PathInsightFlow: A Predictive Model for Rare Disease Diagnosis Using Advanced Image Processing Techniques
Bhavna Bajpai, Alok Dubey, Sheetal Mujoo, Ranjana Ganesh Khade, Ch. Varaha Narasimha Raja, Sumanpreet Kaur · 2025
The CT scan image is very useful in the diagnosis and identification of Atypical Hemolytic Uremic Syndrome (AHUS) in the kidneys. For many medical purposes, it can furnish accurate details about the location and proportions of A-HUS. Conventional and manual medical testing is labor and time intensive. Automatic detection of A-HUS on CT is currently clinically very important as the definitive diagnosis. The existing CAD technique is very important to enhance the efficacy of ballooning of medical testing. However, this is still an important challenge to overcome the existing poor precision and partial detection method. In this paper, we present a way of finding A-HUS with the use of morphological picture enhancement along with Break Neck Pace Modified Convolutional Neural Networks (BNP MCNNs) and multi intersection over union (IoU) threshold. In order to make AHUS (1-3 mm) easy to find and to achieve more stable network, we made four convolutional layers of MCNN network, and we make four IOU threshold cascade MCNNs based on the feature pyramid networks (FPEs). For the A HUS detection problem, the updated CNN was trained within the MATLAB 2022b Tool framework. Once the model is trained, and with the help of the assessment criteria of classifying methods, the model is 99 % accurate on the test dataset. This implies that it can be applied to predict the result of diseases. However, as this machine learning framework was very successful in predicting CKD, the operation of this result to professionals to receive elucidation is dense.