Page 80 - AIH-1-3

Artificial Intelligence in Health Rotational thermography for breast cancer screening

Figure 8. Neural network classification results of the training and testing datasets for breast abnormality detection for PS3 (33 subjects). The results
illustrate model tuning and performance evaluation. The image was created using Matlab software.
Abbreviations: CE: Cross-entropy; %E: Percentage of correctly classified elements.

for population-based case–control studies in PS3 and FS, 4. Discussion
respectively. These matrices offer a visual representation
of the classification performance, aiding in assessing the This system has been installed at a renowned hospital
system’s accuracy and reliability. in North-east India, known for its mass screening
capabilities. The subsequent product deployment will
The developed system’s exceptional accuracy for include installations at various hospitals across India,
screening breast abnormalities and detecting malignant leveraging the system’s superior performance and excellent
tumors was validated at 93.18%, underscoring its reliability output based on the second dataset acquired through our
and effectiveness. proposed IR image acquisition and analysis technique.
Finally, Table 4 provides a comparative analysis of The study utilized a double-blind validation method
studies conducted in population-based case–control where expert doctors and reviewers provided both
settings, elucidating the progression and refinement of quantitative and qualitative feedback. This approach
the system across different phases. This comprehensive ensured an impartial evaluation of the model’s
comparison offers insights into the system’s evolution and performance, as the experts and reviewers were unaware
performance enhancements. of the algorithm’s predictions during the assessment. The

Volume 1 Issue 3 (2024) 74 doi: 10.36922/aih.3312