| Kundu et al. [206]
|
Monkeypox detection to restrict the spread of the virus |
Vision transformer and conventional ML based approaches. |
Insufficient Dataset of the monkeypox disease |
To overcome the challenge of the shortage of data GAN can be applied for producing simulated data. |
| Rahman et al. [207]
|
Patient Length of stay (LOS) Prediction for assisting healthcare professionals in finding a proper treatment plan. |
Federated learning with a linear regression model. |
Only linear models have been used. |
The model can be tested for other datasets as well. |
| Solanki et al. [209]
|
Development of virtual assistant for healthcare sector |
Machine earning based chatbot for medical care system. |
To gain patients' trust |
Provide the ethical part of the system to gain trust of users. |
| Chen et al. [210]
|
Leveraging DL for the detection of pancreatic cancer without human intervention |
Convolutional neural netwrok with integrated CAD based methods. |
A CT scan of the abdomen misses around forty percent of pancreatic cancers that are less than 2 cm. |
Collection of real-time data is a challenging task for many countries. |
| Siar et al. [211]
|
Brain tumour detection through deep learning model. |
CNN, and as a classifier softmax classifier has been used. |
Data collection as it may include sensitive information |
As it includes healthcare data it requires the security measures needs further consideration. |
| Awotunde et al. [212]
|
Breast cancer detection. |
Hybrid rule-driven decision-making technique to find five significant insights. |
Patients' data privacy is not ensured. |
To collaborate with a large number of patient data from patient all around the world. |
| Özdil et al. [213]
|
Fatty liver classification |
CNN with texture-based feature selection method |
The investigation of feature selection using texture is less thorough. |
Pre-processing task of thermal images needs further improvements. |
| Qadri et al. [214]
|
An automated spine segmentation method through deep learning model |
Steps involves preprocessing, regression with sigmoid post processing |
To identify fractured vertebrae. |
Model needs further improvement to identify some cases for example fractured vertebrae. |
| Chieregato et al. [169]
|
Using CNN and CatBoost classifiers, a model that can assist clinical decision-making completely combines imaging and non-imaging data on top of this networking approach. |
CNN and catBoost classifier for clinical decision-making tasks. |
Clinical decision-making depends on data, but collecting datasets of medical is difficult. |
Security needs further discussion, especially for healthcare-based systems. |
| Saheed et al. [173]
|
Using DRNN and SML models, an efficient and effective IDS for classifying and foreseeing unexpected cyberattacks in the IoMT environment was presented. |
DRNN and SML for cyberattack detection in the IoMT area. |
Test within the real-time environment is critical |
Further study for Testing within a real-time scenario. |
| Astha Parihar, Shweta Sharma [215]
|
using unlabeled data and substantial learning throughout the preparation. AI examines how the genomic landscape interacts across characteristics, whereas DL, MRI, and CT scans are used for illness identification and diagnosis. |
AI-based methods for disease diagnosis. |
Integration of AI in the healthcare sector requires a vast dataset. |
Explainability of such a model helps better comprehend the prediction model's result. |
| Bhardwaj et al. [216]
|
Explainable deep learning model for valvular heart disease classification |
Deep learning visualization method, CNN architecture for heart disease detection. |
Model explainability needs to get the trust of the end user. |
Proper training sessions to make the model trustworthy for the end user. |
| Tasin et al. [217]
|
A model for diabetes prediction. |
SHAP and LIME for model explainability and SMOTE analysis for restricting the class minimization in the area of heart disease detection |
dataset contains only 203 samples which might be a critical drawback for model explainability. |
Integration of different hospital datasets for diabetes tests may improve the data quality. |
| Tasnim et al. [218]
|
A explainable model for mortality prediction |
DL methods with integrated SHAP model for explainability of the decision provided by the model for prediction of mortality rate for heart failure based cases. |
Security analysis of the model is challenging. |
To make the model available for rural areas, the explainability of the model needs further improvement. |
| Nancy et al. [219]
|
Smart healthcare monitoring especially for heart disease cases. |
IoT with integrated Deep learning model for heart disease prediction |
IoT devices memory management is difficult as the devices have limited memory spaces. |
Methods that can erase the data saved on the device or a method that does not require saving data, such as Federated learning. |
