African Walnut Multimodal Prostate Progression Model

Repo: Mikecode123/walnut

This repository contains models developed for an interpretable multimodal computational model of prostate cancer progression, integrating:

• Histopathology (H&E)
• Immunohistochemistry (IHC: p63, 34βE12, AMACR)
• Molecular / Laboratory features (phytochemical profile of Plukenetia conophora seed oil)

Biological States

state_id	State Name	Description
0	Normal	Intact architecture, strong basal markers, low AMACR
1	Partial_Restoration	Architecture largely preserved, basal markers recovering
2	Induced_Neoplastic	Distorted architecture, high AMACR, loss of basal integrity

Models Included

File	Description
best_densenet121_image_model.pt	DenseNet121 encoders for H&E + IHC
best_molecular_mlp.pt	MLP on quantitative phytochemicals
best_fusion_model.pt	Fusion of image + molecular features

Feature Details (Molecular)

• Flavonoid: 14.30 mg/ml
• Alkaloid: 4.07 mg/ml
• Terpenoid: 2.57 mg/ml
• Phenol: 1.80 mg/ml
• Steroid: 0.22 mg/ml

These values come from the experimental characterization of African walnut seed oil.

How to Use

from huggingface_hub import hf_hub_download
import torch

# Download model
path = hf_hub_download(repo_id="Mikecode123/walnut", filename="best_molecular_mlp.pt")