficient classification, which enhanced the precision of disease diagnosis. The application of optical coherence Tomography angiogram (OCTA) images in computerized disease prediction employing Contrasting Restricted Accelerated Wavelet Transformation (Contrast limited adaptive histogram) was suggested in [2]. Ophthalmic Scorpion (OS) or Ophthalmic Displacement (OD) was obtained from of the patients' eyes in to determine whether they had hypertension. This Regional Phase Classification (LPQ) technique then was employed again for identification of important components. Inevitably, in order to combine and reduce the features and enhance the accuracy, principal component evaluation and data compression were being used. We propose a novel Asymmetric Pure sine wave Inter Hinged Support Machines Classifier (Assembler) for ophthalmology detection it comprises three stages to address those problems. For the goal of identifying ophthalmology, there are filtering, extraction and classification, and categorization. We examine the approach to use the Discusses the influence sample and a recently gathered Variable will cause data, different clinical datasets. The findings from the research reveal that our methodology is preferable over slashing alternatives for hypertension diagnosis. These foregoing are all this paper's main contributions: 1) hrough considering three separate stages or components of an Optical picture, a machine learning method for ophthalmology screening is proposed. The technique produces promising results because it may be enhanced more according to current retinal identification techniques. 2) In contrast towards the computational framework, they additionally provide a breakthrough Wedge shaped Local Derivative Pre - processor algorithm significantly decreases the duration required to identify ophthalmology by extracting the most important features by combining the positions of neighboring pixels with nearby angular displacement. 3) To develop a Pure sine wave Continuous Single - input single Probabilistic extraction technique that will extract the important feature utilizing geographical values as well as the Stochastic distribution functions, increasing the efficiency of disease prediction.. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 595 https://internationalpubls.com 4) To provide a non - linear and nonlinear hinged vector support ophthalmology detection technique which maximizes its distance from every group towards the subsequent class, thereby enhancing sensitivities across four different classifications (NORMAL, CNV, DME and DRUSEN). 5) The efficiency of the Vertical Harmonic oscillator Inter Hinged Supporting Machine Learners (Assembler) with terms of ophthalmology detection rate, efficiency, and specificity has indeed been assessed using experiments.. The article is constructed out as follows: The topic of Optical images for disease diagnosis employing machine learning techniques is addressed in Part 2 along with some common issues. The Vertical Harmonic oscillator Inter Hinged Support Artificial Learners (ASM-HSML) methodology to ophthalmology detection from of the Image data is discussed in Section 3. These same investigational configurations have been laid out in Section 4 already when Fifth section considers the results of the implementation of the recommended Assembler technique and achievement parameters to determine the suggested work's eye disease time consumption, accurateness, and specificity, in addition to a comprehensive correlation with both the legislature functions. The results and important findings for the suggested method are addressed in Section 6. 1. Related works The most common irrevocable central nervous system vision disorders that contributes to disability is ophthalmology, as according researchers. Thereby, periodic medical checkups for prospective patients presenting were essential for promoting the clinical recognition of the disease. In to automatically identify hypertension, separation of an eye image or cups were carried out in a Machine Learning with CNN investigation [3]. Overall implementation of a selected segmentation technique as well as the characteristics being retrieved was stated to have a significant effect on the majority of the current techniques for ophthalmology assessment employing retina images, that heavily depended upon attributes based on handcrafting. For automatic ophthalmology assessment, five different image network help make decisions were employed [4], which then in return perfectly alright overall overall process. To differentiate among healthful and abnormal inflammatory images, once again another methodology comprising the combination of adaptive filtering statistically or texture features was evaluated from of the identified visual acuity region. Principal open-angle hypertension, angular velocity ophthalmology, tertiary macular degeneration, and medium pressure eye diseases are really the four basic elements of eye diseases which are widely acknowledged. Your naked eye is thus considered to just be susceptible enough in the contemporary world to acquire an irrevocable abnormality. With goal of generating smoother Retinal Images shapes and correctly classifying retina pictures, a novel method focused upon edge edge detection with line contours combining were presented throughout [6]. In respect of the its three different approaches, responsiveness, specificity, and accuracy rate, yet again another splitting machine classification technique [7] were brought out now and improved. The preponderance of the computerized ophthalmology automatic detection currently in practice initially segments their basic structure prior subsequently analyzing the clinical examination for vision diagnosis and monitoring. Even so, these measuring device techniques solely rely just on Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 596 https://internationalpubls.com accuracy of separation and neglect a variety of visual features. A machine learning method is described in [8] the with goal of collecting image-relevant knowledge and effectively diagnosing ophthalmology from images obtained. With the this target in mind, a revolutionary Wedge Evaluation Networks (DENet) for automatic ophthalmology detection was developed. This system integrated the localized visual acuity region with worldwide retina image in a deep structural framework. In the interest of recognizing ophthalmology, an evaluation of retinal images and optical cupping segment and classification methods was developed in [9]. [10] Proposed a study on segmentation and categorization towards disease prediction. In attempt to accurately predict forecast the discovery of ophthalmology well before development of sickness, machine learning technique was applied in [11]. In [12], a review of a database employed for disease detection was recommended. In [13], a study and perspective regarding the application of deep learning to the processing of iris image again for detection of ophthalmology been investigated. The identification of the these ophthalmology diseases is complicated because a majority of them exhibit symptoms which pose a challenge to healthcare experts to determine the exact disease for treatments. In classifying diseases for particular therapeutic interventions, an inter analysis and classification methodology is introduced in [14] through enhancing the algorithm for machine learning as well as a Wavelet transform (wavelet transform transform). As a consequence, overall prediction accuracy for disease diagnosis improved. [15] Describes the development of the revolutionary deep learning approach that utilizes a deep residue neural network (ResNet101) again for automatic detection of ophthalmology employing colour fundus images. Furthermore, the method for generating the forecast was designed, as well as the effect of merging retina images and patient’s previous medical information also was considered. [16] examined machine ophthalmology detection methods that investigated at overall efficiency. Thus further, establishing a machine learning method for detecting ophthalmology disease with very little evidence is significant and difficult. Here [17], a machine learning model was deployed to accurately detect ophthalmology employing retinal image imaging and domain expertise. The deep understanding encompassed a variety of important and relevant regions of information for assessment. A cross human brain (MB-NN) algorithm was developed to address this issue the with aim of extracting useful the relevant and significant information using characteristics from subject matter expertise. This was reported that such a level of accuracy was accomplished. Further splitting Autocad program was designed that [18] to retrieve powerful components and simultaneously classifying those into healthy and ophthalmology groups, achieving the highest accuracy conceivable. For identify the essential aspects engaged in identification, an ensembles approach was developed in [19] using class convolution layer. Retina image has been utilized [20] to design a simple linear classifier and convolutional neural network (CNN) algorithm for detecting either mild or severe ophthalmology. Both hypertension identification rapidity and precision have indeed been increased throughout this work due to these issues first by extracting the crucial element using a Wedge shaped Binary Pattern ( lbp Instruction set. Essential features may additionally be generated utilizing Pure sine wave Discontinuous Single - input single Gaussian Feature Engineering. After which, four types of Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 597 https://internationalpubls.com ophthalmology are detected utilizing Inter Regression Hinged Vector Support Ophthalmology Identification. Finally and not last, a thorough analysis also is provided to demonstrate how and where to correctly, rapidly, and effectively detect ophthalmology. 2. Methodology In this part, we go through our research on the use of optical coherence tomography (OCT) images to diagnose hypertension. We had developed a plan for applying deep learning to process and identify the acquired Optical images. Figure 1 shows the image feature testing procedures in use by our suggested method, including includes I processing, (ii) features are extracted from of the preprocessed image, and (iii) ophthalmology detection according to evaluation of the obtained features. Figure 1 Asymmetric Pure sine wave Inter Hinged Supporting Machine Learning (ASM-HSML) approach schematic diagram 2.1 Three different approaches Binary Pattern ( lbp Pre - processor, Pure sine wave Modular Single - input single Probability distribution Edge Detection, and Inter Hinge Vector Support Eye disease Recognition used it to retrieve resampled Imaging techniques in order to identify the eye disease disease by determining postural assumptions. All these is shown in the preceding figure. 2.2 Local Binary Preprocessor based on Angular By discriminating the underlying retinal image against with a creating an enabling environment for statistical evaluation, preprocessing's main objective is to decrease image deviation. In order to highlight the relevant elements, the pre - processing stage eliminates the variations from of the images that aren't connected to the ophthalmology disease. An Wedge shaped Local Derivative Pre - processor is created in our investigation to identify sickness unrelated differences that cannot be primarily connected to ophthalmology. An Wedge shaped Binary Pattern ( lbp Pre - processor model's block design is shown in Figure 2 below. Preprocessing 𝑖𝑛𝑝𝑒𝑑 π‘–π‘šπ‘Žπ‘”π‘’ 1. adjacent pixel coordinate identificati ons 2. local angular position Feature extraction from preprocessed image Sinusoidal Gaussian Feature Extraction Glaucoma detection from feature extracted image Multi-class Hinge Support Vector Glaucoma Detection Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 598 https://internationalpubls.com Figure 2 Wedge shaped Binary Pattern ( lbp Pre - processor architecture diagram This same Wedge shaped Binary Pattern ( lbp Preprocessor model is built as illustrated in the above figure. The initial choice is indeed a "9*9" cubes that contains the pixel value, "P c." The dimensions of a nearby pixels 'P a' are then evaluated for the appropriate angle. If indeed the frequency of the entering image is greater than the value of something like the neighboring pixel, it is exchanged by "1," otherwise with "0." Even though the inter geometric assessment [1] has been adept at collecting the important variety of features, a pretty sizeable group of feature representation are modestly completely eradicated inside the final image representation recognition, regardless of the fact that they begin making up an important feature that could be essential for disease detection. This study aims to develop a Wedge shaped Binary Pattern ( lbp Pre - processor framework that makes the use alignment tilt between image pixels and component angle vectors that vary between '0Β°' through '180Β°' in order to get around this problem. As just an outcome, the coordinate of the adjacent pixels, denoted as "P a," are as continues to follow: "a[0,..., 7]. 𝑖𝑓 0 ≀ πœƒ ≀ 90 π‘‘β„Žπ‘’π‘›, (𝑀𝑝, 𝑁𝑝) = { 𝑀𝑝 = 𝑀𝑐 + 𝛼 𝑁𝑝 = 𝑁𝑐 + [𝑀𝑝 tan(πœƒ)] (1) 𝑖𝑓 90 ≀ πœƒ ≀ 180 π‘‘β„Žπ‘’π‘›, (𝑀𝑝, 𝑁𝑝) = { 𝑁𝑝 = 𝑁𝑐 + 𝛼 𝑀𝑝 = 𝑀𝑐 + [𝑁𝑝 ( 1 tan(πœƒ) )] (2) Represents four rotational orientations, namely wing positioned, correct position, upturn, and down, respectively. This is apparent from the two variables (1) and (2) above. Furthermore, "M c" and "N c" represent again for respective center of a two dimensions "M" as well as "N" of a input retina "Iri." None the less, the gradients of pixel intensities in the "M" and "N" directions of the cube are looked at with the goal of identifying the angular position. The gradients "O" and "M" at every pixels of width "a,b" therefore are evaluated simultaneously. After which, an estimation of the each cube's local rotational orientation, centred at pixels "(M c,N c)," was produced, which is illustrated below. Retinal OCT images Obtain local angular position Foremost feature Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 599 https://internationalpubls.com 𝐿𝑀(𝑀𝑐, 𝑁𝑐) = 2πœ•π‘€(π‘Ž, 𝑏)πœ•π‘(π‘Ž, 𝑏) (3) 𝐿𝑁(𝑀𝑐, 𝑁𝑐) = [πœ•π‘€ 2 (π‘Ž, 𝑏) βˆ’ πœ•π‘ 2 (π‘Ž, 𝑏)] (4) A fundamental characteristic reflecting the rotational dimension "" of a cube centred at image "(M c,N c)" is performed accordingly following using calculations (3) as well as (4) above: πœƒ(𝑀𝑐, 𝑁𝑐) = 1 2 π‘‘π‘Žπ‘›βˆ’1 [ 𝐿𝑁(𝑀𝑐,𝑁𝑐) 𝐿𝑀(𝑀𝑐,𝑁𝑐) ] (5) By removing the anomalies that are just not connected to ophthalmology using equation (5) previously, the main feature is obtained. Now is presented masquerading representations of the local image preprocessor for Angular. Input: Input Retinal Image β€˜π‘…πΌ = 𝐼1, 𝐼2, … , 𝐼𝑛’, β€˜π›Όβ€™ Output: Prominent feature β€˜πΉ = 𝑓1, 𝑓2, … , 𝑓𝑛’ Step 1: Direction of Initialize pixel angular directions β€˜πœƒβ€™ Step 3: For Each Individual Retinal Image β€˜π‘…πΌβ€™ Step 4: Select β€˜9 βˆ— 9’ cube encircling the center pixel, β€˜π‘ƒπ‘β€™ Step 5: Evaluate coordinates of the adjacent pixels β€˜π‘ƒπ‘Ž=𝑃𝑓𝑝, 𝑃𝑓𝑝+1, π‘ƒπ‘“π‘βˆ’1, π‘šπ‘ƒπ‘“π‘, π‘š + 1𝑃𝑓𝑝’ Step 6: For each coordinates Step 7: Obtain coordinates of the adjacent pixels β€˜π‘ƒπ‘Žβ€™ using (1) and (2) Step 8: Evaluate local angular position of each cube centered at pixel β€˜(𝑀𝑐, 𝑁𝑐)’ using (3) and (4) Step 9: Evaluate foremost feature representing the angular direction β€˜πœƒβ€™ using (5) Step 10: Return foremost feature β€˜πΉβ€™ Step 11: End for Step 12: End for Step 13: End Algorithm 1 Local Binary Preprocessor based on Angular The objective of the algorithm remains to efficiently get the central component required for ophthalmology diagnosis for every retina image given as input, like mentioned in the previous Wedge shaped Binary Pattern ( lbp Preprocessor methodology. This one is performed by obtaining the regional angular displacement as well as the dimensions of something like the pixel that really are close the relevant centre pixels. The main attribute required for ophthalmology detection may be obtained with all these two position data in the shortest possible time. 2.3 Extraction of Sinusoidal Discrete Cosine-based Gaussian Features Despite pure image pixels are able to serve as an image feature, they really aren't adequately descriptive for examples beyond the smallest ones, particularly when working with retina optimal scanning images. The objective is to generate an image using unique traits or characteristics that Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 600 https://internationalpubls.com distinguish the observation through one class from some other (i.e. detected with glaucoma or not). Thus the, extraction of features is a critical step whenever implementing deep learning to retinal images in order to create properties that can differentiate among prominent and ou alors information. In our research, the primary objective obtained using a Wedge shaped Binary Pattern ( lbp Preprocessor is subject to a Stochastic Discrete Extracting Features. The schematic diagram again for Harmonic oscillator Discontinuous Single - input single Stochastic Extracting Features model can be seen in Figure 3. Figure 3 Block diagram of Sinusoidal Discrete Cosine-based Gaussian Feature Extraction model In accordance with the above figure, each image's number of pixels, including includes the primary characteristic "F" with length "A*B," will be first transformed into the sinusoidal discontinuous trigonometric in the manner illustrated below. π·πΆπ‘π‘ž = βˆ‘ βˆ‘ πΉπ‘Žπ‘πΆπ‘œπ‘  ( πœ‹(2π‘Ž+1)𝑝 2𝐴 ) πΆπ‘œπ‘  ( πœ‹(2𝑏+1)π‘ž 2𝐡 )π΅βˆ’1 𝑏=1 π΄βˆ’1 π‘Ž=1 (6) The discontinuous sine 'Direct - current' of magnitude 'Hence both' each pixels 'p,q' from equations (6) previously performs out the modification through the use of a requires new approaches again for corresponding fundamental characteristic 'F'. The feature extraction method employing localized stage quantized [2] plus image distorting is really quite demanding because it's extremely sensitive to alterations in significance level. This work introduces a Stochastic Cumulative distribution function that estimates the positions that are shown below in order to fix this problem. 𝐹𝐸 = 𝑓(𝑅𝑉| 𝐢𝑁 , 𝐢𝑆, πΆπ‘ β„Žπ‘Žπ‘π‘’) = 𝑃𝑁 exp(βˆ’(𝑃𝑆|𝑅𝑉 βˆ’ πœ‡|)) (7) Thus according expression (7) previously, for an average value of "," "RV" represents again for stochastic process, "C N" for normalization criteria, "C S" as trying to scale requirements, as well as "C shape" for shaping requirements. The assessment of the normalization and scalability criteria is presented below. 𝐢𝑁 = 𝐢𝑆,πΆπ‘ β„Žπ‘Žπ‘π‘’ 2(πΆπ‘ β„Žπ‘Žπ‘π‘’) (8) 𝐢𝑆 = 1 𝜎 √ (3 πΆπ‘ β„Žπ‘Žπ‘π‘’ ⁄ ) (1 πΆπ‘ β„Žπ‘Žπ‘π‘’ ⁄ ) (9) Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 601 https://internationalpubls.com Depending just on scalability criterions "C S" and shaped evaluation metrics "C shape," accordingly, the normalization criterions "C N" and scalability criterions "C S" are obtained from the preceding solutions (8) and (9). To every cubes of a Discrete cosine parts that have the most significant characteristics or even the corresponding OCT images of the retina of the eye, the Stochastic distribution function is used. The accompanying is the notional coding for Pure sine wave Discrete Single - input single Stochastic Extraction Of features. Input: Size β€˜π΄ βˆ— 𝐡’, Pixel β€˜π‘, π‘žβ€™, Output: Salient feature extraction β€˜π‘† = 𝑠1, 𝑠2, … , 𝑠𝑛’ Step 1: Initialize Foremost feature β€˜πΉ = 𝑓1, 𝑓2, … , 𝑓𝑛’, Step 2: Begin Step 3: For each Foremost feature β€˜πΉβ€™ Step 4: Perform discrete cosine transform using (6) Step 5: Evaluate positional estimates via Gaussian density function using (7) Step 6: Return salient feature extracted β€˜π‘†β€™ Step 7: End for Step 8: End Algorithm 2 Stochastic Extracting Features model based around pure sine wave Discrete Cosines As mentioned in the above- mentioned Pure sine wave Compact Cosine-based Probability distribution Feature Extraction approach, the objective is to obtain the key characteristics required for disease diagnosis with greatest sensitivity possible for every top extract features using Wedge shaped Binary Pattern ( lbp Preprocessor as insight. Through the use of a harmonic discontinuous cosine and a stochastic density distribution, this is achieved. The sensitivities is believed to be enhanced by the fact that both of these function have had the benefit of having probabilistic attribute values that really can change in reaction to retina image abnormalities. 1.1 Multi-class Hinge Support Vector Glaucoma Detection At last, to use an Inter Vector Support Ophthalmology Detection algorithm, premature hypertension identification is carried out in our research once the salient characteristic has been retrieved. The principle component analysis is used as a classifier through [2], but by using just a few of the principal components while ignoring the rest leads to loss. As just a result, an inter linear pivot vector support ocular hypertension detection technique is being used, that also through using different classes, portrays a clear objective function (early eye disease detectors) for having trained only those 'n' bitwise Support vector machine ( svm. As a result, there's going to be a huge reduction in the losses in disease prediction. Let's take a glance just at collection all 'n' elements that must be detected, '(s i,t i), whereby typically begin, 2,... n. Here,'s i' represents again for data input vectors while 't i' represents again for final prediction to which's i' belong. Let's account for the data for training, that we'll refer to it as "s i" or Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 602 https://internationalpubls.com "s i (n)" where "s i (k)" refers for the output retina imagenet dataset to a "closest k" SVM module. Consequently, π‘˜ = (π‘π‘’π‘šπ‘π‘’π‘Ÿ π‘œπ‘“ π‘ π‘‘π‘Žπ‘”π‘’π‘  βˆ’ 1) (10) 𝑠𝑖(0) = 𝑠𝑖 (11) 𝑠𝑖(1) = 𝑠𝑖 βˆ’ π‘π‘™π‘Žπ‘ π‘ (0) (12) 𝑠𝑖(2) = 𝑠𝑖 βˆ’ [π‘π‘™π‘Žπ‘ π‘ (0) + π‘π‘™π‘Žπ‘ π‘ (1)] (13) The approach is carried out because shown in just using salient feature of a retinal image from the above input that's been obtained to a 'closest k' SVM cubes. 𝑀𝑖𝑛 1 2 π‘Šπ‘˜ π‘‡π‘Šπ‘˜ + 𝐢 βˆ‘ 𝐿𝑖(π‘˜)𝑛 𝑖=1 (14) 𝑆𝑒𝑏𝑗𝑒𝑐𝑑 π‘‘π‘œ 𝑑𝑖(π‘Šπ‘˜ 𝑇𝑠𝑖(π‘˜)) + 𝑏 β‰₯ 1 βˆ’ 𝐿𝑖 (15) The modest weight transpose scalar and support vectors for optimizing with such a continuous 'C' and hinged degradation function 'L i' again for appropriate 'substring' SVM cube were given by 'W kT' and 'W k' in the calculations (14) and (15) respectively. Finally, this hinged algorithm is mathematically stated as follows, wherein t is the ophthalmology detection scores and t seems to be the intended result. The hinged loss of something like the glaucoma detection is stated as follows. 𝐿(𝑦) = max(0,1 βˆ’ 𝑑. 𝑦) (16) Thereby, "y=W(s i)+b,(W,b)" denotes the hyper - plane characteristics with "s i" denoting the incoming retina image for disease detection from formula (16) above. This preceding is the pseudo coding for Inter Nonlinear Hinged Vector Support Disease Diagnosis. Input: Salient feature extraction β€˜π‘†β€™ Output: Accurate and early glaucoma detection Step 1: Initialize β€˜π‘˜β€™ Step 2: Begin Step 3: For each Salient feature extraction β€˜π‘†β€™ Step 4: Equate input retinal image dataset to the β€˜π‘˜π‘‘β„Žβ€™ SVM cube using (11), (12) and (13) Step 5: Perform optimal glaucoma detection using (14) and (15) Step 6: Evaluate hinge loss function using (16) Step 7: If β€˜π‘Š0(π‘˜ = 0)’ then weight vector for first stage detected with normal Step 8: End if Step 9: If β€˜π‘Š0(π‘˜ = 1)’ then weight vector for second stage detected with glaucoma Step 10: If β€˜π‘Š1(π‘˜ = 1)’ then weight vector for first stage detected with CNV Step 11: If β€˜π‘Š1(π‘˜ = 2)’ then weight vector for second stage detected with DME Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 603 https://internationalpubls.com Step 12: If β€˜π‘Š1(π‘˜ = 3)’ then weight vector for third stage detected with DRUSEN Step 13: Return Step 14: End for Step 15: End Algorithm 3 Ophthalmology Identification Employing Inter Linear Hinged Using Support Vector Machines 2. The purpose remains in premature ophthalmology diagnosis with minimal error, as mentioned in the Inter Linear Hinged Gradient Boosting Disease Diagnosis section. This one is accomplished in our instances through the use of a benefits including improved and bending transfer functions, both of which offer the advantage of concurrently learning both 'n' bidirectional Svm classifier, that maximises the margin out of each class for all other categories. None the less, to use the hinges error function provides better accuracy but at a lower sensitivities. 3. Experimental settings The Retina Coherence Optic High - resolution computed (OCT) images that really are made accessible to the general population are produced to use the methodologies that have recently been presented. As an imaging technique, retina confocal laser tomography (OCT) gathers elevated cross - section from of the retina of the eye of patients that remain functioning. The data, which contains subdirectories for every image classification, has indeed been organized into three directories: testing ”, validation accuracy, and test dataset (NORMAL, CNV, DME, DRUSEN). The confocal laser high - resolution computed (OCT) images were taken between July 1, 2013, as well as March 1, 2017, from retroactive collaborators of older patients just at Shenzhen First Women's Clinic, Shanghai Make the correct Eye Hub, California Dermoscopic Research Center, Medical Clinic Ophthalmology Collaborators, and Shiley Eye Research center of the University of California, San Francisco. Matlab was employed to conduct various experiments. The above largely true aims to contrast the pro - posed Geometric Pure sine wave Inter Hinge Support Device Learner (Assembler) technique for disease prediction against the already established Multi Different Feature Deep Convolutional neural Network (MFV-DBN) [1] as well as Contrast Restricted Accelerated Histogram Equalization (CLAHE) [2. The sensitivities with regard to different number of images and the computational duration for disease detection are the parameters that are being investigated. 4. Results and Discussion This section evaluates statistical parameters diagnosis duration, ophthalmology high detection, and responsiveness evaluate the efficacy of disease diagnosis. The same dataframe and similar sample image data were employed to conduct a reasonable evaluation of the suggested Angular Pure sine wave Inter Hinge Support Device Learner (Embedded systems) for disease diagnosis against by the established Inter Different Feature Deep Convolutional neural Internet backbone - Mother to the fetus [1] and Algorithms [1] and Comparison Restricted Accelerated Analytic Hierarchy process ( ahp - CLAHE [2]. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 604 https://internationalpubls.com 4.1 Glaucoma Detection Time In order to maximise treatment plans advantages and minimize financial burden, early detection of ophthalmology is essential. The ophthalmology detection limit is the initial measurement to investigate retinal diagnosis. The faster ophthalmology is diagnosed, the earlier preventive measures can be taken. The quicker the diagnosis time, the mathematical formulation of the ophthalmology detection time appears as follows. πΊπ·π‘‘π‘–π‘šπ‘’ = βˆ‘ 𝑅𝐼𝑖 βˆ— π‘‡π‘–π‘šπ‘’ [𝐺𝐷]𝑛 𝑖=1 (17) Thus according equations (17) above, that time period required to recognize ophthalmology based on the hinge error function is measured as "Length [GD]" and is predicated on the example retina image provided for screening, "RI i." Average ophthalmology identification rates again for three techniques, Mother to the fetus [1] and Convolution [2]β€”are shown in Table 1 beneath. Table 1 ASM-HSNL, MFV-DBN, and CLAHE are tabulated. Sample retinal image Glaucoma detection time (ms) ASM-HSML MFV-DBN CLAHE 500 82.5 95 102.5 1000 102.35 125.55 155.15 1500 135.55 160.75 195.35 2000 150.15 180.25 225.55 2500 165.35 195.55 245.55 3000 190.25 225.35 290 3500 220.55 240.55 310.55 4000 245.55 295.55 335.55 4500 290 345.55 390.15 5000 325.55 360 410.25 Figure 4 Time to diagnose glaucoma in comparison to sample retinal images The geometric representation of ophthalmology diagnosis over time with regard to 5000 different sample retina images obtained from a dataset of retina Mri images collected at multiple periods interval can be seen in Figure 4 above. As shown in the picture, the example retinal images for use as following characteristics impact the speed at which glaucoma is detected. In those other words, whenever the dataset's three directories (labeled training, test dataset, and testing set) and four subdirectories (Regular, Nvr, Mdr, and Combination of chemical) are extended by adding additional example retinal images, the amount of time needed to detect ophthalmology concurrently increases. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 605 https://internationalpubls.com Even so, for just a sample test containing "500" Opt images, ophthalmology identification was demonstrated to require "0.190ms" for Assembler, "0.205ms" with Mother to the fetus, and "0.165ms" for Original image, while the overall diagnostic required "82.5ms," "95ms," and "102.5ms" respectively. From results, it may be determined that now the Assembler product's ophthalmology detection limit is considerably quicker compared to [1] and [2]. The implementation of a Wedge shaped Binary Pattern ( lbp Pre - processor model is what produced the optimization. Through this methodology, semi illness dependent inconsistencies first were eliminated, and afterwards relevant features are extracted by acquiring local angular displacement. The main feature required for ophthalmology detection is generated with any of these two obtained positioning measurements employing the Assembler methodology in the shortest period of time, which is reported to be reduced by 15percentage points compared to [1] and 29% comparing to [2]. Efficiency in ophthalmology diagnosis evaluation metrics Estimations of diagnostic accuracy are quite vulnerable to hypertension system to detect configuration. .however, a lot more analytical problems contribute either to overstatement or overestimate, which decreases the probability that now the results would've been applicable. The algorithm below may be employed to determine the effect of glaucoma detecting design on estimates of diagnostic performance. πΊπ·π‘Žπ‘π‘ = βˆ‘ 𝑅𝐼𝐢𝐷 𝑅𝐼𝑖 βˆ— 100𝑛 𝑖=1 (18) Thus according expression (18), the reliability of ophthalmology prediction is evaluated by classifying correctly identified retinal blood vessels (RI CD) as Healthy, Agc, DME, and Combination of chemical according to testing retina source images (RI i) employed for modeling. The reliability of disease prediction employing the three different approaches, Assembler, Difference between two points [1], and Convolution [2], is shown in Table 2 below. Table 2 ASM-HSNL, MFV-DBN, and CLAHE are tabulated. Sample retinal image Glaucoma detection accuracy (%) ASM-HSML MFV-DBN CLAHE 500 96 94 91 1000 94.55 91.85 89.15 1500 91.35 89.15 88.55 2000 90.85 88.35 86.25 2500 89.35 87.15 85.45 3000 89 85 83.15 3500 88.45 84.35 82 4000 88 82.15 81.15 4500 87.35 81.55 80 500 87.15 80 79.55 Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 606 https://internationalpubls.com Figure 5 Accuracy of glaucoma detection compared to a sample retinal picture In accordance with the experimental sample retina images, which have been separated into four categories and categorized as being either diseased as well as healthy using a randomization individual Identification, Figure 5 above illustrates the correctness of ophthalmology detection. The efficiency of ophthalmology detection is inversely correlated with both the sample retinal images, as shown in the figure. In those other terms, raising the total amount of example retina pictures leads in a rise in the amount of abnormalities therefore lowers the overall accuracy. Regrettably, the as a whole eye disease detection rate has been found to just be 96%, 94%, and 91% including both for different sampling experiments done with "500" number system of Image data along with all four different categories and "480" data points of laboratory testing rectified detecting in accordance with the category utilising Assembler, "470" statistics of test cases rectified trying to detect in accordance with the category using Difference between two points [1], and "455" numbers of test cases tried to correct designed to detect in accordance with the classification using CLAHE [2]. From this result, it can be determined that Assembler HSNL's detection performance is comparative to that of [1] and [2]. The implementation of the Pure sine wave Discontinuous Single - input single Gaussian Extracting Features method is the source of the improvement. Through this method, it was reported that the algorithms harmonic discontinuous cosine and Stochastic concentration were employed in order to produce effective prominent extraction of features. This resulted observed statistically extracted features shifting in reaction to abnormalities in iris image, improving the Assembler HSNL's detection performance approximately 5% comparing to [1] and 7% comparable to [2]. 4.2 Performance measure of sensitivity Responsiveness is a quantitative word that refers to very well how several category classifications procedures work and therefore is particularly employed to the disease diagnosis. In other respects, sensitivity is the proportion of positive instances which are properly identified (for instance, the proportion of individuals who would be correctly given the diagnosis of ophthalmology without really have vision). Sensitive in glaucoma screening testing is an assessment that indicates whether the test can spot real positive results. To put it another way, a test with a highly sensitive value properly classifies a specimen retinal vision alone without precondition and positive frequently compared to a diagnostic with such a reduced sensitivity. The below is the manner in which this is logically represented. 𝑆𝑒𝑛𝑠𝑖𝑑𝑖𝑣𝑖𝑑𝑦 = βˆ‘ 𝑇𝑃 πΊπ·π‘ π‘Žπ‘šπ‘π‘™π‘’π‘  βˆ— 100𝑛 𝑖=1 (19) Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 607 https://internationalpubls.com The percentage of sensitivities, "Responsiveness," is determined using equation (19) above based on the positive instances, "TP," as well as the real ophthalmology diagnosed instances, "GD samples." It is expressed in percentages (%) terms. All three methodologies' susceptibility is shown in Table 3 below: Assembler, Inclusion [1] as well as CLAHE [2]. Table 3 ASM-HSNL, MFV-DBN, and CLAHE are tabulated. Sample retinal image Sensitivity (%) ASM-HSML MFV-DBN CLAHE 500 86.36 81.81 77.27 1000 85.25 80.25 75.35 1500 84.15 80 75.15 2000 83.25 79.15 75 2500 82.85 79 74.75 3000 81 77.85 74.35 3500 80.45 77.35 74.15 4000 80.25 77 74 4500 80 76.15 73.25 5000 79.15 75.25 73 Figure 6 sample retinal images versus Sensitivity Finally and not most, figure 6 above shows the responsiveness rate in relationship to 5,000 samples retinal images taken by different patients at distinct periods throughout the course of 2017. The graph suggests that when numerous sample mri scans are used, the quality measures falls. Even so, the overarching frequency response rate was discovered to be '86.36%', '81.81%', as well as '77.27%' including both with '500' number of inhabitants regarded for checking and '110' individuals who were genuinely discovered to eye problems and '95' becoming the rate of true positives using ASM- HSNL, '90' becoming the rate of true positives using MFV-DBN [1], and '85' being the rate of true positives utilising CLAHE [2]. The application of the Inter Linear Hinged Vector Support Ophthalmology Detection algorithm is just what generated the improvements. In recognition of the benefits of concurrently training numerous Svm classifier, the technique is applicable two main functions stochastic optimization and hinges losses to the significant features that were already collected. As an outcome, the tolerances from each method to the other class are increased, which improves sensitivity by 5% comparing to [1] as well as 10% comparable to [2] when employing ASM-HSNL. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 608 https://internationalpubls.com Conclusion In the investigation, the evaluation of ophthalmology detection through retina Optic Coherent Ultrasonography (OCT) images has indeed been proposed using Angular Sine wave Mega Hinged Supporting Deep Learning (ASM-HSML). To identify neighboring pixels values and local directional locations, the OCT images are first normalised using a Wedge shaped Binary Pattern ( lbp Preprocessor. The Pure sine wave Discrete Single - input single Probabilistic Extracting Features model would then be given the heavily processed foremost input features. Next, to use a harmonic discontinuous cosine and a stochastic distribution function, characteristics are extracted in this instance. At last, the Inter Linear Hinged Vector Support Ophthalmology Detection models inputs is constructed to use the salient features extracted. Here, this weighing scale is adopted as just significant criteria for differentiating the four categories of Healthy, Nvr, DME, and Combination of chemical. In term of ophthalmology detection rate, correctness, and specificity, the proposed ASM- HSML method achieved superior results in the experiment conducted. In contrast to prior investigations, simulation results employing retina Images reveal superior perspective in the context of the three parameters. References [1] F. Ajesh, R. Ravi, G. Rajakumar, β€œEarly diagnosis of glaucoma using multi‑feature analysis and DBN based classification”, Journal of Ambient Intelligence and Humanized Computing, Springer, Feb 2020 (Multi Feature Vector Deep Belief Network – MFV-DBN) [2] Chan Yam Meng, E.Y.K. Ng, V. Jahmunah, Joel En Wei Koh, Oh Shu Lih, Leonard Yip Wei Leon, U. Rajendra Acharya, β€œAutomated detection of glaucoma using optical coherence tomography angiogram images”, Computers in Biology and Medicine, Elsevier, Oct 2019 (Contrast Limited Advanced Histogram Equalization – CLAHE) [3] Mamta Juneja, Shaswat Singh, Naman Agarwal, Shivank Bali, Shubham Gupta, Niharika Thakur, Prashant Jindal, β€œAutomated detection of Glaucoma using deep learning convolution network (G-net)”, Multimedia Tools and Applications, Springer Nature, Mar 2019 [4] Andres Diaz‑Pinto, Sandra Morales, Valery Naranjo, Thomas KΓΆhler, Jose M. Mossi, Amparo Navea, β€œCNNs for automatic glaucoma assessment using fundus images: an extensive validation”, Bio Medical Engineering OnLine, Jul 2019 [5] Lamiaa Abdel-Hamid, β€œGlaucoma Detection from Retinal Images Using Statistical and Textural Wavelet Features”, Journal of Digital Imaging, Springer, Feb 2019 [6] Soorya M., Ashish Issac, Malay Kishore Dutta, β€œAn automated and robust image processing algorithm for glaucoma diagnosis from fundus images using novel blood vessel tracking and bend point detection”, International Journal of Medical Informatics, Elsevier, Feb 2018 [7] Shruti Jain, Ayodeji Olalekan Salau, β€œDetection of glaucoma using two dimensional tensor empirical wavelet transform”, Springer Nature, Oct 2019 [8] Huazhu Fu, Jun Cheng, Yanwu Xu, Changqing Zhang, Damon Wing Kee Wong, Jiang Liu, and Xiaochun Cao,” Disc-aware Ensemble Network for Glaucoma Screening from Fundus Image”, IEEE Transactions on Medical Imaging, Feb 2018 [9] H N Veena, A Muruganandham, T Senthil Kumaran, β€œA Review on the optic disc and optic cup segmentation and classification approaches over retinal fundus images for detection of glaucoma”, Springer Nature, Aug 2020 [10] Niharika Thakur, Mamta Juneja, β€œSurvey on segmentation and classification approaches of optic cupand optic disc for diagnosis of glaucoma”, Biomedical Signal Processing and Control, Elsevier, Jan 2018 [11] Anshul Thakur, Michael Goldbaum, Siamak Yousefi, β€œPredicting Glaucoma before Onset Using Deep Learning”, American Academy of Ophthalmology, Elsevier, Jul 2020 [12] Muhammad Mateen, Junhao Wen, Mehdi Hassan, Nasrullah, Song Sun, Shaukat Hayat, β€œAutomatic Detection of Diabetic Retinopathy: A Review on Datasets, Methods and Evaluation Metrics”, IEEE Access, Mar 2020 Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 2 (2025) 609 https://internationalpubls.com [13] Daniele M. S. Barros, Julio C. C. Moura, Cefas R. Freire, Alexandre C. Taleb, Ricardo A. M. Valentim and Philippi S. G. Morais1, β€œMachine learning applied to retinal image processing for glaucoma detection: review and perspective”, BioMedical Engineering OnLine, Jul 2020 [14] F. Ajesh, R. Ravi, G. Rajakumar, β€œEarly diagnosis of glaucoma using multi‑feature analysis and DBN based classification”, Journal of Ambient Intelligence and Humanized Computing, Springer Nature, Feb 2020 [15] Feng Li, Lei Yan, Yuguang Wang, Jianxun Shi, Hua Chen, Xuedian Zhang, Minshan Jiang, Zhizheng Wu, Kaiqian Zhou, β€œDeep learning-based automated detection of glaucomatous optic neuropathy on color fundus photographs”, Graefe's Archive for Clinical and Experimental Ophthalmology, Springer Nature, Jan 2020 [16] Yuki Hagiwara, Joel En Wei Koh, Jen Hong Tan, Sulatha V. Bhandary, Augustinus Laude, Edward J. Ciaccio, Louis Tong, U. Rajendra Acharya, β€œ Computer-aided diagnosis of glaucoma using fundus images: A review”, Computer Methods and Programs in Biomedicine, Elsevier, Jul 2018 [17] Yidong Chai, Hongyan Liu, Jie Xu, β€œGlaucoma Diagnosis Based on Both Hidden Features and Domain Knowledge through Deep Learning Models”, Knowledge-Based Systems, Elsevier, Jul 2018 [18] U Raghavendra, Hamido Fujita, Sulatha V Bhandary, Anjan Gudigar, Jen Hong Tan , U Rajendra Acharya, β€œDeep Convolution Neural Network for Accurate Diagnosis of Glaucoma Using Digital Fundus Images”, Information Sciences, Elsevier, Jan 2018 [19] Yu-Chieh Ko, Shih-Yu Wey, Wei-Ta Chen, Yu-Fan Chang, Mei-Ju Chen, Shih-Hwa Chiou, Catherine Jui-Ling Liu, Chen-Yi Lee, β€œDeep learning assisted detection of glaucomatous optic neuropathy and potential designs for a generalizable model”, PLOS ONE | https://doi.org/10.1371/journal.pone.0233079 May 14, 2020 [20] Jin Mo Ahn, Sangsoo Kim, Kwang-Sung Ahn, Sung-Hoon Cho, Kwan Bok Lee, Ungsoo Samuel Kim, β€œA deep learning model for the detection of both advanced and early glaucoma using fundus photography”, PLOS ONE | https://doi.org/10.1371/journal.pone.0207982 November 27, 2018