A prototype breast cancer diagnosis aid that analyzes radiology images with YOLO and lab test data with Random Forest, with input via images, Excel files, or manual entry.
An academic project carried out by a team of students with technical assistance from Techno Enjaz to develop an intelligent system to support breast cancer diagnosis through two complementary tracks: a visual track that analyzes breast radiology images with the YOLO algorithm to detect visual indicators of a tumor, and a numerical track that analyzes laboratory test data with the Random Forest algorithm to produce a prediction of the case.
The system was built in Python with the Ultralytics, TensorFlow, and scikit-learn libraries and a Tkinter graphical interface that accepts input in three ways: uploading a radiology image, uploading an Excel file of test results, or entering values manually, and then saves the predicted result to a new file. The project describes the system's performance as accurate and fast but reports no figures or test dataset, so the current version is presented as a prototype assistive tool for physicians, not an approved clinical diagnostic tool.
| Item | Details |
|---|---|
| Project type | AI-assisted breast cancer diagnosis system |
| Field | Computer vision, machine learning, medical informatics |
| Project status | Prototype with a working desktop graphical interface |
| Techno Enjaz role | Technical support and assistance to the students during project development |
| Core technologies | Python, YOLO (Ultralytics), Random Forest, scikit-learn, TensorFlow, Tkinter |
| Outputs | Tumor detection result on the radiology image, case prediction from lab data, Excel file with the result |
Breast cancer is among the most common and dangerous diseases affecting women's health, and early detection plays a central role in improving the chances of treatment. Conventional diagnosis, however, relies heavily on breast radiography (mammography) and on radiologists' expertise in reading the images, a process that is complex, costly, and prone to human error that can lead to false positives or false negatives.
From its review of earlier studies, the project also notes that AI models in this field face challenges such as limited diversity in training data, the difficulty of explaining deep models' decisions (the "black box" problem), and the need for large amounts of data and computing resources. This led to the idea of a system that brings together visual data (radiology) and numerical data (lab tests) to give a fuller picture of the patient's condition.
The project was carried out academically by the students, with technical assistance and support from Techno Enjaz during development. It is presented among the office's work as an example of its support for projects that apply computer vision and machine learning in medical applications.
The process starts at the main interface, where the user chooses one of two tracks:
| Interface | Function |
|---|---|
| Main interface | Starting point for choosing the image track or the lab data track |
| Image upload interface (YOLO) | Select a radiology image from the device and send it for analysis |
| Detection result window | Show the image with the tumor detection result |
| Data entry interface | Enter values manually, upload an Excel file, or attach a lab report image |
| Prediction interface | Show the predicted result based on the entered data |
| Results file | Excel file containing the patient's data and the predicted result |
| Tool | Role in the project |
|---|---|
| Ultralytics | Building, training, and running YOLO models in the image track |
| scikit-learn | Classification algorithms and data processing in the lab track |
| TensorFlow | Deep learning library within the software environment |
| Tkinter | Building the graphical interfaces and windows |
| VS Code and Jupyter Notebook | Environments for writing and running code and analyzing data |
The theoretical part of the project reviews image classification models such as ResNet50V2, ResNet152V2, InceptionV3, Xception, and MobileNet, along with decision trees and the SVC algorithm, as scientific background. The two tracks implemented in the system rely on YOLO and Random Forest.
The project concludes that the system performed well in supporting breast cancer diagnosis, that processing of images and data was fast compared with conventional methods, and that the multiple input methods (images, Excel files, manual entry) made the system flexible and easy to use. It also states that combining the two tracks raises accuracy and reliability compared with using either one alone.
However, the book reports no quantitative values such as accuracy, sensitivity, mAP, or processing times, does not identify the training and test datasets or their sizes, and does not describe a specific mechanism that merges the outputs of the two tracks into one decision; in the interfaces, the two tracks appear as separate paths that the user chooses between. The published results therefore reflect the functional success of a prototype, not a clinical evaluation. Figures quoted in the literature review chapter belong to other studies.
According to the recommendations set out in the project, the system could be developed by:
The system handles sensitive radiology images and laboratory data, so any real-world use requires patient consent, clear controls over access to the Excel files and saved results and how long they are kept, and an independent clinical evaluation before relying on it.
If you are working on a deep-learning medical diagnosis system and need technical help developing the idea or building a prototype, you can contact the Techno Enjaz office to discuss your project's requirements and the appropriate scope of support.
It is a prototype assistive system for breast cancer diagnosis built by a team of students with technical assistance from Techno Enjaz. It works through two tracks: analyzing radiology images with the YOLO algorithm to detect visual indicators of a tumor, and analyzing laboratory test data with the Random Forest algorithm to predict the case.
The system loads the prediction model, then receives test data entered manually, from an Excel file, or from an image of a lab report. The data is cleaned of missing or invalid values and passed to the Random Forest model, and the result is shown in a dedicated window and saved to a new Excel file containing the patient's data and the predicted result.
The system was built in Python. It uses the Ultralytics library for YOLO models in the image track, scikit-learn for the Random Forest algorithm and data processing, plus TensorFlow, and Tkinter for the graphical interfaces. VS Code and Jupyter Notebook were used as development environments.
The project describes the system's performance as good and fast, but reports no quantitative values such as accuracy, sensitivity, or mAP, and does not identify the training and test datasets. No specific accuracy figure is therefore attributed to the system; the results represent the functional success of a prototype, not a clinical evaluation.
No. The system is an assistive tool and has not yet been tested in a real working environment inside a medical institution, which the project proposes as a next step. The final diagnosis remains with the radiologist and confirmatory tests, and any real-world use requires an independent clinical evaluation and safeguards for patient data.
Reach out to Techno Enjaz engineering team to discuss technical feasibility, prototype hardware, and development scope.
Discover more engineering prototypes and academic implementations
A virtual board for writing, drawing, picking colors and erasing with hand gestures in front of a camera, built in Python with MediaPipe hand tracking and OpenCV.
A prototype that detects people with YOLOv8 and tracks them with DeepSORT in defined work areas, counts presence time in seconds, and exports records to Excel.
An exam-monitoring prototype using YOLO, MediaPipe and OpenCV to analyze a student's head direction, flagging head turns in red for review and forward gaze in green.
