Page 62 - AIH-2-3

Artificial Intelligence in Health AI in medical diagnostics: A multi-disease approach

IoT-based systems, using algorithms such as random
forest, have been developed to monitor patient activities
and predict health conditions in real time. One example is
a hybrid IoT model that utilizes random forest techniques
to predict T2D, demonstrating high predictive capability
and aiding in the early intervention. AI-powered mobile
platforms for real-time disease monitoring further improve
patient care and support healthcare providers in managing
complex cases effectively.

8.4. Superior performance of DL in medical image
analysis

DL models, particularly through CNNs, have demonstrated
remarkable performance in medical image analysis. CNNs
have been effectively used to identify malaria-infected
Figure 4. The accuracy of AI techniques in disease diagnosis. The bar
chart illustrates the classification accuracy of two AI techniques: Deep blood cells, while other ML models, including Naïve
learning and machine learning. The accuracy percentage for each Bayes, SVM, and gradient boosting, have shown success in
technique is calculated as the ratio of correctly classified instances to the diagnosing various diseases.
total number of instances in the dataset.
Notes: Classification Accuracy: This metric measures how often the AI For instance, a DL model called LAPNet, using a pyramid-
model correctly predicts the class or category of a given data point. It’s based architecture and attention mechanisms, proved highly
calculated by dividing the number of correct predictions by the total effective in detecting and grading diabetic retinopathy from
number of predictions. Percentage Derivation: In the context of the medical images. Region aggregation graph convolutional
bar chart, the height of each bar represents the classification accuracy
percentage. For instance, if the Deep Learning bar reaches 80 on the y-axis, networks have also been applied in medical imaging tasks,
it means the model achieved an 80% accuracy in classifying the data. such as bone age estimation using CT and MRI scans, further
Abbreviations: AI: Artificial intelligence; DL: Deep learning; enhancing diagnostic accuracy in radiology.
ML: Machine learning.
The research findings illustrate that DL methods generally
outperform traditional ML models, especially when
working with large datasets and complex medical images.
DL techniques not only provide high diagnostic accuracy
but also offer frameworks for real-time patient monitoring
and tailored treatment options, paving the way for improved
patient outcomes and healthcare efficiency. Through
AI-driven medical diagnostics, healthcare providers can
achieve early detection, personalized treatments, and better
patient management, setting a new standard for care quality.
8.5. Challenges and future directions
While AI’s potential in healthcare is immense, the field
Figure 5. AI applications in medical imaging faces several challenges, including issues related to data
Abbreviations: AI: Artificial intelligence; CNN: Convolutional neural privacy, algorithmic bias, and the need for diverse and
networks; RAGCN: Region aggregation graph convolutional networks.
high-quality datasets. In addition, ethical and regulatory
forest algorithms demonstrated over 98% accuracy in challenges must be addressed to promote responsible AI
tuberculosis detection. RNN and LSTM networks further use in healthcare.
achieved approximately 97% accuracy in classifying Future research may focus on refining AI algorithms,
gastrointestinal (GI) diseases, underscoring AI’s potential improving multi-source data integration, and ensuring
in predictive diagnostics. that AI applications in healthcare are equitable, reliable,
and adaptable to a broad spectrum of healthcare settings.
8.3. Integrating AI with IoT for healthcare
advancements 9. Discussions and future directions

A promising area of research combines AI with the IoT The findings from this exploration underline the
to enhance patient monitoring and disease prediction. transformative impact of AI on medical diagnostics,

Volume 2 Issue 3 (2025) 56 doi: 10.36922/aih.5173