Field of Invention
Artificial intelligence (AI), the Internet of Things (IoT), and cloud computing have all lately gained popularity in the healthcare industry, allowing radiologists to make better decisions. PET imaging, also known as positron emission tomography, is one of the most reliable ways for a radiologist to diagnose a variety of malignancies, including lung tumours. We presented stage categorization of lung tumours in this paper, which is a more difficult issue in computer-aided diagnosis. As a result, a modified computer-aided diagnosis is being investigated as a solution to relieve radiologists' onerous workloads and provide second opinions. This invention includes a method for identifying and verifying distinct phases of lung tumour advancement, as well as a deep neural model and data gathering utilizing a cloud system for categorizing lung tumour progression. Many individuals throughout the world were recently hit by the COVID-19 epidemic. To eliminate physical personalising and disease propagation, the contemporary situation has turned the whole healthcare sector into electronic diagnostic, remote access, virtual consulting, and E-treatment. Due to a scarcity of physicians, carers, and radiologists, this pandemic has brought the healthcare business to a new level of severity, increasing the death rate of chronic illness patients, notably those with cancer, diabetes, and heart ailments.

Background of Invention
Lung cancer is one of the leading causes of cancer-related mortality. Early identification of lung cancer can greatly improve the chances of survival. Radiologists must manually delineate lung nodules, which is a time-consuming operation. For supporting radiologists, a novel computer-aided decision support system for lung nodule diagnosis based on 3D Deep Convolutional Neural Network (3DDCNN) was created. Our decision assistance technology gives radiologists a second opinion when making lung cancer diagnostic decisions. Median intensity projection and the multi-Region Proposal Network (mRPN) for automated identification of possible region-of-interests to harness 3-dimensional information from CT images have been used. This research demonstrates the potential of deep learning combined with cloud computing for accurate and efficient lung nodule diagnosis using CT imaging, which might aid doctors and radiologists in the treatment of lung cancer patients. Pulmonary cancer is an uncontrollable aberrant lung cell development known as nodules, whose early discovery is critical for successful disease progression management and therefore potentially increasing the patient's life rate. On high-resolution and high-quality chest radiographs, radiologists commonly utilise manual lung nodule delineation. CT scanning is a difficult, time-consuming, and tiresome procedure.

Object of Invention
Dataset Description
♦ LIDC-IDRI Dataset are utilized for pre-training and testing purpose.
<♦ The proposed E-CNN model is trained on 650 Cases and tested on 368 Cases. Software
♦ The proposed models is performed in MATLAB2018b.
♦ LC-Cloud Server is accessed as private AWS server Hardware

♦ The E-CNN is developed and validated in the hardware PC with i7 processor with NVIDIA JETSON GPU System-on-module with 256-core NVIDIA Pascal™ GPU architecture with 256 NVIDIA CUDA cores, and memory 64 GB 128-bit LPDDR4 Memory 1866 MHz- 59.7 GB/s
♦ The performance test was performed using a Dell R730 server (2-way E5-2603 v3, 1.6 GHz, 6-core, 48 GB RAM). The terminals were personal computers (Dell 3020, dual-core 3.5 GHz, 4 GB RAM), Android(based tablet computers (Samsung Note 8), smartphones with iOS (iPhone 6), and iOS-based tablet computers (iPad mini 2).
♦ The network bandwidth was set to 7-Mbps to simulate low bandwidth in the regional network.

Summary of Invention
The World Health Organization (WHO) recently reported that the lung tumor was the leading cause of death worldwide. Here, a practical computer-aided diagnosis (CAD) system is developed to increase a patient's chance of survival. Segmentation is acritical analysis tool for dividing a lung image into several sub-regions. This work characterized an automated 3-D lung segmentation tool modeled by an active contour model for computed tomography (CT) images and PET-Scans. The proposed segmentation model is used to integrate the local image bias field formulation with the active contour model (ACM). Here, a local energy term is specified by using the mean squared error to reconcile severely in homogeneous CT images and used to detect and segment tumor regions efficiently with intensity inhomogeneity. In addition, a Multiscale Gaussian distribution was applied to the CT images for smoothening the evolution process, and features were determined. We used the Lung Image Database Consortium (LIDC-IDRI) data set that consisted of 850 lung nodule-lesion images that were segmented and refined to generate accurate 3D lesions of lung tumor CT images. Tumor portions were extracted with 97% accuracy. Using continuous feature extraction of 3-D images leads to attribute the deformation and quantifies the centroid displacement. In this work, we predicted the centroid displacement and contour points by a curve evolution method which results in more accurate predictions of contour changes and than the extracted images were classified using an Enhanced Convolutional Neural Network (CNN) Classifier. The experimental result shows that the modified Computer Aided Diagnosis (CAD) system has a high ability to acquire good accuracy and assures automated diagnosis of a lung tumor.The novelty of work are as follows:
♦ An Automated 3D-Lung Cancer Detector Prototype called "Cloud-based 3D-Lung Cancer Segmentation and Detection (CB-LCSD)" that include proposed ACM segmentation and ECNN detection model with cloud-prototype design that minimizes the overall death rate.
♦ A novel two-stage classifier developed for detecting the lung-tumor severity level at its early state to prolong the lifespan of lung-cancer patients.
♦ The proposed framework performs both segmentation and detection technique using pre-processed segmented images.
♦ Cloud-LTDSC framework is capable to detect lung-tumor nodules using both standard CT-Scans and PET-Scans.

♦ The proposed prototype design increase the e-treatment diagnosis system and provides the second-opinion for patients in an automated way.
♦ Reduces the radiologist's diagnosis time with this smart detector.
♦ Patients Report are automatically stored in the proposed LC-Cloud server for clinical study and E-treatment.

Detailed Description of Invention
The proposed framework includes segmentation and detection automated deep learning module for lung-cancer.
LT-Segmentation Module
In compared to other organs such as the brain, .detecting lung nodule in a Dicom CT lung imaging is tough. Likely, bronchus and intensity of lung nodule identification in a CT Dicom lung image are tedious. The bronchus and intensities of blood vessels within the tumour area are then measured, and the region of lung parenchyma is partitioned from back to front. The steps involved for the detection of lung tumor are as follows:
1. Remove the region of mediastinum and thoracic wall if the lung parenchyma was not reconstructed.
2. The tumour part of the lung picture is segmented using the Active Contour Model (ACM)
3. A total of 925 lesions from different patients were collected, together with nodule information.
Due to a gradient value, the edge-based ACM causes some segmentation modifications. The leaking problem occurs when there is a weak boundary in the image. The approach of edge-based segmentation is not possible due to noise in the image border. Border leakage does not occur in region-based ACM, and undesired sections of the picture are removed. When the roundness rule was applied to the objects, the result was approximately 1, resulting in a solid nodule. Snake's model was utilized to develop a curve that aids in the detection of a tumor portion in the relevant photos. The curve must be drawn around the identified item, and then it must alter its location towards the interior and end at the object borders.For each contour line that gather a similar amount of control point and distinct point, the initial prediction model of active contour is utilized; it predict contour point that helps for the contour lines and then generates the active contour model's first prediction.The proposed approach generates 3D features from 2D stochastic characteristics, which are then fed into the CNN classifier.
Consider Gray-scale Dicom image [ (y,z) € R2 ], Contour of the segmented portion was mentioned as:

Ck(s) = [yfc(s).Zir(s)]r.{s G (0,1)] (1)
In the proposed ACM, we reduced the number of points to fit the curve in the tumor portion which is segmented. Let the Sample contour Ct(s) into the given n points and the overall curve Uk was given by,
"lr = {ClU'CM CkJ (2)
The modified ACM is evaluated using gradient value which helps for an edge detection. It is derived from the image as,
Ek(C)=ocf*\C'(s)\2ds+ f!fi\C"(s)\ds~ l*f*Wfo(C(y))\2dy (3)
Such as a, p and u are constant and Ek refers to the parametric curve with kth image.
Using Mumford-Shah model, we derived an expression for an intensity outside and inside curves with level set method and Energy approximation AE is given by,
^('i,/2,e»= 4 (/(y)-/1)2/a(0)dy + /n (rty)-i2)2/.(-0)dy+r Jn /.(«*y +
fila S(0)\V(0)\dy
(4) j
Using Euler-Lagrange equation, inside and outside curve intensities II and 12 is expressed as,
i /2(0) = f" nyMi-jMy^y (6)
These expressions (5) and (6) gives the weight calculation which is given for inside and outside curve intensities II and 12 is derived by kernel function f(y) decreased. It becomes constant and represented average intensities to be determined.

LT- Detection Module
♦ E-CNN is the multilayer architecture which consists of 6 layers of CNN.
♦ E-CNN estimates the exact or neighboring classes of images with less than 5% error.
♦ The layers of E-CNN are 3 convolutional layers, followed by fully connected layers which have a Rectified Linear Unit (Relu), normalization and dropout layers.
♦ The dataset was trained in E-CNN technique, while this validation data was helpful for fine - tuning parameter results.
♦ The patch is extracted with the size of 50x50 mm in classification, which is resized to a size of 64*64 pixels with a resolution of 0.69 mm from tumor portion.
♦ Data augmentation is classified as test-data augmentation (TDA) and train-data augmentation (TRDA).
♦ The LIDC dataset was trained by pointing the lesions by 1mm and patch by 40, 45, 50 and 55 mm.
♦ If the nodule is more than 3 mm, it will be captured as the patches, TDA improves the patches by rescaling the trained data.
♦ In the training dataset, 5-fold cross-validation for evaluating the LIDC dataset is used.
*> 850 lesions were divided into five subsets, one validation and testing subset and three subsets for training.
LC-Cloud Server
The LC-Cloud server architecture comprises of three layers called "input, processing and storage layer" which is connected with firewall and LC-server.The input layer designed to adopt any category of lung tumor data such as "dicomimages and PET images" directly accessible through users or radiologists through internet. These data are feed-forwarded into next layer for processing under certain per-defined conditions.The new user data is checked with the LC-Cloud engine for availability match. The conditions are (i) if the new user record is matched with existing file then directly forwarded to physicians and (ii) if the data is not matched then it sent to the processing layer for diagnosis and sent to the physician. Finally, the physician results and

patient records are gathered in the storage layer for clinical study and for virtual treatment process.The following are the significances of this proposed server:
♦ Useful for patients in terms of fast processing particularly in emergency situation
♦ The records are analysed by worldwide due to its distributed sharing
♦ Beneficial for both physicians and patients in terms of cost
♦ Avoiding the data collection and transferring repetition process which saves the analysis time and improves the performance
♦ Easily accessible
♦ Lower-level structure with limited maintenance cost
♦ Used for clinical study purpose or for future references

Detailed Description of Drawings
Figure 1 illustrates the overview structure of proposed CB-LCSD framework. It contains the following blocks such as: an Automated 3D-Lung Cancer Detector Prototype called "Cloud-based 3D-Lung Cancer Segmentation and Detection (CB-LCSD)" that include proposed ACM segmentation and ECNN detection model with cloud-prototype design.A two-stage classifier developed for detecting the lung-tumor severity level at its early state.The proposed framework performs both segmentation and detection technique using pre-processed segmented images.Patients Report are automatically stored in the LC-Cloud server for clinical study and E-treatment.
Figure 2 and 3 shows the resultant segmentation image of PET scan.Remove the region of mediastinum and thoracic wall if the lung parenchyma was not reconstructed .The tumor part of the lung picture is segmented using the Active Contour Model (ACM). A total of 925 lesions from different patients were collected, together with nodule information.
Figure 4 depicts the architecture of ECNN. E-CNN is the multilayer architecture which consists . of 6 layers of CNN.E-CNN estimates the exact or neighboring classes of images with less than 5% error. The layers of E-CNN are 3 convolutional layers, followed by fully connected layers which has a Rectified Linear Unit (Relu), normalization and dropout layers.The dataset was trained in E-CNN technique, while this validation data was helpful for fine-tuning parameter results.
Figure 5 illustrates the overview structure of proposed LC-Cloud Server framework.The LC-Cloud server architecture comprises of three layers called input, processing and storage layer which is connected with firewall and LC-server.The input layer designed to adopt any category of lung tumor data such as dicomimages and PET images, directly accessible through users or radiologists through internet. These data are feed-forwarded into next layer for processing under certain per-defined conditions.The new user data is checked with the LC-Cloud engine for availability match. The conditions are (i) if the new user record is matched with existing file then directly forwarded to physicians and (ii) if the data is not matched then it sent to the processing layer for diagnosis and sent to the physician. Finally, the physician results and patient records are gathered in the storage layer for clinical study and for virtual treatment process.

Figure 6 shows the segmented results. Using LIDC-IDRI dataset, we segmented the tumor portion.The tumor part of the lung picture is segmented using the proposed Active Contour Model (ACM).The proposed segmentation model is used to integrate the local image bias field formulation with ACM. A local energy term is specified by using the mean squared error to reconcile severely in homogeneous CT images and used to detect and segment tumor regions efficiently with intensity inhomogeneity. In addition, a Multiscale Gaussian distribution is applied to the CT images for smoothening the evolution process for determing the features.
Figure 7 shows the detection results of E-CNN model. Our proposed enhanced CNN model has a sensitivity of 89% and 97% with respect to 1.7 FPs/image and 3.8 FPs/images in the LIDC-IDRI data set. A sensitivity of 94% with 2.5 FPs/images and specificity of 91%was determined in the 5-fold cross-validation tests. The proposed model shows that the developed CAD system has a high potential for automatic diagnosis of lung nodules.
Figure 8 shows that the proposed E-CNN uses nodules graphs and also a comparison with other CNN methods. In addition, a multi-scale Gaussian distribution is applied to the CT images for smoothening the evolution process, and features were determined. In this paper, we introduced three CNN networks, and an enhanced CNN model was used for detecting lung nodules in CT images.

Different Embodiment of Invention
i. This work focused on developing a cloud-based lung tumour detector that contains a segmentation module, detection module, and stage classifiers in order to reduce lung tumour death rates.
ii. To more precisely detect the tumour, the Cloud-LTDSC module proposed uses unsupervised learning neural networks as a predictor and stage classifier,
iii. The segmented pictures are sent into the M-CNN model, which determines the patient's classification and severity level.
iv. The e-record contains the patient's name, severity level, and tumour information, which are immediately transmitted to clinicians through cloud for virtual monitoring and E-diagnosis.

Application of Invention
a. These deep learning and smart lung tumour models use excellent virtual monitoring and
E-diagnosis to detect the tumour at an earlier stage.
b. The analyse of this suggested system for huge databases in future work and then publish
the findings on how the system is capable of properly managing hospital enormous data.
This prototype is also planned to evolve into a complete cloud hardware module in future
work.
c. CAD systems for lung cancer diagnosis are designed to aid the radiologist, in the process
of nodule detection by giving a reference opinion.
d. From the results of the performance comparisons, it is clear that our suggested model
3DDCNN outperformed other state-of-the-art systems in terms of sensitivity and FPs per
scan.
e. Although 3DDCNN's measured performance metric is pretty high, it might be enhanced
further. Because the performance was less accurate in recognising micro nodules, future
research will focus on detecting micro nodules with diameters smaller than 3 mm.