JMPI presentation

SCHOOL OF INFORMATICS
DEPARTMENT OF
INFORMATION TECHNOLOGY

SNo

1.
Full Name Id Number
Pgw/87922/15
Title review of: four articles on
Matusal Debebe
deep learning)

Submitted to D/r Siraje S.(PHD )

Information Management System
 Outlines
 Introduction
 General objective
 Specific objective
 Title and authors
 Method
 Finding
 Limitation
 Comparison of Existing Approaches
 Future work
 Recommendation and conclusion IMS
 1. Introduction

 Lung disease refers to many disorders affecting the lung such as asthma,
Chronic Obstructive Pulmonary Disease (COPD), tuberculosis, influenza, lung
cancer, pneumonia and other breathing problems. Lung diseases are respiratory
diseases that affect the various Organs and tissues associated with breathing.
 The papers discusses the growing importance of lung disease detection and
characterization, as medical imaging Datasets are expanding rapidly.
 Common signs are; trouble in breathing, shortness of breath, feeling like not
getting enough air, decreased ability to exercise, a cough that won't gon away,
coughing up blood or mucus, and pain or discomfort when breathing in or out.
Such diseases are caused by infection.
1 IMS
 [Link] ……..
 Lung diseases, including asthma, tuberculosis, influenza, lung cancer, and
pneumonia, are common Medical conditions caused by smoking, infections, and
genes.
 To address this, the researchers propose using deep learning techniques, which
have shown promising results in image classification tasks.
 The researchers implemented three deep learning models - Sequential,
Functional, and Transfer - and trained them on open-source lung disease image
datasets.
 The paper emphasizes the need for effective predictive technologies to diagnose
lung diseases early, highlighting the use of computer vision and deep Learning
techniques. IMS
 Objective

General objective
 The general objective of this papers is to detect and classify various lung
diseases such as pneumonia, tuberculosis, and lung cancer from standard X-ray
images and Computerized Tomography (CT) scan images with the help of
volume datasets
Specific objective
 To implement dynamic feature extraction to improve accuracy of lung disease
detection. To early detection and diagnosis of lung diseases
 Seeking effective solutions for automated processing of medical images
3
N TITLE AND METHODE FINDING LIMITATION
o AUTHORS
1  Title:-Detection  VGG (visual geometry  Performance of VDSNet.  Dataset limitations.
of lung disease group)network Using the full chest X-ray  Generalization to other lung
from x-ray image  Information enrichment dataset, the VDSNet model diseases.
using deep Spatiotemporal Transducer achieves a predictive
performance of 83%
 Lack of external validation.
learning frame System (STS) + CNN
work by  Potential Impact -The VDSNet architecture,  Limited model interpretability.
employing which combines VGG,
VDSnet information enrichment, and
 Authors:- the spatiotemporal transducer
[Link] and system (STS) with CNN,
[Link] demonstrates superior
2  Title:-Lung  Sequential Model performance compared to
disease detection  Functional Model existing methods
using various  Transfer Learning
deep learning Model
algorithm
 Authors:-M
JasminePemeena  Limited dataset size
Priyadarsini,Keta  Pneumonia detection.  Lack of clinical validation.
[Link],  Tuberculosis detection.  Single-modality input.
[Link]  Lung cancer detection  Interpretability limitations
Raj,[Link]  Computational complexity
.
4
N TITLE&AUTHORS METHOD FINDING LIMITATION
O
3  Title:-Lung disease detection  Dataset  The comparative .
using deep learning.  Model Architecture analysis done by the  Lack of detailed
 Authors:-SiddhanthTripathi,  Evaluation metrics authors demonstrated evaluation
Sinchana Shetty,Somil Jain, the effectiveness of metrics
Vanshika Sharma their proposed deep  Lack of clinical
learning approach insight.
compared to other
techniques.

4  Title :-Recent Study of Lung  Deep convolutional  Lack of Detailed

Disease Detection Using neural network(DCNN)  The paper does not Methodology
Deep Learning Techniques  Model architecture provide a clear and The document does not
 Authors :-Urvashi B.  Comparison other detailed summary of go into the specifics of
Deshmukh, Apurva S. methods the key findings from the methods and
experimental setups
Solanke, and Prapti D. the study. used in the reviewed
Deshmukh from Dr. G. Y  The information about studies..
the findings is limited  Limitedevaluation
in the paper. However Metrics:
based on limited The document only
information provided: briefly mentions the
use of classification
 Dataset limitation accuracy as a
5  Model limitation performance metric,
No DISEASE DETECTION RESULT CHALLENGE EVALUATION
1  The results demonstrate that the proposed  Dataset limitations Performance metrics.
VDSNet model outperforms other state-of-  Model design and Comparison with state-of-the-art
the-art methods in detecting lung diseases architecture methods
from chest X-ray images, particularly  Interpretability and explain Dataset split and cross-validation
when using the full dataset. ability Ablation studies or sensitivity
analysis
2  According to the results reported in the
paper, the deep learning models  Image variability  Performance metrics
demonstrated strong performance in  Class imbalance  Cross-validation
detecting and classifying various lung  Computational complexity
 Comparison to existing
diseases: methods
 Ablation studies
 The abstract only gives a high-level  The Semantic Gap
3 summary of the proposed methodology  Handling Complex Lung
and the promising overall classification  Disease Categorization  Evaluation metrics
accuracy achieved.  Per-class Performance
Analysis
 Potential Additional
 The model achieved a good overall  Dataset limitations Evaluation
4 multiclass classification accuracy on the  Model design and
Chest X-Ray 14 dataset, though the exact architecture  The researchers plan to
percentage is not provided.  Interpretability and explain provide a "comparative
 The authors state that the model "gives a ability analysis of the model's per-
good result in multiclass" classification of class performance" to
 Feature works

ARTICLE 1:-
 As mentioned in the previous responses, the paper lacks crucial details that would
help fully evaluate the VDSNet model's capabilities and potential real-world
impact. Addressing the identified gaps, such as providing more information about
the dataset, conducting comprehensive evaluations, and validating the model in
clinical settings, would be essential to strengthen the overall contribution of this
work.
ARTICLE 2:-
 The authors emphasize the need for stronger collaboration between researchers,
clinicians, and medical experts to bridge the gap between the technical
advancements and the practical clinical application of deep learning-based lung
disease detection solutions. 7
 Cont.…..

ARTICLE 3:-
 This paper aims to develop an automated lung disease detection system using deep
learning techniques applied to chest X-ray images and other patient data. Based on the
information provided in the introduction, the researchers seem to be aware of a key
challenge in developing their deep learning-based lung disease detection system,
which is referred to as the "semantic gap“.
ARTICLE 4:-
 The document provides a review of recent studies on using deep learning and other
techniques for automated detection and classification of lung diseases from medical
images. The document does not mention anything about the datasets used in the
reviewed studies, such as their size, diversity, or quality. 8
Recommendation to future improvement
 As my opinion the research could explore applying VDSNet to other medical
imaging modalities or expanding the lung disease classification capabilities
 To make it better Conduct further research on automated processing and
segmentation of lung regions in medical images to facilitate timely and accurate
disease diagnosis.
 Validate the performance of the proposed deep learning techniques on larger, diverse
datasets to ensure robustness and generalizability.
 There is a need for more advanced deep learning architectures that can effectively
capture and learn relevant features from chest X-ray images for improved lung
disease classification. I would recommend the paper for publication in a reputable
journal in the healthcare engineering or medical imaging
IMS domain. 9
 [Link]
 Generally, lung diseases are respiratory diseases that affect the various organs and
tissues associated with breathing, leading to airway diseases, lung tissue diseases, and
lung circulation diseases.
 The paper emphasizes the need for effective Predictive technologies to diagnose lung
diseases early, highlighting the use of computer vision and deep Learning techniques,
so deep learning has emerged as a powerful tool for automated lung disease detection
from medical images.
 Additionally, the document provides a valuable overview of the current state of deep
learning-based lung disease detection, highlighting the promising results achieved and
the need for continued research and development in this important field of medical
imaging analysis.
 10 IMS
THANK YOU
SO MUCH.

IMS