Alzheimer's disease (AD), a leading type of dementia, is a prevalent neurodegenerative

condition amongst the older population. Though incurable, treatment options for AD are

available, thus, recognizing its presence early on is beneficial. However, despite its effectiveness,

getting a diagnosis via brain scan can be costly and time-consuming. In our research, we focused

on using cognitive tests, a quick and inexpensive option that can be administered by primary care

physicians, to accurately diagnose AD. The National Alzheimer’s Coordinating Center (NACC)

dataset contained the scores of 2,700 test subjects taking six different cognitive tests and their

diagnosis of either healthy (0), mild cognitive impairment (MCI) (1), or dementia due to AD (2).

We constructed three box plots for each test, separated by diagnosis level, and altered our dataset

so that diagnosis was a binary variable with value 0 for healthy subjects and value 1 for subjects

with MCI or dementia due to AD. To measure the accuracy of our model through

cross-validation, we split our modified dataset into training (80%) and testing (20%) and created

logistic regression models for diagnosis, with various arrangements of predictors, eliminating

irrelevant predictors (p-value > 0.05). The models that were used included multiple simple

logistic regression models, a logistic regression model with multi-layered neural network (NN),

and a Random Forest model. We found that the simple logistic regression model worked the best

and fitting logistic regression models made from a dataset by the NACC, the combination of four

different cognitive tests out of the six we studied—the Animals Test, the Trail Making Test Part

B, the Logical Memory Test, and the Mini-Mental State Examination (MMSE)—produced the

best result, with an 85% accuracy. This research offers primary care physicians a prompt and

efficient method for the early diagnosis of AD in the elderly, enabling timely initiation of

symptom management.

In the context of bone fracture detection, the research study titled “Revolutionizing Bone Fracture Detection: YOLOv5 vs. YOLOv8 Face-off” comprehensively assessed the performance of two state of-the-art YOLO (You Only Look Once) object detection algorithms, namely YOLOv5 and YOLOv8, across different model configurations, including small (s), medium (m), and large (l) variants. These six YOLO models underwent rigorous training, validation, and testing phases, employing three distinct datasets. One dataset exclusively consisted of wrist fractures, while the remaining two contained a mix of bone fracture types. The evaluation of model performance was based on precision, F1 score, and recall confidence ratings, and the results were presented through various visualization methods, including tables, images, and graphs. Ultimately, the study established that YOLOv5 and YOLOv8 demonstrate comparable detection capabilities, with YOLOv5 occasionally exhibiting superior accuracy in detecting fractures of varying sizes.