Page 71 - AIH-1-3

Artificial Intelligence in Health Rotational thermography for breast cancer screening

breast tumors and elevated temperatures, prompting However, one of the primary challenges lies in establishing
consideration of IR imaging as a valuable tool for breast precise temperature thresholds to differentiate malignant
cancer detection. tumors amidst variations in individual heat sources due to
diverse medical and physical conditions. Addressing this
In recent years, the global rise in breast cancer incidence
has spurred extensive research into IR imaging for breast challenge is crucial in developing effective breast cancer
screening systems. This study presents a novel, non-contact,
cancer diagnosis. Prior research has explored thermal non-invasive breast imaging method capable of capturing
imaging for various medical applications, including comprehensive abnormalities. Several obstacles confront
diabetic foot disease, ocular surface temperature analysis,
and tumor detection Click or tap here to enter text. researchers in the development of IR imaging-based breast
1-6
cancer screening systems, including the surge in breast
Click or tap here to enter text. Click or tap here to enter cancer cases, the cost and invasiveness of current screening
text. Click or tap here to enter text. Click or tap here to modalities, challenges in tracking malignancy progression,
enter text. Studies investigating breast cancer detection poor visibility of affected anatomical areas, patient
through thermal imaging have examined bi-spectral discomfort during examinations, and shortages of skilled
invariant features, color segmentation techniques, and medical professionals. Table 1 summarizes the earlier works
deep learning-based segmentation methods 7-14 Click or tap and comparison with the approach used in this study. This
here to enter text. Click or tap here to enter text. Click or research endeavors to overcome these challenges through
tap here to enter text. Click or tap here to enter text. Click an efficient data collection process and systematic stages of
or tap here to enter text. Click or tap here to enter text. system development for IR image processing techniques.
Click or tap here to enter text. A novel method combining Conducted at a prominent hospital in northeast India, the
rotational thermography and dynamic temperature study comprises four research phases: Three pilot studies
analysis has been developed to address challenges in (Phase 1 [PS1], Phase 2 [PS2], and Phase 3 [PS3]) and a final
breast cancer screening 15-17 Click or tap here to enter text. study (FS). The study evolves to optimize imaging quality
Click or tap here to enter text. This non-contact, non- by transitioning from low-resolution forward-looking IR
invasive approach enhances the visibility of abnormalities (FLIR) SC-325 cameras in the initial phases to Infratec HD
in breast tissue, improving detection accuracy. Machine 600 and FLIR T-650 cameras in subsequent phases. The
learning algorithms trained on extracted features aid in study aimed to develop an efficient and accurate thermal
distinguishing normal from abnormal patterns. Earlier, imaging system for breast cancer screening. It progressed
EtehadTavakol et al. achieved high detection rates using through four phases, refining data collection methods and
7,10
bi-spectral invariant features and K-means clustering for improving imaging results.
segmentation. Garduño-Ramón et al. introduced
7,10
11
a non-invasive tool utilizing temperature and texture In PS1 and PS2, conventional imaging approaches were
features, yielding promising results Various segmentation employed, but challenges such as varied focal lengths and
11
techniques, including K-means, fuzzy c-means (FCM), and patient movement resulted in unsatisfactory images. PS3
expectation-maximization (EM) algorithms, have been introduced a semi-circular arc for camera movement,
explored, with the EM algorithm demonstrating superior allowing for precise focus on one breast at a time, though
accuracy 11,12 Click or tap here to enter text. Venkataramani manual adjustments and open environments presented
et al. 10,13,18-22 proposed a semi-automated method using difficulties. Findings from PS3 and FS underscore the
morphological filtering and thresholding, achieving high necessity of mounting the IR camera on a motorized
sensitivity and specificity. 10,13,18-22 Deep learning-based mechanical arm capable of rotational thermography,
approaches, such as the level-set method, have shown enabling comprehensive imaging coverage. In the final
high accuracy in segmenting suspicious regions in breast phase (FS), a temperature-controlled chamber and
thermograms. 9,23,24 Thermal imaging has also been utilized automated hardware and software provided a standardized,
touchless, and painless imaging process with an accuracy
for brain tumor detection, highlighting its versatility in rate of 93.18% for detecting abnormalities. Data analysis
medical diagnostics. 13,20,25-27
evolved from conventional methods to advanced
While IR imaging shows promise for breast cancer techniques in PS3 and FS, employing clustering methods
screening, challenges remain, including standardization and references from ultrasonography (USG) and biopsy
of temperature values and the need for trained personnel. reports. Mean temperature and standard deviation analyses
IR Imaging researches also focus on optimizing machine further enhance detection precision. The study included
learning algorithms, integrating advanced imaging diverse breast cancer subtypes and stages, demonstrating
techniques, and exploring novel optimization algorithms the system’s applicability across various scenarios. Control
for enhanced diagnosis and classification. 28-34 groups provided benchmarks for assessing accuracy and

Volume 1 Issue 3 (2024) 65 doi: 10.36922/aih.3312